SlideShare a Scribd company logo
IMPERIAL COLLEGE LONDON
The impacts of data on the online advertising industry
by
Mutlu Dogus Yildirim
A Project submitted in partial fulfilment of the
requirements for the MBA degree
July 2016
II
SYNOPSIS
Data is a game-changer in the online advertising industry. Companies that
did not exist several years ago generate more revenues than The New York
Times now. WPP, the largest advertising agency in the world, wants to be
known as a “data” company. Google and Facebook generate their incomes
mostly through online advertising by utilising data, and they are among the
most valuable companies in the world now. In Germany, rival publishers are
about to start a “data alliance”. Meanwhile, governments try to regulate the
usage of data, and many people use ad-blockers to protect their privacies.
Still, many advertising companies track users collectively, and companies
like MasterCard sell user data to advertisers.
One of the goals of this research is to investigate how data is used in online
advertising. One of the main findings is that companies use many techniques
to track users and collect many types of data like search, behavioural or
location data, then this data is used for purposes like targeting,
personalisation, measurement, optimisation and attribution.
Another goal of this research is to address the benefits and risks for the
players in the ecosystem. One important conclusion is that there are indeed
many benefits as well as risks for publishers, advertisers, agencies,
technology providers and users, and all parties should analyse these benefits
and risks thoroughly.
The final goal is to assess impacts of data on the online advertising industry.
Some of the findings show that data has a huge impact in the industry, from
creating the most valuable companies in the world, to data alliances among
rival companies, to legislations and ad-blockers which may force the
advertising companies to change their business models.
III
ACKNOWLEDGEMENTS
I would like to thank my supervisor Dr John Beasley for guiding me with his
valuable feedbacks. I would like to thank my associate Mr Alp Ayhan for his
great help.
Finally, I would like to express my deepest appreciation to my wife Maria for
all of her support. Without her, I could never finish this project.
IV
LIST OF CONTENT
Page
1 INTRODUCTION 1
1.1 BACKGROUND 1
1.1.1 What Does Data Mean for Advertising? 1
1.1.2 Is Data Really Important? 2
1.2 LITERATURE REVIEW 5
1.3 OBJECTIVES 6
2 APPLICATIONS OF DATA IN THE ONLINE ADVERTISING INDUSTRY 7
2.1 TYPES OF DATA 7
2.1.1 Search 7
2.1.2 Contextual 7
2.1.3 Intent 8
2.1.4 Image & Video 9
2.1.5 Ethnic 9
2.1.6 Location 10
2.1.7 Social Media 10
2.1.8 Social Sharing 11
2.1.9 Business 11
2.1.10 CRM 12
2.1.11 Offline 12
2.1.12 Behavioural 13
2.2 CONNECTING TO USER 14
2.2.1 Cookies 15
2.2.2 Flash Cookies 15
2.2.3 IP Address 15
2.2.4 Browser Fingerprinting 16
2.2.5 Local Storage 16
2.2.6 Canvas Fingerprinting 16
2.2.7 Carrier Injected Headers 16
2.2.8 Evercookie 17
2.2.9 ISP-Level Tracking 17
2.2.10 Others 17
2.3 DATA QUALITY 17
2.3.1 One Device & Multiple Users 18
2.3.2 One User & Multiple Devices 18
2.3.3 Deleted Cookies 19
2.3.4 Data Loss 19
2.3.5 Recency 19
2.3.6 Missing Data Points 20
2.3.7 Poor Integration / Implementation 20
2.3.8 Technological Challenges 20
2.4 SEGMENTATION 21
2.5 USAGE OF DATA 21
2.5.1 Traditional Advertising 21
2.5.2 Programmatic Advertising 22
2.5.3 Targeted Media Buying 24
2.5.4 Campaign Measurement & Optimisation 28
2.5.5 Attribution 29
2.5.6 Yield Optimisation 30
V
2.5.7 Audience Optimisation 30
2.5.8 Data Sales 30
2.5.9 Frequency Optimisation 33
2.6 FRAUD DETECTION 33
2.7 VIEWABILITY 34
2.8 MOBILE-SPECIFIC CHALLENGES & BENEFITS 34
2.8.1 Publishers 35
2.8.2 Advertisers 35
2.8.3 Cookies 36
2.8.4 Other Tracking Methods 36
2.8.5 Benefits of Mobile Devices 37
3 BENEFITS AND RISKS FOR THE PLAYERS IN THE ECOSYSTEM 38
3.1 BENEFITS FOR PUBLISHERS 38
3.1.1 Data + Media Bundle Sales 39
3.1.2 Data Sales 39
3.1.3 Data Partnerships 40
3.1.4 Audience Enrichment 40
3.2 BENEFITS FOR ADVERTISERS 41
3.2.1 Targeting a Specific Audience 41
3.2.2 Banner Optimisation 42
3.2.3 Understanding Customers Better 42
3.2.4 Viewability 42
3.2.5 Fraud 43
3.2.6 Spending Budgets Efficiently 43
3.2.7 Brand Safety 43
3.2.8 Data Sales 44
3.3 BENEFITS FOR ADVERTISING AGENCIES 44
3.3.1 Optimising Media Investments 45
3.3.2 Tailoring Relevant Messages 45
3.3.3 ProvideAccountability and Knowledge to Marketers 45
3.4 BENEFITS FOR TECHNOLOGY PROVIDERS 45
3.4.1 Data at the Core 45
3.4.2 New Business Opportunities 46
3.5 BENEFITS FOR USERS 46
3.5.1 Free Content 47
3.5.2 Inevitability 48
3.5.3 Discounts & Special Offers 49
3.5.4 Additional Benefits 49
3.6 RISKS FOR PUBLISHERS 49
3.6.1 Privacy Concerns 49
3.6.2 From Publisher to Audience 49
3.6.3 Too Targeted = Too Low Spending? 50
3.6.4 Information Asymmetry 50
3.6.5 Data Leakage 50
3.6.6 Impact on Revenues 51
3.7 RISKS FOR ADVERTISERS 51
3.7.1 Being Annoying 51
3.7.2 Data Leakage 52
3.7.3 Increasing Complexity 52
3.7.4 Impacts on HR 53
3.7.5 Impacts on Advertising Costs 54
VI
3.8 RISKS FOR ADVERTISING AGENCIES 55
3.8.1 Increasing Complexity 55
3.8.2 Low Spending 55
3.8.3 Benefits(!) of Fraud 56
3.8.4 Impacts on HR and Mindsetsof Employees 56
3.8.5 Over-relianceon Technology and Data 56
3.9 RISKS FOR TECHNOLOGY PROVIDERS 56
3.9.1 Missing Data Points 57
3.9.2 Crowded Industry 57
3.10 RISKS FOR USERS 57
3.10.1 Privacy 57
3.10.2 Increased Spending 58
4 IMPACTS ON THE INDUSTRY 60
4.1 PRIVACY CONCERNS 60
4.1.1 Legislations 61
4.1.2 Targeting Children 62
4.1.3 Other Sensitive Data 62
4.1.4 Self-Regulation 63
4.1.5 Do-Not-Track 63
4.1.6 Ad-Blockers 64
4.2 MAJOR IMPACTS OF DATA ON THE INDUSTRY 64
4.2.1 Google and Facebook 65
4.2.2 Other Companies 66
4.2.3 Publisher Partnerships & Alliances 67
4.2.4 Telecommunication Companies 67
4.2.5 Credit Card Companies 68
4.2.6 Purely Data-Focused Players 68
4.2.7 Ad Blockers 69
4.2.8 Big Data Technologies 69
4.2.9 Mergers and Acquisitions 70
4.2.10 New Business Models 70
4.3 NEAR FUTURE 71
4.3.1 TV Advertising 71
4.3.2 Internet of Things (IOT) 72
4.3.3 Virtual/Augmented Reality 72
4.3.4 Outdoor & Radio Advertising 72
4.3.5 Cross-Device Attribution / Targeting 73
4.3.6 Converging Technologies 73
4.3.7 Improving Technologies & Decreasing Costs 73
4.3.8 Change in Employee Skills 74
4.3.9 Privacy Concerns & Legislations 74
5 CONCLUSIONS 75
6 FUTURE WORK 77
7 REFERENCES 78
1
1 INTRODUCTION
1.1 Background
In the last few years, people have created more data than in the entire
human history (Marr, 2015). This data is analysed and used in many different
ways in many industries, but for one of them, the advertising industry, it has a
special value.
John Wanamaker famously said, “Half my advertising is wasted, I just don't
know which half.” (AdvertisingAge, 1999). Data helps advertisers to
understand how their advertising budgets are spent, who has seen their ads
and how many times, did it result in any action and so on. This information
gives advertisers great insights regarding their advertising campaigns and
helps to reduce the money they “waste”.
In fact, the advertising industry, online advertising, in particular, seems to be
“have a crush” on data (Marshall, 2013a). It affected the industry so much,
that the largest advertising agency company in the world, WPP, does not
want to be known as an “advertising” company anymore, but instead wants
to be seen as a “data” company (Marshall, 2013b).
1.1.1 What Does Data Mean for Advertising?
In mid-2009, McKinsey suggested that consumers are changing the ways
they research and buy products. They suggested that one of the things
marketers should do is to “customise advertising by viewing the context and
the consumer” (Court et al., 2009). The good news is; current technologies
allow advertisers to do just that - by using data.
In the online advertising world, data can be collected from many sources like
what product the user is viewing on the advertiser’s website, what s/he reads
on a news publisher’s website, his/her location and so on. Mostly users
generate this data, so this research will mainly be concerned with user data.
2
However, there are other types of data sources as well, like the weather
condition, which can be useful to advertise an umbrella for instance. Hence,
some of these sources will also be discussed in this project.
It is possible to change the messages in real-time considering the audience
or context - as suggested by McKinsey, which is possible only through data.
It will be discussed later in more detail, but in the online advertising world,
data can especially be helpful with targeting, personalisation, measurement,
optimisation and attribution. By using data from various sources, it is possible
to show the ads to a specific audience like “women aged between 18-34” or
“men who searched for Audi cars recently”. The messages can even be
tailored at the individual level, and a 10% discount can be offered only to
those who visited advertiser’s website and spent more than 3 minutes
exploring a product for instance. Thus, the term “data-driven advertising”
became popular in the industry (GlobalDMA, 2014).
1.1.2 Is Data Really Important?
Two examples; Google and Facebook, prove the importance of data in online
advertising.
The famous search engine, Google, was born in 1998 (Google, n.d.g). In
2000, Google introduced AdWords, a self-service tool for creating online
advertising campaigns. In 2004, after going public, Google was valued at $23
billion (Ritter, 2014). Today, Google’s market capitalisation is $469.05 billion
(Yahoo, 2016a) and it is one of the most valuable brands in the world
(Forbes, 2016). Around 90% of Google’s revenue comes from advertising
(Statista, 2016a), which amounted to $67,39 billion in 2015. The fast growth
of Google in years can be seen in Figure 1.1.
3
Figure 1.1 Google advertising revenues (Statista, 2016b)
Google achieved this big success thanks to AdWords – a platform which
allows advertisers to reach people as they search for some specific words or
phrases and who are in specific geographical areas. It also enables
measurement and optimisation of campaigns in real-time (Google, n.d.a).
These capabilities, which are possible thanks to search and geographical
data of the users, are very precious to advertisers (Rampton, 2014).
Facebook is another good example of how data helped a company grow very
quickly and become one of the biggest companies in the world. It was
founded in 2004 (Facebook, n.d.a) and held its initial public offering (IPO) in
2012. During its IPO, it was valued at $104 billion, which was the largest ever
valuation of a U.S. company at the time of its offering. Today, Facebook’s
market capitalisation is $290.35 billion (Yahoo, 2016b).
Facebook’s success in advertising is related directly to data. Facebook
knows where its users live, what they like, who their friends are, their age
and so on, which are extremely valuable to the advertisers (Martin, 2012).
Hence, advertisers shifted much of their budgets to Facebook advertising
campaigns, and Facebook generated $17.93 billion revenue in 2015, from
which $17.08 billion comes from advertising, as can be seen in Figure 1.2
(Statista, 2016c).
4
Figure 1.2 Facebook advertising revenue (Statista, 2016c)
Facebook is expected to grow even further thanks to the continuously
advancing technologies in targeted advertising, which is helping them to be
way ahead of their competitors (DiChristopher, 2015).
Obviously, data is not the only reason why Google and Facebook became so
successful; there are many other factors too (e.g. product quality, the number
of users, etc.). However, the shift in advertising budgets so quickly and so
substantially can only be explained by data and its benefits to the
advertisers.
Data does not only help Google and Facebook; it also helps to the whole
online advertising industry and the players in it. In 2017, online advertising
spending is expected to surpass TV spending in the US for the first time in
history (Nicolaou, 2016). Comcast - a broadcasting and cable television
giant, is planning to give advertisers some targeting capabilities by utilising
data (Levy, 2016b), like their “purely digital” rivals do. Criteo, a company that
offers personalised advertising products by utilising data, has a market
capitalisation of $2.8 billion (Bloomberg, n.d.). Amazon.com generated
around $850 million from advertising in the US in 2015 (Statista, 2016e),
whereas The New York Times generated only $638 million (The New York
5
Times Company, 2015). Hence, traditional publishers around the world like
Financial Times, The Guardian and Reuters join their forces and create
partnerships to combine their data and offer advertisers better products
(Bilton, 2016). Meanwhile, companies like MasterCard or Visa sell data to
advertisers to allow them to target a specific audience with specific
purchasing behaviours.
As can be seen, data has been shaping an entire industry quite a bit. As
Economist puts it, data is crucial to the online advertising economy (Koen,
2014).
1.2 Literature Review
There are many studies on data and online advertising in the literature.
Evans (2009) discusses the economic benefits of online advertising, which
can be communicated to the most relevant people, and how it disrupts
traditional advertising methods. Fulgoni (2013) discusses different ways
marketers can take advantage of (big) data, where Chen et al. (2014)
discuss how to maximise revenues for publishers. Gironda (2014) tries to
analyse consumer attitudes towards personalised ads. Langheinrich et al.
(1999) show that the usage of data started at the early ages of the internet,
and they suggest ways of “customisation” of ads without being intrusive.
Englehardt et al. (2016) identify many different techniques of tracking users
and show how different companies share data with each other. Karlsson et
al. (2015) discuss why marketers should be transparent about how and why
they collect personal information. Goldfarb et al. (2010) analyse how privacy
regulations affect campaign effectiveness negatively – around 65% reduction
where EU privacy laws are applied for instance. Toubiana et al. (2010) come
up with a new approach, which allows behavioural tracking while respecting
user privacy.
6
Actually, behavioural advertising and privacy related issues are researched
quite well by many others (Mayer and Mitchell (2012), Ur et al. (2012),
Hoofnagle et al. (2012), Corniere and de Nijs (2013)).
1.3 Objectives
The purpose of this research is to extend the overall knowledge of data and
its usage in the online advertising industry, and mainly answer the questions
below:
1- How is data used in online advertising?
2- What are the benefits and risks of data for the major players in the online
advertising industry?
3- What is the impact of data to the online advertising industry?
7
2 APPLICATIONS OF DATA IN THE ONLINE ADVERTISING INDUSTRY
2.1 Types of Data
There are many types of data that are useful to advertisers. Some of them
are briefly explained below.
2.1.1 Search
One of the important data points for advertisers is search data, which shows
what the user is actively looking for. It is the type of data that made Google
so big and powerful, as Google shows ads to people who look for specific
information (e.g. audi a3 price), and this is much more efficient than showing
ads to people who are not interested in anything specific (Rampton, 2014).
Search data is not limited to queries made on search engines like Google.
People also search within websites. For instance, someone can visit
Amazon.com and search for a product. This data, along with others, can be
used by Amazon to show its visitors some personalised ads, and it is one of
the reasons why Amazon generates more revenue than The New York
Times from advertising business (Liyakasa, 2015).
The same logic can be applied to a newspaper or to any other website,
where the search queries are used to segment users and for ad optimisation
purposes. Chango, a company who specialised in this area, was acquired by
Rubicon Project for $122 million (Rubicon Project, 2015).
2.1.2 Contextual
Contextual data is about the content a user views (Zhang & Katona, 2011).
For instance, if someone visits Theguardian.com and starts reading an article
called “Zurich city guide”, this might indicate that s/he could potentially be
interested in visiting Zurich. Therefore, it may be a good idea to show this
person a Zurich hotel ad.
8
Contextual data can also be used to segment users based on the content
they consume. For instance, people who read articles about “Samsung
smartphones” can be grouped into a segment called “Samsung-smart-phone-
prospects”, and they can be targeted with relevant ads, either while they read
Samsung smartphone related content, or somewhere else. This can be done
through advanced technologies like data management platforms (DMP),
which will be covered later.
Companies like Grapeshot collect contextual data from various sources,
segment users, and sell these segments to the advertisers (GrapeShot, n.d).
When an advertiser buys a user segment from a company like Grapeshot,
the advertiser knows that this segment consumed relevant content; hence,
they have a potential interest in the advertiser’s product or service.
There are also ad networks like Clicksor that allow their advertisers do
contextual targeting on the publishers.
2.1.3 Intent
Intent data indicates if a person is in-market for a product or service (Sharp,
2015). This is an important type of data since it can connect the advertiser
directly to the potential buyer.
One of the most powerful types of intent data is “search” data. When
someone searches for “Audi A4 prices” on Google, it is a signal that s/he is
potentially interested in buying that car.
Intent data can also be classified as internal and external intent data:
● Internal intent data is captured by the actions of the user made on
company’s own properties
● External intent data is captured by other companies and enabled
through either third parties or directly (Singh & Grennev, 2015)
9
For instance, if a user views different colours of Audi A4 on Audi’s website
and spends more than ten minutes to review the features of the car, s/he is
giving a strong signal regarding his/her intention, which can be considered as
“internal intent” data.
On the other hand, if a person goes to cars.com and compares Audi A4 with
Volkswagen Passat, it is also a signal regarding his/her intention. Audi (or
Volkswagen in this case) can obtain this “external intent” data either by
partnering with cars.com or through third party companies and use it to target
the relevant person.
2.1.4 Image & Video
An interesting data point, which is hard to extract, is the content of the
images. Understanding that a user is looking at a beach photo, for instance,
would allow an advertiser to target him/her with an advertisement of a hotel.
Companies like Visenze or Cortica are working on this problem and try to
convert images to texts, which can be used for targeting purposes.
Video content is another data point that can be used by advertisers to
understand their customers better or show an ad at a particular time of a
video. Picturesafe, a company that tries to recognise video speech and
converting to text is one of the companies working on this problem.
2.1.5 Ethnic
Some advertisers believe ethnicity is the key to people’s identity (The
Economist, 2011). Dublis (2001) recommends that advertisers should have
an understanding of the linguistic acculturation, ethno-cultural stereotypes,
intensity of ethnic identification and language differentiations. Advertising
agencies like Ethnicity work on how to communicate to different cultures
even if they are in the same country. Companies like Ethnic Online Network
allow advertisers to target a particular ethnic group. Facebook is one of the
effective channels where advertisers are able to target people from different
cultures (Boykiv, 2015).
10
2.1.6 Location
An advertiser offering its products or services in only one country would not
want to show its ads in another country. In the online advertising world, it is
quite possible to show ads to irrelevant audiences from other countries. For
instance, CNN.com can be visited from many countries - so an advertiser
selling products only to people live in U.S. would have to target U.S. visitors
on CNN.com specifically, otherwise, a portion of the advertising budget
would be wasted.
Nowadays especially with the high usage of mobile phones, advertising
technologies can understand the location of a user and allow advertisers do
“hyper-local targeting” (Improve Digital, n.d.; Kusek, 2014; Andrews: 2015).
Facebook is working on using location data in new forms like understanding
which stores the users visit after seeing the ads (Constine, 2016).
Companies like Digital Element and Maxmind are providing data about the
location of the users.
2.1.7 Social Media
Social media advertising proved to be very useful for the marketers (DeMers,
2014). The market size is estimated to be around $24 billion worldwide,
around 65% of this market belongs to the Facebook and it is expected to
grow further in the next years (eMarketer, 2015b).
The reason is simple: not only people spend a lot of time on social media
channels, but they also provide a lot of valuable data about themselves. On
Facebook, it is very easy to target women who are between 18-24, engaged
and like music. Facebook knows this data because people give it to them.
Actually, Facebook collects a lot of data from everyone, things people do,
content they provide, their location and so on (Facebook, n.d.b).
It is important to note that some social media channels like Snapchat do not
allow targeting at an advanced level - yet. Moreover, Snapchat CEO Evan
11
Spiegel claims targeting belongs to the early days of internet advertising
(Snapchat, 2015). Ironically, the first question Snapchat answers in the
F.A.Q. section of their website is “What type of ad targeting does Snapchat
offer?” (SnapChat, n.d.), and some industry experts think that they work on a
new advertising technology allowing more targeting options to the advertisers
(Sloane, 2016).
2.1.8 Social Sharing
When people see something they like, dislike, important news, etc. they tend
to share it with their friends or families who they think would also be
interested. They can do it through email, social media channels like Twitter or
SMS. It gives advertisers an idea about people’s intentions or interests.
For instance, if a person shares news about dropping smartphone prices with
his/her friends, the advertiser could assume that the person who shares the
news, and the person who receives the shared message would be interested
in buying a phone. Thus, showing smartphone ads to these people would
make sense.
Companies like RadiumOne use algorithms to analyse this social sharing
data and make it actionable for the advertisers (RadiumOne, n.d.). Other
companies like AddThis collect this data from the publishers and sell it to
advertisers (AddThis, n.d.).
2.1.9 Business
Unlike B2C companies, B2B companies would not have many potential
customers, and they would want to show their ads only to a specific
audience. Moreover, they would want to be able to tailor their messages
considering the industry, position and seniority level of the person who sees
their ads. For instance, a company called SAP serves in 26 industries;
hence, when they run an advertising campaign, they want their messages to
be tailored accordingly (Anon, 2015).
12
LinkedIn provides advertisers with different types of targeting options like job
title and function, industry, and seniority level of the person who sees the ad
(LinkedIn, n.d.).
Other companies like Bombora and Multiview also provide different types of
targeting options for B2B advertisers.
2.1.10 CRM
A Customer Relationship Management (CRM) is a software that gives a
holistic view of each customer’s interactions with the organisation (Cross,
2004). To be able to do that, companies collect different types of data from
their existing/potential clients like their names, phone numbers, last time they
made a purchase, how much they spent and so on.
This information is very precious to advertisers as they collect it directly from
channels like their own website, their call-centre or their store. Since the data
is collected directly from their own sources, its probability to be more reliable
and accurate is higher than others. It means, tailoring the advertising
message considering this data would be very beneficial, and one of the
examples show 390% ROI (Ravesencio, 2015).
Companies like Adroll and Retargeter help with CRM targeting. It is also
possible to use data-management-platforms (DMP) to activate CRM data,
which will be discussed later.
2.1.11 Offline
Many people still use offline channels to buy something and interact with the
brands. This valuable “offline” data is becoming more actionable for “online”
targeting (Vegliante, 2015).
Connecting offline data to the online world is very attractive, and advertisers
start to put more focus on that. Companies like Datalogix and Liveramp help
13
with this process, which is also known as “on-boarding” in the industry (Kaye,
2014).
2.1.12 Behavioural
Current technologies allow advertisers and publishers to collect data from
visitors about their online activities like how much time they spend on a
particular page, what type of content they read, what kind of news they share
with their friends and so on. This data is used to profile the visitors, and these
profiles can be used to show relevant ads to different types of people (Ur at
al., 2012).
Behavioural data is related to the most of the topics mentioned previously.
For instance, if a user searches for “iphone” on the internet, reads content
about “smartphones”, likes or shares an iPhone-related news or clicks an
iPhone ad, it could potentially mean that s/he is interested in buying an
iPhone.
Advertisers use this behavioural data to find and target potential customers.
Likewise, publishers can use this data to show relevant ads to their visitors. It
is also known as “interest-based advertising” (Internet Advertising Bureau
UK, n.d.).
Behavioural data can be collected from many different sources. News sites,
price comparison sites, responses to the existing ads, time spent on a
product page, etc. can be used to collect data from a user and create a
profile about him/her.
Behavioural advertising helps advertisers greatly, as it results in higher
conversion to sales. In some cases, it even modifies consumer behaviours
by changing users’ self-perceptions (Walker Reczek, Summers & Smith,
2016). In one study, it is observed that when people see an advertisement of
an environmental product, they feel “greener” and they even tend to donate
14
money to an environmental charity (Summers, Smith & Walker Reczek,
2016).
Moreover, behavioural data can be used to “predict” consumer behaviour
and target them accordingly. Some people believe it is the future of
advertising (Goodwin, 2014), and companies like Affectv are focusing in this
area.
Behavioural data can also be used to find similar audiences, which are
discovered algorithmically. This is also called “look-alike targeting” (Rajeck,
2014). Some advertisers who have tried this targeting model achieved very
good results (Goodfellow, 2014).
Two companies, in particular, have access to a very rich consumer data
through many different channels: Google and Facebook.
Thanks to the tools or services like Google.com, Youtube.com, Google
Analytics, Chrome, Gmail or Android, Google can collect a massive amount
of data about its users. This data enables Google to be very powerful in
online advertising space (Team, 2015).
Similarly, thanks to the tools or services like Facebook.com, WhatsApp,
Instagram or “Like” buttons, Facebook also collects a vast amount of data
from its users and use it to personalise ads (Facebook, n.d.b).
Behavioural data is crucial in the online advertising industry, and it can be
expected to be more important in the future.
2.2 Connecting to User
As discussed previously, there are many types of data that can be collected.
But this data needs to be “connected” to its owner so that it would be
meaningful. It also needs to be up-to-date to be more useful; hence, its
owner should constantly be tracked.
15
There are many techniques to do these; some major ones are explained
below.
2.2.1 Cookies
Cookies are the most popular method to track people online. It is so popular
that European Commission has a dedicated page about cookies describing
them as a piece of data websites ask browsers to store on people’s
computers or mobile devices. This data can be about many things like
remembering users’ preferences, identify them, helping them complete tasks
without re-entering data and so on.
There are different types of cookies like the session cookie which is deleted
when the user closes the browser, the permanent cookie which is not deleted
until a pre-determined date, the first-party cookie which is set by the domain
user visits, and the third-party cookie which is set by a different domain than
the visited one (European Commission, n.d.).
Many premium publishers like The Economist or The Guardian notifies users
about their cookie policies when people visit them the first time.
2.2.2 Flash Cookies
Adobe’s Flash software can also be used to store information - if installed by
the user. It has advantages like not being easy to be deleted and cross-
browser tracking capabilities (Ullrich, 2015). Since Flash is expected to be
“dead” very soon (Singleton, 2016), this type of cookie may not be used in
the future.
2.2.3 IP Address
An Internet Protocol (IP) address is an identifier every computer uses when
connecting each other and share data (Guthrie Weissman, 2015). IP address
can be used to track the online behaviour of users. However, it is not very
16
reliable, since users can change their IP addresses or multiple users can
share the same IP address (Ullrich, 2015).
2.2.4 Browser Fingerprinting
Browser fingerprinting is done through configuration and settings made in the
browsers (Panopticlick, n.d.), which are likely to be unique. Hence, it can be
used to track users (Ullrich, 2015).
2.2.5 Local Storage
HTML5 is a web language promising to launch the next era of websites,
smartphone apps and online videos (Goldman, 2012). It has introduced
some mechanisms like session storage and local storage, which can help
with storing data on the browser of the user; hence, helping with tracking
their behaviours (Anon., 2016).
2.2.6 Canvas Fingerprinting
Mowery and Shacham (2012) introduced canvas fingerprinting by utilising a
feature of HTML5, which became popular in time (Wagstaff, 2014). It is a
very persistent way of tracking users, and there is not an easy way of
stopping it (Ullrich, 2015).
2.2.7 Carrier Injected Headers
Some mobile carriers track users using some methods like altering the
internet (HTTP) requests (Ullrich, 2015). Verizon uses this technique and
calls it “Unique Identifier Header”, which they claim to be used for more
personalised advertising (Verizon, n.d). It is discovered that Verizon has
been doing this practice since 2012 but did not declare it until 2014, which
resulted in a fine of $1,35 million (Ha, 2016a).
Carriers like AT&T also use unique trackers, and they also claim that it is
used to show relevant ads (AT&T, n.d.).
17
2.2.8 Evercookie
By combining many tracking methods, Samy Kamkar developed a software
which can track users persistently. Even if the user deletes cookies, the
Evercookie script could use canvas fingerprinting or another technique to
track him/her, and deleted cookies would be recovered (Kamkar, 2010).
This method became very popular and even featured in The New York Times
(Vega, 2010). On the other hand, there are some privacy and ethical issues
surrounding this method (Cleere, 2015).
2.2.9 ISP-Level Tracking
Some companies like Phorm track virtually all online activities through their
integrations at the internet service provider (ISP) level. Obviously, this is
potentially the best way of tracking users, as it is possible to analyse every
action a user does online. However, it is also the most harmful way from user
privacy point of view, and it led to some scandals in the past (Williams,
2011).
2.2.10 Others
There are many other ways of tracking. Some companies even started to use
more innovative ways like AudioContext fingerprinting (Englehardt &
Narayanan, 2016). However, there is no need to go into details of all of them.
Many companies track users on the web and collect data about them. Online
tools like Ghostery and Disconnect help users understand which companies
track them.
2.3 Data Quality
As discussed before, data can be collected from many sources; from online
behaviours to offline actions. If it is not done properly, there might be some
issues like poor data quality. In an interview, Bob Rupczynski, VP of media,
data and CRM at Kraft Foods, gave an example regarding one of their
campaigns, where they wanted to target females but noticed that female data
18
was only half-accurate. In another campaign, they targeted some people
because of outdated data (Morrisey, 2015).
There are several factors which could cause poor data quality.
2.3.1 One Device & Multiple Users
In the United Kingdom, the average number of people per household is 2.4
(Knipe, 2015), and on average there are 1.3 laptop and 1.2 tablets per
household (Press Association, 2015). Therefore, it is fair to assume that
more than one family member might use the same computer.
When different people use the same computer, it is difficult to understand
who is behind the computer and which ad would be interesting for him/her.
Since smartphones are more personal devices, the case might be different
for them, though mobile devices have their own complications, which will be
explained later.
2.3.2 One User & Multiple Devices
The average household in the United Kingdom owns 7.4 internet enabled
devices (Press Association, 2015), and the average number of people in a
household is 2.4 (Knipe, 2015). Hence, it would be fair to assume that one
user can use multiple devices to connect to the internet.
When one user uses multiple devices, it is difficult to connect his/her data
across devices. However, advertisers want to be able to target the same
person on multiple digital devices, which is called cross-device or cross-
screen targeting (Tan, 2014).
There are two main approaches to cross-device targeting: deterministic and
probabilistic. The deterministic method requires a user to use a unique
identifier, which would help to recognise him/her across devices (Schiff,
2015). Facebook, for instance, is one of the “lucky” companies that is widely
used with login data, and this login information helps Facebook to recognise
19
the same person across devices. Hence, one of the Facebook products,
Atlas, is claiming to help to the advertisers in delivering and measuring ads
across “everything” (Atlas Solutions, n.d.).
Probabilistic method, on the other hand, relies on some algorithms that are
analysing many data points to understand if a person is the same across
devices (Schiff, 2015). Tapad and Drawbridge are two companies that apply
this method.
Both methods are not working perfectly and there is still improvement
needed in this area (eMarketer, n.d.a).
2.3.3 Deleted Cookies
Some people delete their cookies from time to time, and as soon as cookies
are deleted, the data that was kept in cookies and connection to the user is
lost. A 2014 research shows that 28% of users in Europe delete their cookies
four times a month (comScore, 2014). It affects data quality negatively and
may result in poor campaign performance.
2.3.4 Data Loss
Different technology players in the online advertising ecosystem synchronise
cookies with each other (Bonder, 2014), so that they can all track the same
user in a more efficient way. A recent study shows that most of the
companies apply this method (Englehardt & Narayanan, 2016).
However, the more players synchronise data with each other, the more
problematic it becomes, as the number of network requests increases
significantly (Mitchell, 2014). It may also cause discrepancies and data loss
during this process, which can be up to 30% (Ad Ops Insider, 2011a).
2.3.5 Recency
It is important to show users timely ads. It is an obvious situation, but as
mentioned before, some people can be targeted due to outdated data.
20
Therefore, it is important to understand user’s “recent” behaviours and show
ads accordingly, which is called “recency” in the industry.
2.3.6 Missing Data Points
Unfortunately, it is not always possible to gather all necessary data or to fully
merge different data points of a user. Some of the reasons are:
● Users visit various websites, and not all these websites have a
centralised way of collecting and sharing data. For instance, Google
does not share search data with Facebook.
● One user may use multiple devices - as discussed before.
● Users have “offline” behaviours as well. For instance, someone might
buy a car from a car dealer after doing some online research and
seeing some online ads, but this offline data point might not be
synchronised with the online world.
2.3.7 Poor Integration / Implementation
As will be explained later, using “pixels” is a popular way of tracking users,
which are small codes placed in the web pages of the publishers or
advertisers. If this technical implementation is not done properly, it might
cause poor data quality. For instance, if pixel implementation is problematic,
it will not be possible to understand whether the user made a purchase;
hence, will be continued to show ads to the user, which could ruin the user
experience (Rosenbaum, 2014).
2.3.8 Technological Challenges
As mentioned before, cookies are the most popular way of tracking users.
However, mobile devices do not fully support cookies, and this causes some
technical difficulties (Internet Advertising Bureau, n.d.).
Another problem is with cross-device targeting. It is already mentioned that
existing methods are not functioning perfectly yet.
21
Finally, one computer can be used by multiple users, and existing
technologies are not advanced enough to immediately detect who uses it.
2.4 Segmentation
Once data is collected, publishers or advertisers create segments of people
that have similar interests and behaviours. For instance, people who visited a
car company’s (e.g. Audi) page and checked a particular car model (e.g. A4)
several times in the last few weeks would fall into the same segment. In this
example, Audi would be able to tailor its message to this specific A4-
interested-segment. Hence, all people under this segment would see the
same banner.
Similarly, publishers can segment their users based on their behaviours on
their website. For instance, people who read some car-related news in the
last few weeks are potentially interested in cars, and they can be potential
customers of car advertisers.
It is important to note that the smaller a segment is, the less attractive it
becomes for publishers due to economic reasons. A segment of 450 people
would not help with increasing revenues, especially in a world where a
publisher is paid per one thousand ad impressions (CPM). If the publisher is
paid per click (CPC), even worse.
2.5 Usage of Data
Data can be used in many areas of online advertising. The major ways will
be explained below.
2.5.1 Traditional Advertising
One of the potential use cases for data usage is with traditional ways of
selling media. Publishers can create audience segments from visitors who
visit their websites. They should be using an ad-server (a software managing
advertising campaigns) which has support for audience segmentation and
22
targeting. DFP, Google’s ad-server for publishers, can create audience
segments and target them (Google, n.d.c).
Alternatively, the publisher can use a DMP, which is a data-focused
technology that will be discussed later. In each case, when an advertiser
asks the publisher to target a specific audience, the publisher can create
these segments if they do not exist yet, and then run an advertising
campaign targeting that specific audience.
2.5.2 Programmatic Advertising
Programmatic advertising is a relatively new and growing way of showing
ads, where machines calculate many variables and decide whether to show
an ad, with a minimal human involvement in the process (Rayport, 2015).
Data is vital for programmatic advertising (Olenski, 2015), as will be
explained shortly.
In programmatic advertising, the most important players in the value chain
are demand-side platforms (DSP), supply side platforms (SSP) and ad
exchanges. For simplicity purposes, the term SSP will be used for both SSPs
and ad exchanges as they are doing a similar job nowadays.
A DSP is a technology mainly used by advertisers or agencies to buy
advertising spaces on publishers. Data is a crucial part of this buying process
since DSPs usually try to target specific users (Marshall, 2014a). As there
are billions of transactions happening in a day (Rayport, 2015), many DSPs
use machine learning to optimise their campaigns.
Similarly, SSPs help publishers to sell their inventory in an automated way
(Marshall, 2014a). Again, data is critical here since advertisers require
publishers to send them information about users, the more data, the better.
These technologies communicate with each other mostly through IAB’s
OpenRTB protocol (Internet Advertising Bureau, 2016). This protocol enables
23
the sellers and buyers transact many data about users like location, gender,
content they read, segment they belong to and so on (Internet Advertising
Bureau, n.d.)
2.5.2.1 Cookie Syncing
Since a DSP gives decisions based on data, it is important to understand
who the user is to make decisions like which ad to show, how much to pay
and so on. On the other hand, users visit publishers and publishers usually
work with SSPs, which are enabling DSPs to show ads to the users.
However, DSPs and SSPs are different systems, they both have their ways
to track users, hence, the same user would probably have different IDs in
these systems. There are many different SSPs and DSPs, and a user is
likely to have a different ID in each system. Through a process called “cookie
syncing”, DSPs and SSPs share user IDs with each other so that they can
recognise the same person.
Cookie syncing does not happen only between DSPs and SSPs, but also
between other programmatic (and non-programmatic) players. A recent study
shows that it is a quite popular method used by many companies (Englehardt
& Narayanan, 2016)
2.5.2.2 DSP & SSP
A DSP uses data mainly through the methods below:
● Cookies: When a DSP shows an ad to a user, it can put a cookie on
his/her browser or read data from it. Through cookie syncing with
other platforms, DSP can track users.
● Web beacons: Web beacons are also known as tracking pixels (or
simply “pixels”). They can be placed on advertiser’s pages to see
which pages a user is visiting and how long s/he stays there, which
helps to understand the intentions of the user. When a user purchases
a product or fills a form, s/he is taken to a “Thank You” page, where
another pixel (a.k.a. conversion pixel) is placed. It is also an important
piece of data since the DSP can analyse the profile of the users who
24
“convert”, and optimise the campaign accordingly (Internet Advertising
Bureau, 2014).
● DMPs: A DMP is used to collect data about audiences, segment them
and pass these segments to DSPs to show relevant ads (Marshall,
2014a). DMPs and DSPs also synchronise cookies so that they can
identify the same person (Mediagician, 2013).
● Third Parties: A DSP can be integrated to many other third party
companies to measure how much of the ad is seen, whether the page
is “safe” for the brand (i.e. there is no content damaging the brand),
whether the user activated the mobile app s/he downloaded and so
on. A DSP can also collect some of this information and use it for
optimisation or reporting purposes. For instance, if the ads are not
seen by anyone on a particular website, DSP can stop targeting that
site.
Similarly, an SSP can utilise data through:
● Cookies: Using cookies, an SSP can track users, collect user
information and operate the technology (Rubicon Project, 2016). One
of the primary purposes for cookie usage in SSPs is to match their IDs
with other third parties like DSPs.
● Web beacons (Pixels): As explained before, pixels can be used to
track browsing behaviour.
● Web Logs: Web-logs also can be used to collect information regarding
how a user interacts with the ads s/he sees (Pubmatic, 2015).
● DMPs: Publishers can use DMPs to segment their users and offer
these segments to the advertisers to show relevant ads.
2.5.3 Targeted Media Buying
Data can be used to target audiences. Instead of showing ads to everyone,
only people who would potentially be interested in the product or service can
be targeted, which would result in efficient budget usage.
25
Targeted media buying can be done in various ways. The major models are
described below.
2.5.3.1 Behavioural
As discussed before, depending on the information like what a user reads,
searches or buys, his/her intentions can be analysed and used for
advertising purposes. An ad-server or DSP can utilise this behavioural data,
and as soon as the related user is detected, an ad can be shown to him/her.
2.5.3.2 Predictive
The concept of "big data" is mostly about human behaviours, and these past
behaviours can help with predicting the person's future actions (Pentland,
2014). When this approach is applied to online advertising, it can be
predicted what the user will buy, when is the right time to show the ad, what
is the chance of conversion and so on.
Companies like Adtheorent and Trusignal use data to make predictions and
to optimise the advertising campaigns.
2.5.3.3 Lookalike (Audience Extension)
By analysing behavioural data of some users, it is possible to find similar
audiences who acted in similar ways and who would likely be interested in
the same product or service (Facebook, n.d.c), which is a form of predictive
targeting (Sukornyk, 2013).
Another name for look-alike targeting is “audience extension”. DSPs like
Adgear and Fiksu have this capability.
2.5.3.4 Retargeting
In simple terms, retargeting is tracking a person who visited an advertiser’s
website and showing him/her an ad on another site (AdRoll, n.d.). For
instance, when a person checks a Samsung Galaxy phone on Samsung’s
website, this product’s ads start to appear to him/her on many other
26
websites. Advertisers use retargeting to reconnect with their prospects or
customers and pull them back to their websites.
There are several types of retargeting; some major ones are explained
below:
● An advertiser can retarget people who visited the advertiser’s website,
which is called site-retargeting.
● Search targeting is the natural capability of search engines like
Google. Using Google’s advertising technologies like AdWords, it is
possible to target people who searched for specific keywords. People
do searches on other websites as well, and their search keywords can
be used to retarget them. Moreover, when people do a search on a
search engine and land to a website, it may be possible to retrieve the
search keyword made on the search engine, and retarget people
accordingly.
● Social retargeting is retargeting people on social media channels.
Facebook allows advertisers to retarget people through Facebook
Exchange (FBX) (Smith, 2014), though this service may be shut down
soon (Shields, 2016). Twitter also has similar capabilities
(Shrivastava, 2014).
● Sequential retargeting is showing ads to a person in a sequence
(Gaudio, 2014). For instance, after seeing the first ad of a car, the
person can be retargeted with another ad giving more details about
the car or showing the interior design of it.
● CRM retargeting, which is a type of retargeting, will be explained
separately.
Based on the banner type, there are two types of retargeting: static and
dynamic. A static retargeting is simply retargeting people with a non-
personalised banner, whereas dynamic retargeting is personalising the
banner and show the audience the most relevant products or services
(Perfect Audience, n.d.). For instance, Samsung could retarget its website’s
visitors with the same banner, which is an example of static retargeting.
27
Alternatively, it could show to each of them the product they have viewed, so
one visitor could see a TV ad whereas the other could see a smartphone ad,
which is an example of dynamic retargeting. In any case, utilisation of data is
crucial. Especially for dynamic retargeting, a deeper integration between
retargeting technology and the advertiser is necessary.
Retargeting has proven to be one of the most effective targeting methods
(Koetsier, 2014). Companies like Criteo, Adroll and Retargeter are focusing
on this area.
2.5.3.5 CRM
Customer relationship management (CRM) allows businesses to manage
their relationships with their clients and prospects (Burnham, 2013).
Customer and prospect data can be integrated with advertising technologies
and make customers and prospects available for (re)targeting. As an
example, a bank can target customers who spend more than $5,000 in the
last two months and offer them a premium credit card.
Since CRM data is about a company’s existing clients or prospects, it is very
precious. Existing customer and prospect data can be activated in many
different combinations (e.g. a bank can target people who spent more than
$5,000 monthly and have a loan less than $100,000). It is also possible to
add offline data into the mix, which is obtained from stores or call centres.
For instance, a retailer can target customers who spent more than $500 in its
stores in the last few weeks, and offer them a special discount.
Companies like Adroll, Retargeter and Liveramp can use CRM data to target
audiences. DMP companies like Bluekai, Lotame or CXense are also able to
utilise CRM data (online and offline).
2.5.3.6 Cross Device
To be able to target the same person across devices, s/he needs to be
identified on different devices. It has been discussed before that this can be
achieved through two methods: deterministic, which uses unique identifiers
28
to understand who the user is, or probabilistic, which uses an algorithmic
approach to “guess” who the user is.
To be able to use Facebook, a user should login first - on any device. It helps
Facebook to identify the same user on different devices. Google tries to use
login data as much as possible, asking its users to login on Chrome, on
Android and so on. Adobe is also working on a technology which uses a
“pool” of login data from different parties, so that if any of the parties
recognises the user, it can share this information with others (Ha, 2016b).
These are all examples of deterministic approach.
Companies like Bluecava, Drawbridge, Tapad and Crosswise uses
probabilistic models to identify the users.
Cross-device targeting is one of the important challenges for the advertisers,
and the existing technologies are not very useful yet (eMarketer, n.d.a).
2.5.4 Campaign Measurement & Optimisation
Data can be used to measure and optimise advertising campaigns.
2.5.4.1 Campaign Measurement
One of the obvious applications of data is campaign measurement. How
many times an ad is shown and clicked can be measured easily, but it is also
possible to measure how much of the banner stayed in the viewing area of
users, whether a user made a purchase after seeing the ad and so on.
2.5.4.2 Banner Optimisation
Usually, advertisers create a few different versions of their banners and show
them to users in a rotation. Some technologies can optimise the advertiser's
campaign after a while by checking which banner is clicked more by the
users, then show only that banner to the users and drop the rest.
29
2.5.4.3 Dynamic Creative Optimisation (DCO)
Dynamic creative optimisation (DCO) is creating banners automatically in
real time, based on interest or action data of the user (myThings, n.d.). This
would help the same banner to be tailored for each user considering his/her
data and show him/her the latest product s/he viewed for instance, which can
also be called "personalised banner". DCO results in much better campaign
results than static banners (Sizmek, 2014; Ad Ops Insider, 2011).
2.5.4.4 Bidding Optimisation
Real-time bidding (RTB) is the most popular way of trading media in the
programmatic advertising ecosystem. It is buying or selling ad impressions in
an auction which happens in milliseconds. When a user loads the page,
publisher’s SSP auctions him/her off to the advertiser’s DSP that pays the
highest price (Marshall, 2014a).
DSP decides the bid value by analysing data. If a user has viewed a book
and a house recently, for instance, a real estate company would be willing to
pay a very high price to show an ad to that user. On the other hand, the
bookstore can calculate how much they would make from selling a book to
this person, and calculate the value of showing an ad to him/her accordingly.
2.5.5 Attribution
Before making a purchase, a customer goes through some decision-making
stages. Some call these stages “purchase funnel”, some “consumer decision
journey” (Court et al., 2009). People are exposed to many marketing touch-
points during the purchasing process, and marketers should determine what
affected the consumers before purchasing their product or service, which is
also called “attribution” or “attribution modelling” (Convertro, n.d.).
Many marketers do not have any attribution modelling in place (eMarketer,
2015c), and there is also offline actions of the users which should be
considered, but still, most marketers who use attribution modelling see great
benefits in it (Kehrer, 2013).
30
Companies like Convertro and VisualIQ work on attribution modelling.
2.5.6 Yield Optimisation
Many publishers start to understand how valuable their audience data is
(Stenberg, 2013) and although some of the publishers are still sensitive
about their data (Reagan, 2013), they sell more and more audiences to their
clients (AdExchanger, 2010).
By understanding which audience segments are more demanded by the
advertisers, publishers can adjust prices and maximise revenues.
2.5.7 Audience Optimisation
By targeting different audience segments in advertising campaigns,
advertisers can understand who is more interested in their products or
services, the cost of these segments, their contribution to the revenues, and
make optimisations accordingly.
For instance, if an advertiser tries to sell cars to women, and if young women
click the ads and book test drives more than the others, the advertiser may
start targeting young women more aggressively and spend the budget more
efficiently.
2.5.8 Data Sales
There are three different types of data from ownership perspective: first,
second and third party data.
First-party data is the data owned directly by publisher or advertiser, and
created from sources they own, as demonstrated in Figure 2.1. CRM data, or
behavioural data, which is obtained by the actions made on the publisher’s or
advertiser’s website is first-party data for instance. Publisher or advertiser
can be sure about the quality of the first-party data since it is generated from
their own sources.
31
Second-party data is the data obtained from a partner company. Two (or
more) different companies can make an agreement and share their first-party
data with each other, as demonstrated in Figure 2.1, in which case each
company’s data would be second-party to the other.
Third-party data is the data owned by other companies, aggregated from
different sources and usually sold on marketplaces to any buyer (Lotame,
2013), as presented in Figure 2.1. In this case, quality of the data may be
questioned since the source of the data would not be fully visible to the
buyer.
Figure 2.1 Types of Data (Muir, 2016)
All three types of data can be beneficial.
32
● First-party data can be very precious, especially to those who own it.
An airline company can offer a discount to its customers who have
flown to Japan in the last three months for instance. It is also possible
to sell the first-party data. For example, many advertisers would want
to know if a specific person is reading Financial Times. Thus,
Financial Times could potentially sell this data to the advertisers.
● Second-party data can be very helpful in case existing first-party data
is not enough or can be enriched by additional data from partners. For
instance, Rolex and Mercedes could partner with each other and
share their website visitor cookies since a Rolex owner is likely to be
interested in having a prestigious car and vice versa.
● Third-party data sales will be explained along with data marketplaces
(data exchanges) below.
2.5.8.1 Data Marketplace (Data Exchange)
A data marketplace helps companies to buy and sell audience data
(Techopedia, n.d.). Third-party data is traded in these marketplaces.
For instance, owner of a blog on prestigious cars can put its visitor data to
these marketplaces and any buyer who is interested in this segment could
buy and show ads to those people. MasterCard is one of the companies that
sells data on data marketplaces in a similar fashion (BlueKai, 2012).
Data in the marketplaces can be branded or anonymous (Exelate, n.d.). If the
buyer can see the owner of the data, it is a branded data. If the seller name
is not declared, the data is anonymous. One of the question marks about
data marketplaces is who contributes to the anonymous data segments, how
much they contribute and what is the quality of them.
Some of the data management platforms (DMP) like Bluekai, Exelate and
Lotame also have a data marketplace solution. DMPs will be explained later
in detail, but typically a DMP’s focus is helping publishers or advertisers to
analyse and segment their data and make this data actionable, whereas a
data marketplace helps them to trade it. Having an anonymous data selling
33
option in a marketplace may raise questions by the DMP owners regarding
how much of their data is contributing to the quality or quantity of data sold in
the marketplaces.
2.5.9 Frequency Optimisation
Frequency capping is limiting the number of times a person sees the same
ad (Will Marlow Agency, n.d.).
It is important to use frequency capping in advertising campaigns and
deciding optimal frequency correctly since a user might receive campaign
message when s/he is exposed to the creative five times, and targeting
him/her more would be a waste of money. Moreover, being exposed to the
same banner over and over may disturb user, so frequency capping can
protect the brand from being “annoying” (Kessler, 2011).
Research shows that frequency capping can decrease user acquisition costs
up to 30% (Chandler-Pepelnjak & Song, 2009).
Nowadays many people use more than one device, so setting a frequency
capping across devices is becoming more important (BlueCava, n.d.). Data
can help to understand and limit how many times a user has seen the same
ad, not only on the same device but also across devices.
Usually, frequency capping is set at campaign level, so all users are targeted
at the same level of frequency. Finding optimal frequency capping is a
challenging task (Buchbinder et al., 2014) and it may be difficult to decide
this number at the user level. On the other hand, advanced technologies
might make it possible in the future.
2.6 Fraud Detection
Ad fraud is one of the biggest problems in the online advertising industry. It is
mostly caused by non-human traffic, also known as “bots” (Heine, 2016), and
it is expected to cost $7.2 billion to the industry in 2016 (Stefo, 2016).
34
Software, acting like a human, visits the publishers’ website and “see” the
ads that are paid by the advertisers, and sometimes the software “clicks” the
ads, visits the advertiser’s site and spend some time there. Thus, they
consume advertisers’ budgets and harm the industry.
Companies like Forensiq and Integral Ads offer solutions to analyse data and
detect fraudulent traffic generated by bots.
2.7 Viewability
Viewability is another important topic in the online advertising industry. Many
publishers have more than one banner space on their pages, and not all of
them enter the viewing area of their visitors; hence, the advertisers question
if their budgets are spent wisely.
A banner ad displayed on a desktop device is considered to be “viewable” if
50% of it is shown at least 1 second to the user (Internet Advertising Bureau,
2014b). On the other hand, companies like Facebook have different
approaches to the definition of viewability (Marshall, 2015b).
So, there is a debate about what a “viewable ad” means (Marshall, 2015c)
and different vendors provide different results under same test conditions
(Internet Advertising Bureau, 2014b), but still, it is an important subject for
the industry. Also, the “person” who views the ad should be real, not fraud,
so this topic is intersecting with fraud detection at some points.
Companies like Meetrics and Moat help advertisers solve this problem.
2.8 Mobile-Specific Challenges & Benefits
Smartphones become more and more popular, helping mainly to
communicate and to access information among other things. Hence, mobile
internet penetration is increasing rapidly - it is forecasted to reach 71% by
35
2019 (Internet Society, 2015), and in 2015, Google announced that global
mobile searches surpassed PC searches first time (Macmillan, 2015).
Although it may not mean that internet access through PCs will die (Marshall,
2015a), it shows how dominant mobile devices started to be in peoples’ lives.
High usage of mobile devices bring some challenges to publishers and
advertisers, since tracking a user on mobile devices is not easy (Internet
Advertising Bureau, n.d.).
2.8.1 Publishers
An ordinary premium publisher would have a regular website, a mobile
website and mobile apps for both iOS and Android supporting phones. As a
result, running a cross-platform campaign might be challenging. Doubleclick
for Publishers (DFP), the widely used ad-server of Google, provides some
mechanisms to track users on mobile devices, one of them being publisher
provided identifiers - PPID (Google, n.d.e). But not all ad-servers have similar
functionalities, and not all publishers can provide such unique identifiers.
Another challenge is about frequency capping, which is limiting the number
of times a person sees the same ad in a given period. Since it is hard to track
the same person across devices, it is also difficult to prevent the same
person seeing the same ad more than desired, which can potentially impact
campaign results and reputation of the publisher negatively.
2.8.2 Advertisers
Tracking a user on a single mobile device is hard, but it is not the only
problem for advertisers. Many people do not use a single device to connect
to the internet. 90% of people shift between devices to complete a task
(Taube, 2013). When British people are at home, they switch devices 21
times an hour (Ridley, 2014). It creates further complications for the
advertisers.
36
On the other hand, advertisers would want to collect data about who sees
their ads and how many times, how people react to these ads, on which
channels they react positively or negatively and so on. Existing technologies
do not fully help with this need due to difficulties with tracking the same
person across devices.
2.8.3 Cookies
Using cookies is still a popular method to track users on mobile devices. On
the other hand, there are different mobile devices, web browsers and
applications. First party cookies are allowed by almost all web browsers.
However, third party cookies are not allowed by some of them. Moreover,
there is only a limited allowance on mobile apps (Internet Advertising Bureau,
2013).
2.8.4 Other Tracking Methods
There are also other methods of tracking mobile behaviours. Major ones are
briefly described below.
● Client / Device Generated ID: These are unique identifiers provided by
mobile operating systems. Both iOS and Android offers identifiers to
marketers (Google, n.d.f). These IDs mainly help with tracking users
across mobile applications within the same device (Tune, 2016).
● Statistical ID: By analysing various data points and applying some
probabilistic algorithms, the uniqueness of a user can be inferred
(Internet Advertising Bureau, 2013), which can help with identifying
the same user across apps, websites and devices. Companies like
Drawbridge and Tapad use this approach.
● HTML5 Cookie Tracking: HTML5’s local storage can be used to store
information and can help with tracking. It can be used only when the
browser is open (Internet Advertising Bureau, 2013).
● Universal Login: If a user logs into the mobile application or website,
this login data can be used to track the user. Some companies like
Facebook require their users to be always logged in, which is very
helpful for tracking purposes (Internet Advertising Bureau, 2013). This
37
identifier can be used to track the same user across browsers, apps
and devices.
Sometimes tracking users is done at mobile operator level; Verizon and
AT&T use special techniques to track users within their network (Verizon
Wireless, 2016; Ha, 2016a).
Cross-device targeting, which is tracking the same person across devices
(Bilton, 2015), is a complicated problem and mainly done through two
models: deterministic and probabilistic. The deterministic model relies on
data of logged-in users - as in the universal login method explained earlier,
and probabilistic model relies on analysing different data points to identify the
same user - as in the statistical ID method discussed before (Heine, 2016).
Although it is in high-demand, existing technologies are still underdeveloped
(eMarketer, n.d.a).
2.8.5 Benefits of Mobile Devices
One of the biggest benefits of advertising on mobile devices is being able to
target people in a particular geographic location since people usually carry
their mobile devices with themselves (Internet Advertising Bureau, 2013).
The location information can be obtained through various methods; hence,
the advertisers can target people in a particular shopping mall for instance.
38
3 BENEFITS AND RISKS FOR THE PLAYERS IN THE ECOSYSTEM
So far it has been discussed how data is collected and used in various
contexts. This section focuses on potential benefits and risks of data for
major players in the online advertising industry; namely publishers,
advertisers, advertising agencies, technology providers and of course users
themselves.
The section will start with the discussion of benefits first, and then will
continue with risks.
3.1 Benefits for Publishers
Publishers, especially large ones like newspapers, are visited by many
people every day. For instance, Nytimes.com is visited by 65.8 million people
(The New York Times Media Kit, 2016) and Forbes.com is visited by 46
million people monthly (Forbes Media, n.d.). It means a company’s existing
and potential customers would likely to be among those millions of people.
Also, the publishers are one of the main channels to show ads to people in
the online advertising world. The only question is; how can an advertiser
show its ads to the relevant people, and not waste their money by showing
ads to others?
In the past, advertisers were using a model called “spray and pray”, showing
their ads to as many people as possible and hoping some of them would
become their customers (Pomerantz, 2014). However, those days are
coming to an end since the consumer tolerance for seeing irrelevant ads are
very low nowadays (Roman, 2015), and obviously, it is not an efficient way of
spending advertising budgets.
Therefore, publishers offer different advertising models to the advertisers, so
that they can reach to the audiences they target. One of the major ways
publishers offer is categorical targeting, so a technology company could
39
show its ads on the -related news, whereas a car company could show its
ads on the car-related news for instance. The other major and trending way
is audience targeting, where the publisher would show technology-related
ads to the technology related people, even if they read other types of news at
that moment.
The major ways data helps to the publishers are explained below.
3.1.1 Data + Media Bundle Sales
Publishers can bundle their data with their media and offer it to the
advertisers. For example, Forbes.com could offer Volkswagen an advertising
campaign on its main page, where ads are shown only to people who are
interested in cars.
This way of showing ads would be beneficial for publishers in various ways:
● It would help organising the inventory and use different pricing models
for different products. For instance, showing ads to the car-lovers on
the main page could be priced differently than showing ads to the
book-lovers.
● Showing ads to relevant people would help to achieve better
campaign results, which could increase publisher’s reputation and
repeat sales.
● Showing relevant ads to the visitors would be less disturbing for them,
which could again increase publisher’s reputation and traffic.
3.1.2 Data Sales
Publishers can offer their data (audience segments) to buyers, without
bundling them to their media, which would allow advertisers target these
audiences on any publisher they want. For instance, Forbes.com could sell
cookies of car-lovers to Volkswagen, and Volkswagen could target these
people on TheGuardian.com.
Publishers can do this through different ways like:
40
● They could allow a buyer to collect data directly from their websites by
putting advertiser’s pixels into their website.
● They could allow an advertiser to put a cookie into user’s browser
each time an ad is shown.
● They could use a DMP (a technology which will be explained later)
and pass segment cookies to the advertiser’s ad-server or DSP.
3.1.3 Data Partnerships
One of the ways publishers can benefit from data is through partnerships
with other companies. For instance, Telegraph.co.uk has a section dedicated
to cars (The Telegraph, 2016), but they can partner with Autoblog.com to
access to more people who are interested in cars. Thus, even if a visitor of
Telegraph.co.uk does not read car related news on its site, s/he may read on
Autoblog.com, and The Telegraph would know this through its partnership
with Autoblog.com. As explained before, this is called second-party data, and
it can be very beneficial.
Data partnerships can help publishers in the ways below:
● It can help contribute to understanding visitors better, providing a
complete view by enhancing existing or integrating non-existing data
points.
● When selling data, directly or bundled with media, additional data
points would contribute to data quality, and in some cases, it would
increase revenues further thanks to the new audiences discovered in
partner’s website and higher unit prices.
● Data partnerships with prominent companies may contribute to the
brand image of a publisher.
3.1.4 Audience Enrichment
A user usually visits more than one publisher in a day; so a publisher can
have only a partial understanding of its users. Buying data from other
websites through data marketplaces (a.k.a. third party data) and enriching
publisher’s own data points about the user (a.k.a. first party data) is called
41
“audience enrichment” (Lotame, 2016). Actually, it is possible to combine first
party, second party and third party data and enrich existing data points.
Some publishers offer “guaranteed” audiences, combining advertiser’s first
party data with other data points, aiming to make sure that only people that
advertiser targets see the ads (Friedman, 2016).
The benefits of audience enrichment are similar to the data partnerships
explained before.
3.2 Benefits for advertisers
Data is crucial to advertisers. It can help mainly in the ways below.
3.2.1 Targeting a Specific Audience
A car company would usually want to show its ads to the potential car
buyers. To be able to “find” this audience, they can do contextual targeting -
showing ads in car related websites or car related news. However, this would
help only partially, because a potential car buyer would not spend his/her
whole time on a car-related website or page, s/he would read other news or
visit other websites as well. Missing him/her on those pages, or worse,
allowing a competitor to target him/her on other sites would reduce the
chances of selling him/her a car.
So, the advertisers would want to show their ads to potential clients wherever
they are. Data can help to achieve this goal.
● An advertiser can use first-party data to (re)target potential customers.
Thus, whenever a potential client is “spotted” on a publisher,
advertiser would show an ad to him/her.
● An advertiser can partner with another brand to share data with each
other. It would help them to use second party data to target potential
customers. For example, Rolex and Mercedes could share data with
each other, and whenever a Mercedes client is spotted on a publisher,
Rolex can show an ad to him/her. On the other hand, Mercedes may
42
also want to show an ad to this particular user, which should be taken
into consideration by both parties.
● An advertiser can use third party data to target potential customers.
For example, MasterCard sells audiences based on their spending
behaviour; hence, a company like Rolex could buy data from
MasterCard and target people who have high monthly spending.
3.2.2 Banner Optimisation
Showing dynamic banners with relevant products to the users can result in
huge benefits (DoubleClick Advertiser Blog, n.d.) - as discussed before in
dynamic creative optimisation (DCO) section.
By using data, it is possible to show personalised banners that may attract
customers, either by retargeting or predicting what could be interesting for
them.
3.2.3 Understanding Customers Better
By analysing different data points about users, advertisers can understand
what type of banners or channels drive their customers to sales (or another
action they want), what is the optimal price to show them ads, what product
or service their customers are interested in and so on.
This understanding would result in optimisations in campaign budgets and
channels, also would potentially increase revenues by selling more products
or services.
3.2.4 Viewability
Data allows advertisers to understand how many and how much of their ads
are actually “viewed” by users, which is called “viewability” (Marshall, 2014b),
and make optimisations accordingly.
43
As discussed before, the definition of “viewability” is not entirely standardised
yet, but it will definitely evolve in time, and eventually help advertisers spend
their budgets on viewable ads only.
3.2.5 Fraud
It is estimated that ad fraud will cost advertisers $7.2 billion in 2016 (Bond,
2016), which is a big concern for the advertisers. One of the important
players in programmatic advertising space, AppNexus, lost around 90% of its
traffic after removing its fraudulent traffic (Wiessbrot, 2016).
Data helps advertisers to detect fraudulent activities and stop them as much
as possible.
3.2.6 Spending Budgets Efficiently
Showing ads only to target audience helps with spending the campaign
budgets more efficiently as well.
● Using data, instead of showing ads to everyone and pay for it,
advertisers can show ads only to existing & potential customers.
● After a user sees an ad a predetermined number of times, s/he can be
excluded from the targeting group (a.k.a. frequency capping).
● By understanding who responds positively to ads, shifting budgets
towards that audience would help with budget optimisation.
● Programmatic advertising technologies use sophisticated algorithms
to evaluate many data points and decide when and where to show an
ad to an audience, and optimal price to show the ads.
3.2.7 Brand Safety
Data can help advertisers to show their ads only on websites or pages where
it is “safe” for them. For instance, an airline company would not want to show
its ads on a page where there is news about a plane crash.
44
3.2.8 Data Sales
Some brands that look for new revenue streams can decide to sell their data.
Companies like MasterCard and American Express are two examples that
use this business model (Kaye, 2013).
3.3 Benefits for Advertising Agencies
Usually, advertisers work with an advertising agency to run their advertising
campaigns. Therefore, an agency would have access to data of its clients. As
explained before, this data can be very useful for the advertiser; hence, the
agency should also be able to utilise it.
The programmatic specialist agency Xaxis, owned by the largest advertising
and marketing agency WPP, has invested $25 million in developing a data
management platform (DMP) to help them gain leverage and make a more
efficient buying, and it is considered part of their technology stack
(Kantrowitz, 2014).
Annalect, owned by the second largest advertising and marketing agency
Omnicom, provides solutions and helps marketers to make data actionable
(Omnicom Media Group, 2016).
Another programmatic specialist agency, Vivaki, owned by the third largest
advertising agency Publicis, also owns a DMP to manage data (Vivaki,
2015).
So, all of the large agencies have created their own solutions to utilise data.
Chief Strategy Officer of Accuen, Murphy (2016), thinks data would help with
optimising media investments, tailoring relevant messages to audiences and
provide accountability and knowledge to marketers.
45
3.3.1 Optimising Media Investments
Some advertising agencies are specialised in media buying. Depending on
campaign goals of their clients, they plan where to show the ads, and how
much to spend on each media channel.
Data would help greatly with this planning and buying process since the
agency would know which channels are more “viewable”, less fraudulent,
how much they contributed to sales and so on.
If the agency benefits from programmatic advertising methods, data can help
further by helping with real-time optimisation.
3.3.2 Tailoring Relevant Messages
As explained before, data helps with targeting a specific audience and
personalise the ads considering context or audience data.
3.3.3 Provide Accountability and Knowledge to Marketers
An agency can use data to justify its media spending. It would help to
agency’s client to understand the rationale behind agency’s actions.
Moreover, data can help with analysing the types of banners that lead to
sales, the profiles of people who respond positively to the campaign, what
could improve the results and so on. This information is very important for the
advertisers, and agencies can use data to gain and pass this knowledge to
their clients.
3.4 Benefits for Technology Providers
Data helps advertising technology providers in many ways.
3.4.1 Data at the Core
By their nature, almost all technologies depend on data to be able to function
properly. An ad-server system like DFP (Doubleclick for Publishers) uses
46
data to decide which campaign to prioritise or which banner to show and
when. A DSP like DataXu uses data to determine the optimal bid price to
show an ad to a specific audience. An SSP like Pubmatic uses data to
decide the winner of an auction. A fraud detection system like Forensiq uses
data to determine whether an activity is fraudulent. A DMP like Cxense
directly promises to manage “data”.
It is possible to give more examples, but it should be obvious that data is at
the core of almost all technologies in the online advertising ecosystem.
3.4.2 New Business Opportunities
Different types of data can be used to solve different problems. In the past,
there was just a simple interaction between media seller and media buyer,
and an ad-server was enough to run an advertising campaign. Then some
start-ups wanted to automate this process with real-time auctions. Some
companies wanted to understand if the ads were “really” viewed. Some
companies wanted to expose fraudulent activities. Some companies wanted
to know if it is the same person on different devices.
It is possible to give more examples, but as it can be seen, data helped many
start-ups to be borne, which are specialised in analysing different types of
data and helping advertisers or publishers to do their jobs better. Some of
these start-ups like Criteo now worth billions of dollars.
3.5 Benefits for Users
Users have different attitudes against the fact that their data is used to show
them relevant ads. Some people desire to see relevant ads (Sasaki, n.d.)
and some of the research support this approach (eMarketer, 2016a),
whereas some people find seeing targeted ads “creepy” (eMarketer, 2015b).
Still, there are some benefits in seeing relevant ads.
47
3.5.1 Free Content
One of the most important advantages of seeing ads is that they translate to
free content. Not many people would pay to avoid seeing ads (eMarketer,
2015d).
Many large publishers block their content to those who use ad-blockers.
Figure 3.1 demonstrates how Forbes.com warns a user who uses an ad-
blocker.
Figure 3.1 Forbes.com main page with ad-blocker
Not using or disabling an ad-blocker and visiting a publisher means giving
consent to the publisher to collect data and showing relevant ads whenever
possible. Figure 3.2 demonstrates how The Telegraph notifies its visitors
about the usage of cookies.
48
Figure 3.2 The Telegraph main page with cookie usage notification
As a result, users can either choose not to access to free content or access it
knowing the fact that their data is used to show them relevant ads.
3.5.2 Inevitability
Related to the topic above, if seeing an ad is inevitable, why not seeing a
relevant one? Consumer tolerance for irrelevant ads is very low nowadays
(Roman, 2015). Users may have realised that there is no escape of seeing
ads; hence, prefer seeing a relevant one.
On the other hand, with the rise of ad-blockers, seeing an ad is not entirely
inevitable anymore. Ad-blockers are used by around 200 million people
(Richards, 2015). To fight the ad-blockers, publishers use technologies that
detect if a user is using an ad-blocker and “forces” him/her to see an ad by
bypassing the ad-blocker, or offering other options like paid subscriptions.
Some of the publishers do not allow their content to be viewed by a user if
s/he is using an ad-blocker.
49
3.5.3 Discounts & Special Offers
Users may benefit from discounts and special offers that are tailored
considering their data. 52% of Canadian internet users would like to see such
ads (eMarketer, 2015d)
3.5.4 Additional Benefits
Some companies offer various benefits to users when they share their data
for online advertising purposes. Datacoup, for instance, offers users money
in return of their data and sells this data to the advertisers (Van Allen, 2014)
along with other potential buyers (Datacoup, n.d.).
Similarly, People.io offers users credits in return of their data, which can be
used in the exchange of products, services, offers and so on (People, n.d.).
3.6 Risks for Publishers
Data has its disadvantages for publishers as well. Some of them are
explained below.
3.6.1 Privacy Concerns
Tracking users and collecting data about their behaviours triggered some
privacy concerns among users - many people do not like the fact that a “big
brother” is watching them constantly.
31% of British people who use ad-blockers use them due to privacy concerns
(Internet Advertising Bureau UK, 2015). Similarly, 50% of people in the US
who use ad-blockers feel that their personal data is misused (Richards,
2015). It shows how serious the problem is, especially for the publishers
since they are the ones who cannot sell their traffic.
3.6.2 From Publisher to Audience
Especially with the rise of programmatic advertising, advertisers started to be
able to target specific people wherever they are (Battelle, 2014). It means
50
that the audience might be on a premium website or might be reading a small
blog; in any case, the advertiser can target him/her.
This shift in the mindsets of the advertisers, from publisher to audience,
might affect premium publishers negatively. The New York Times mentions
this in its 2015 annual report; regards this problem as “audience
fragmentation” and explains how it affects their business negatively (The
New York Times Company, 2015).
3.6.3 Too Targeted = Too Low Spending?
If advertisers target only specific audiences, publishers would not be able to
sell enough inventory. If there are only 1000 potential real estate buyers out
of millions of visitors, showing ads only to these people would result in very
little revenue.
3.6.4 Information Asymmetry
In some cases, a user might be too valuable to an advertiser and publisher
may not be aware of this. For instance, a car company may know that a user
is a potential car-buyer thanks to its first-party data, while the publisher may
not be aware of this. Therefore, even if the car company would be willing to
pay a high price to show an ad to that user, lack of data and knowledge
would not allow the publisher to increase the price for that user, which would
result in leaving money on the table.
3.6.5 Data Leakage
Some publishers want to sell their data only bundled with their own media.
Data leakage happens when a third party company collects data about
publisher’s users and use this data elsewhere (Moses, 2015).
Some of the companies that work with publishers directly might cause this
issue, “inviting” other companies to drop pixels on users who visit the
publisher’s website. Sometimes even companies that work directly with the
51
publisher might not be aware that another company is using “them” to reach
to the users of the publisher.
Although data leakage is a big concern of publishers, many of them use
Google Analytics, Facebook buttons, or other technologies like AddThis.
These companies explicitly announce that they use their tools to collect data
and use it either for internal purposes or sell it to other companies.
3.6.6 Impact on Revenues
Many different companies in the online advertising industry use and analyse
data in a variety of ways; some use it for brand safety purposes, some use it
for viewability analysis, some use it for fraud analysis and so on. All these
different players get a share from each advertising dollar spent. Research
shows that publishers receive around 65% of the advertising spending in EU
- including agency commissions (IHS Inc., 2015), while some think they
receive only 30% (Polisetti, 2014).
So, it seems like the more intermediaries analyse and interpret different parts
of data, the more revenues publishers lose.
3.7 Risks for Advertisers
Data may create some disadvantages for advertisers in the ways explained
below.
3.7.1 Being Annoying
Particularly with the retargeting method, there is a risk of being annoying to
users and damage the brand’s reputation. It happens when the user feels like
s/he is being “stalked”, particularly when s/he has already purchased the
product, but still seeing the same ad again and again (Rosenbaum, 2014).
52
3.7.2 Data Leakage
As with publishers, there is a data leakage risk for advertisers as well. It
happens when the advertiser works with a company to retarget people for
instance, and willingly or unwillingly, this company calls “other” companies to
track the customers of the advertiser, or use data of one advertiser for the
campaign of another.
Similarly, advertising agencies usually work with multiple advertisers, and
they might use one advertisers’ data in another one’s campaign.
3.7.3 Increasing Complexity
One of the advantages online advertising brought to advertisers was being
able to measure how many times an ad was shown and how many times it
was clicked. It was a big improvement in comparison to conventional
advertising channels like TV or newspaper, where there was no way of
measuring how many times an ad was shown.
Being able to measure impressions and clicks required further data like who
saw the ads, how much of a banner stayed in the viewing area of a user and
for how long, what banner size contributed to sales more, was there any
external factors contributing to the decision process of the user and so on.
With the popularity of smartphones, things started to get more complicated
since one user was using multiple devices, so the need of unifying data
points from different devices increased a lot.
It seems like the more data advertisers get, the more data they need to
analyse further and make sense of the data at hand. This is not necessarily a
bad thing, and obviously, it is better than having no data at all, but in
comparison to the old days where there were some simple media plans and
simple reports to make sense of, the current way of online advertising can be
a bit “confusing” for some advertisers.
53
Another challenge is about the players in the industry. Some of them claim
they do fraud detection very well, some of them focus on viewability, some of
them are pure video DSP, some of them are pure mobile DSP, some of them
are pure DMP, some of them are both DSP and DMP and so on. It is not
easy for an advertiser to keep up with new companies that are potentially
adding another type of value and test all the claims in an easy way.
Thus, there is an increasing complexity in the market, and not many
advertisers can stay up-to-date with all the changes in the industry.
3.7.4 Impacts on HR
While data and technologies around it evolve, the needed skills and
qualifications of people working in the industry are also evolving. WPP’s Sir
Martin Sorrell referred to this change as “from mad men to maths men”, and
many companies in the industry started to hire people with skills in math,
engineering and analytics (Cookson, 2015). It is predicted that by 2017, Chief
Marketing Officer (CMO) of a company will spend more time on technology
than Chief Information Officer (CIO), therefore, new roles like Chief
Marketing Technologist (CMT) started to appear and companies like
Kimberly-Clark have roles like “Global Head of Marketing Technology”
(Brinker & McLellan, 2014).
As a result, the workforce in the industry is in change and this change may
affect companies concerning training costs, employee costs, hiring
procedures and so on.
Programmatic advertising brings further complications. Its market size will
grow from 60% in 2015 to 80-90% in 2019 in the UK (Internet Advertising
Bureau UK, 2016), and it almost entirely depends on data. The difficulty of
hiring and training people with the right skill set is seen to be the biggest
obstacle in the adaptation of programmatic advertising models (IHS Inc.,
2015).
54
3.7.5 Impacts on Advertising Costs
As discussed before, there are many intermediary companies analysing a
different set of data and getting a cut from advertising budgets in return. As a
result, out of €40.3 billion advertising spend in 2014, publishers received
€30,7 billion - including agency commissions (IHS Inc., 2015), and this figure
should be even lower when agency commissions and publisher-side
technology fees are included into calculation.
Figure 3.3 Display Advertising Ecosystem (Luma Partners, n.d.)
As it can be seen in Figure 3.3, there are many players between a marketer
and a publisher. Hence, advertising costs are impacted negatively. On the
other hand, advertisers would not want to show their ads in brand-damaging
pages, meaning benefits of working with these intermediary companies
should also be considered carefully.
55
3.8 Risks for Advertising Agencies
Adverting agencies may also suffer from the usage of data, mainly in the
ways explained below.
3.8.1 Increasing Complexity
As mentioned before, in the earlier days of online advertising it was possible
to measure how many times an ad was shown and how many times it was
clicked. This data required more data like who saw the ads, how much of
banners stayed in viewing areas of users and for how long, what banner
sizes contributed more to sales and so on. Now the industry needs further
data about smartphones, cross-channel attributions and so on. To make
sense of all these data is not easy - if not impossible in some cases, thus,
daily tasks of agencies are getting more challenging and confusing.
Similarly, it is becoming harder to keep up with the industry since there are
many technology players in the ecosystem, and each of them claims
something else like doing a task better or specialising in another area.
3.8.2 Low Spending
The business model of advertising agencies is mainly taking a share from the
advertising budgets of the advertisers. If an advertiser spends $100K for
instance, the agency may make around $20K (naturally this rate can vary).
However, data may limit this spending in the ways explained below, which
would affect agency revenues negatively.
3.8.2.1 Too Targeted = Low Spending?
Data helps with targeted campaigns; however, spotting advertiser’s target
audience on the web and showing ads to him/her may not be easy. This
would cause the advertising budgets to be spent slower than usual; hence, it
would affect agency’s revenues negatively.
3.8.2.2 Brand Safety = Low Spending?
Data also helps with showing banners in websites or pages where there is no
harmful content for the advertiser. But if a campaign’s safety rules are too
56
strict, spending may be slow as there will be few pages where the ads will be
able to appear.
3.8.3 Benefits(!) of Fraud
Although ad-fraud is a big problem for the online advertising industry, it may
not be necessarily so for all the players in it. Some agencies benefit from low
costs of fraudulent activities and their reports look “excellent” to advertisers
when costs are low. In one example, an agency’s client was so happy with a
campaign, where 90% of activity was fraudulent, that the client renewed
contract with the agency (Davies, 2015).
Thus, the more advertisers utilise data to detect fraudulent activities, the
more agencies may suffer from “client happiness”.
3.8.4 Impacts on HR and Mindsets of Employees
As discussed before, needed skills and qualifications from people who work
in the online advertising industry have been changing. It is even more
challenging from an agency point of view since there is a risk of losing a
client if results are not satisfying. Hence, some agency employees ignore
fraudulent activities, or falsify data, or do other “shady practices” (Davies,
2015). These practices may have some unexpected consequences for the
industry in the future.
3.8.5 Over-reliance on Technology and Data
As mentioned before, poor data quality is one of the problems in the industry.
Also, not all technologies are perfect yet. Murphy (2016) believes agencies'
over-reliance on technology and poor quality 3rd party data may paralyse
and neglect core strategy.
3.9 Risks for Technology Providers
Data creates some challenges for technology providers, mainly in the ways
explained below.
57
3.9.1 Missing Data Points
Data is at the core of many technology providers in the online advertising
space. However, it is not always possible to capture all data points, which
may result in poor decisions by technologies.
For instance, when a user views a product, s/he might be retargeted. When
s/he purchases that product, retargeting should be stopped immediately.
However, this is not always the case, since either technical integration is not
done properly or it is not possible to capture some data points at all.
3.9.2 Crowded Industry
Apart from major companies like Google and Facebook, there are dozens of
other companies (if not hundreds worldwide) in the advertising technology
industry, and this number is growing constantly (Display AdTech, n.d.). It
results not only in intense competition, but also getting “noticed” among so
many players is becoming hard - even if the technology company’s
proposition is unique. Therefore the industry has been consolidating
(Ledbetter, 2015) and small players either go out of business or acquired by
bigger ones.
3.10 Risks for Users
There are two major ways data impacts users negatively.
3.10.1 Privacy
Almost all companies in the online advertising industry are after data of
users. Every bit of data like what they read, search, share or intend to buy is
important. Therefore, many companies track the users collectively.
However, not everyone welcomes being tracked. As discussed before, one of
the top reasons why ad-blockers are used is because of concerns about
privacy (Internet Advertising Bureau UK, 2015; Richards, 2015)
58
3.10.2 Increased Spending
One of the ways data-driven marketing activities help to companies is; they
drive sales and profitability (Forbes, 2015). For instance, retargeted users
are 70% more likely to “convert” on advertisers website (Abramovich, 2013).
It also means users spend more money online (and probably offline as well).
Figure 3.4 and Figure 3.5 show how online spending increased in Europe
and US.
Figure 3.4 Online Spending in Europe (Statista, 2016d)
Figure 3.5 Online Spending in the US (Statista, 2016e)
59
The increase in spending may not be purely due to the usage of data in
online advertising, but it is clearly one of the factors. However, spending
more and saving less is not necessarily a good thing for some users.
60
4 IMPACTS ON THE INDUSTRY
4.1 Privacy Concerns
One of the biggest challenges the online advertising industry faces is privacy
concerns of users and related regulations, which may impact the industry
negatively.
Many advertising technology companies, including the largest ones like
Google and Facebook, base their algorithms mostly on user data. This data
is collected and related to a user mostly through cookies (European
Commission, n.d.) and in a way that keeps the user anonymous. However,
some companies like Facebook work on other types of tracking methods
(Jakubowski & Johnson, 2014) since a cookie is dependent on the browser
of the user and it is difficult to identify the same user across browsers or
devices. It might also mean a threat to the privacy of users.
As a result, many users have some concerns about their online privacy. A
recent study shows that 91% of US consumers do not want companies to
collect information about them without their knowledge, even if they get
discounts in return (Turow, Hennesy & Draper, 2015). Concerns about online
privacy are rising in the UK as well (Ofcom, 2015).
Google collect data from many sources, including videos watched on
Youtube, content of emails, location and so on, which are used for various
reasons including showing tailored ads (Google, 2016).
Similarly, Facebook collects data from many sources, including what users
like, their location, what they do on other Facebook services like WhatsApp
and so on (Facebook, n.d.b.).
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry
The Impacts of Data on the Online Advertising Industry

More Related Content

What's hot

Smart Social Quarterly - 2014 Q4
Smart Social Quarterly - 2014 Q4Smart Social Quarterly - 2014 Q4
Smart Social Quarterly - 2014 Q4
Unified
 
Horizon Partners Newsletter 2013-11-22
Horizon Partners Newsletter 2013-11-22Horizon Partners Newsletter 2013-11-22
Horizon Partners Newsletter 2013-11-22Tyson Hendricksen
 
Measuring Social Return on Investment
Measuring Social Return on Investment Measuring Social Return on Investment
Measuring Social Return on Investment
Goodbuzz Inc.
 
Altimeter Report: Social Marketing Analytics
Altimeter Report: Social Marketing AnalyticsAltimeter Report: Social Marketing Analytics
Altimeter Report: Social Marketing Analytics
Jeremiah Owyang
 
[Infographic] The State of Social Business: Social Media Matures, Yet Many Co...
[Infographic] The State of Social Business: Social Media Matures, Yet Many Co...[Infographic] The State of Social Business: Social Media Matures, Yet Many Co...
[Infographic] The State of Social Business: Social Media Matures, Yet Many Co...
Altimeter, a Prophet Company
 
Marketing in a Digital World
Marketing in a Digital WorldMarketing in a Digital World
Marketing in a Digital World
Ogilvy
 
Wcp ad-tech-report-20140821 (1)
Wcp ad-tech-report-20140821 (1)Wcp ad-tech-report-20140821 (1)
Wcp ad-tech-report-20140821 (1)Sumit Roy
 
Digital marketing trends 2022
Digital marketing trends 2022Digital marketing trends 2022
Digital marketing trends 2022
Amulya Gautam
 
Members, Subcommittee on Digital Commerce and Consumer Protection Re: Underst...
Members, Subcommittee on Digital Commerce and Consumer Protection Re: Underst...Members, Subcommittee on Digital Commerce and Consumer Protection Re: Underst...
Members, Subcommittee on Digital Commerce and Consumer Protection Re: Underst...
Jeff Martinez
 
Flathau unit1 slidshow
Flathau unit1 slidshowFlathau unit1 slidshow
Flathau unit1 slidshow
brentflathau
 
Slideshow
SlideshowSlideshow
Slideshow
AngeliqueThomas4
 
Mindshare at DES: Programmatic: It's Not Really About Cheap Media
Mindshare at DES: Programmatic: It's Not Really About Cheap MediaMindshare at DES: Programmatic: It's Not Really About Cheap Media
Mindshare at DES: Programmatic: It's Not Really About Cheap MediaDigiday
 
[Report] Social Data Intelligence, by Susan Etlinger
[Report] Social Data Intelligence, by Susan Etlinger [Report] Social Data Intelligence, by Susan Etlinger
[Report] Social Data Intelligence, by Susan Etlinger
Altimeter, a Prophet Company
 
What's Next: Social Media Trends 2020
What's Next: Social Media Trends 2020 What's Next: Social Media Trends 2020
What's Next: Social Media Trends 2020
Ogilvy Consulting
 
The open data economy unlocking economic value by opening government and publ...
The open data economy unlocking economic value by opening government and publ...The open data economy unlocking economic value by opening government and publ...
The open data economy unlocking economic value by opening government and publ...
Rick Bouter
 

What's hot (16)

Smart Social Quarterly - 2014 Q4
Smart Social Quarterly - 2014 Q4Smart Social Quarterly - 2014 Q4
Smart Social Quarterly - 2014 Q4
 
Horizon Partners Newsletter 2013-11-22
Horizon Partners Newsletter 2013-11-22Horizon Partners Newsletter 2013-11-22
Horizon Partners Newsletter 2013-11-22
 
Measuring Social Return on Investment
Measuring Social Return on Investment Measuring Social Return on Investment
Measuring Social Return on Investment
 
Altimeter Report: Social Marketing Analytics
Altimeter Report: Social Marketing AnalyticsAltimeter Report: Social Marketing Analytics
Altimeter Report: Social Marketing Analytics
 
[Infographic] The State of Social Business: Social Media Matures, Yet Many Co...
[Infographic] The State of Social Business: Social Media Matures, Yet Many Co...[Infographic] The State of Social Business: Social Media Matures, Yet Many Co...
[Infographic] The State of Social Business: Social Media Matures, Yet Many Co...
 
Marketing in a Digital World
Marketing in a Digital WorldMarketing in a Digital World
Marketing in a Digital World
 
Wcp ad-tech-report-20140821 (1)
Wcp ad-tech-report-20140821 (1)Wcp ad-tech-report-20140821 (1)
Wcp ad-tech-report-20140821 (1)
 
Internet advertising
Internet advertisingInternet advertising
Internet advertising
 
Digital marketing trends 2022
Digital marketing trends 2022Digital marketing trends 2022
Digital marketing trends 2022
 
Members, Subcommittee on Digital Commerce and Consumer Protection Re: Underst...
Members, Subcommittee on Digital Commerce and Consumer Protection Re: Underst...Members, Subcommittee on Digital Commerce and Consumer Protection Re: Underst...
Members, Subcommittee on Digital Commerce and Consumer Protection Re: Underst...
 
Flathau unit1 slidshow
Flathau unit1 slidshowFlathau unit1 slidshow
Flathau unit1 slidshow
 
Slideshow
SlideshowSlideshow
Slideshow
 
Mindshare at DES: Programmatic: It's Not Really About Cheap Media
Mindshare at DES: Programmatic: It's Not Really About Cheap MediaMindshare at DES: Programmatic: It's Not Really About Cheap Media
Mindshare at DES: Programmatic: It's Not Really About Cheap Media
 
[Report] Social Data Intelligence, by Susan Etlinger
[Report] Social Data Intelligence, by Susan Etlinger [Report] Social Data Intelligence, by Susan Etlinger
[Report] Social Data Intelligence, by Susan Etlinger
 
What's Next: Social Media Trends 2020
What's Next: Social Media Trends 2020 What's Next: Social Media Trends 2020
What's Next: Social Media Trends 2020
 
The open data economy unlocking economic value by opening government and publ...
The open data economy unlocking economic value by opening government and publ...The open data economy unlocking economic value by opening government and publ...
The open data economy unlocking economic value by opening government and publ...
 

Similar to The Impacts of Data on the Online Advertising Industry

Evolution of Digital Marketing and its Impact on Privacy
Evolution of Digital Marketing and its Impact on PrivacyEvolution of Digital Marketing and its Impact on Privacy
Evolution of Digital Marketing and its Impact on Privacy
ijtsrd
 
GroupM Brand Safety Playbook For Marketers
GroupM Brand Safety Playbook For MarketersGroupM Brand Safety Playbook For Marketers
GroupM Brand Safety Playbook For Marketers
Social Samosa
 
GroupM Brand Safety Playbook 2020
GroupM Brand Safety Playbook 2020GroupM Brand Safety Playbook 2020
GroupM Brand Safety Playbook 2020
GroupM THailand
 
ClickZ Buyers Guide - Bid Management - PPC, Display, Social Media
ClickZ Buyers Guide - Bid Management - PPC, Display, Social MediaClickZ Buyers Guide - Bid Management - PPC, Display, Social Media
ClickZ Buyers Guide - Bid Management - PPC, Display, Social Media
Clark Boyd
 
360i Report: Big Data
360i Report: Big Data360i Report: Big Data
360i Report: Big Data360i
 
Rocket Fuel Big Data Report
Rocket Fuel Big Data ReportRocket Fuel Big Data Report
Rocket Fuel Big Data ReportCarat Turkiye
 
LUMA Digital Brief 010 - Power to the People
LUMA Digital Brief 010 - Power to the PeopleLUMA Digital Brief 010 - Power to the People
LUMA Digital Brief 010 - Power to the People
LUMA Partners
 
The display landscape
The display landscapeThe display landscape
The display landscapeAdCMO
 
E book mma_eng_short
E book mma_eng_shortE book mma_eng_short
E book mma_eng_short
Mobile Marketing Association
 
Think programmatic event_report_2014
Think programmatic event_report_2014Think programmatic event_report_2014
Think programmatic event_report_2014AdCMO
 
Emerging Media
Emerging MediaEmerging Media
Data - Its a big deal
Data - Its a big dealData - Its a big deal
Data - Its a big dealSubarna Gupta
 
CES 2015: A No Gadget Report for Marketers & Brands
CES 2015: A No Gadget Report for Marketers & BrandsCES 2015: A No Gadget Report for Marketers & Brands
CES 2015: A No Gadget Report for Marketers & Brands
Ben Grossman
 
Future of Digital Marketing [Free Download]
 Future of  Digital Marketing  [Free Download] Future of  Digital Marketing  [Free Download]
Future of Digital Marketing [Free Download]
Visitor Analytics
 
Analysis Of An Online Digital Marketing A Case Study
Analysis Of An Online Digital Marketing  A Case StudyAnalysis Of An Online Digital Marketing  A Case Study
Analysis Of An Online Digital Marketing A Case Study
Cheryl Brown
 
When Device Recognitio an Programmatic Buying Intersect
When Device Recognitio an Programmatic Buying IntersectWhen Device Recognitio an Programmatic Buying Intersect
When Device Recognitio an Programmatic Buying Intersect
AdTruth
 
Thinking_Beyond_the_Third_Party_Cookie_Guide_Refresh_Offer.pdf
Thinking_Beyond_the_Third_Party_Cookie_Guide_Refresh_Offer.pdfThinking_Beyond_the_Third_Party_Cookie_Guide_Refresh_Offer.pdf
Thinking_Beyond_the_Third_Party_Cookie_Guide_Refresh_Offer.pdf
PrasadraoBucche
 
What is big data ? | Big Data Applications
What is big data ? | Big Data ApplicationsWhat is big data ? | Big Data Applications
What is big data ? | Big Data Applications
ShilpaKrishna6
 
ClickZ/Fospha: The State of Marketing Measurement, Attribution, and Data Mana...
ClickZ/Fospha: The State of Marketing Measurement, Attribution, and Data Mana...ClickZ/Fospha: The State of Marketing Measurement, Attribution, and Data Mana...
ClickZ/Fospha: The State of Marketing Measurement, Attribution, and Data Mana...
Clark Boyd
 

Similar to The Impacts of Data on the Online Advertising Industry (20)

Evolution of Digital Marketing and its Impact on Privacy
Evolution of Digital Marketing and its Impact on PrivacyEvolution of Digital Marketing and its Impact on Privacy
Evolution of Digital Marketing and its Impact on Privacy
 
GroupM Brand Safety Playbook For Marketers
GroupM Brand Safety Playbook For MarketersGroupM Brand Safety Playbook For Marketers
GroupM Brand Safety Playbook For Marketers
 
GroupM Brand Safety Playbook 2020
GroupM Brand Safety Playbook 2020GroupM Brand Safety Playbook 2020
GroupM Brand Safety Playbook 2020
 
ClickZ Buyers Guide - Bid Management - PPC, Display, Social Media
ClickZ Buyers Guide - Bid Management - PPC, Display, Social MediaClickZ Buyers Guide - Bid Management - PPC, Display, Social Media
ClickZ Buyers Guide - Bid Management - PPC, Display, Social Media
 
360i Report: Big Data
360i Report: Big Data360i Report: Big Data
360i Report: Big Data
 
Rocket Fuel Big Data Report
Rocket Fuel Big Data ReportRocket Fuel Big Data Report
Rocket Fuel Big Data Report
 
LUMA Digital Brief 010 - Power to the People
LUMA Digital Brief 010 - Power to the PeopleLUMA Digital Brief 010 - Power to the People
LUMA Digital Brief 010 - Power to the People
 
The display landscape
The display landscapeThe display landscape
The display landscape
 
E book mma_eng_short
E book mma_eng_shortE book mma_eng_short
E book mma_eng_short
 
Think programmatic event_report_2014
Think programmatic event_report_2014Think programmatic event_report_2014
Think programmatic event_report_2014
 
Big Group
Big GroupBig Group
Big Group
 
Emerging Media
Emerging MediaEmerging Media
Emerging Media
 
Data - Its a big deal
Data - Its a big dealData - Its a big deal
Data - Its a big deal
 
CES 2015: A No Gadget Report for Marketers & Brands
CES 2015: A No Gadget Report for Marketers & BrandsCES 2015: A No Gadget Report for Marketers & Brands
CES 2015: A No Gadget Report for Marketers & Brands
 
Future of Digital Marketing [Free Download]
 Future of  Digital Marketing  [Free Download] Future of  Digital Marketing  [Free Download]
Future of Digital Marketing [Free Download]
 
Analysis Of An Online Digital Marketing A Case Study
Analysis Of An Online Digital Marketing  A Case StudyAnalysis Of An Online Digital Marketing  A Case Study
Analysis Of An Online Digital Marketing A Case Study
 
When Device Recognitio an Programmatic Buying Intersect
When Device Recognitio an Programmatic Buying IntersectWhen Device Recognitio an Programmatic Buying Intersect
When Device Recognitio an Programmatic Buying Intersect
 
Thinking_Beyond_the_Third_Party_Cookie_Guide_Refresh_Offer.pdf
Thinking_Beyond_the_Third_Party_Cookie_Guide_Refresh_Offer.pdfThinking_Beyond_the_Third_Party_Cookie_Guide_Refresh_Offer.pdf
Thinking_Beyond_the_Third_Party_Cookie_Guide_Refresh_Offer.pdf
 
What is big data ? | Big Data Applications
What is big data ? | Big Data ApplicationsWhat is big data ? | Big Data Applications
What is big data ? | Big Data Applications
 
ClickZ/Fospha: The State of Marketing Measurement, Attribution, and Data Mana...
ClickZ/Fospha: The State of Marketing Measurement, Attribution, and Data Mana...ClickZ/Fospha: The State of Marketing Measurement, Attribution, and Data Mana...
ClickZ/Fospha: The State of Marketing Measurement, Attribution, and Data Mana...
 

More from Mutlu Dogus Yildirim

Anadolu Üniversitesi - Programatik Reklamcılık Eğitimi
Anadolu Üniversitesi - Programatik Reklamcılık EğitimiAnadolu Üniversitesi - Programatik Reklamcılık Eğitimi
Anadolu Üniversitesi - Programatik Reklamcılık Eğitimi
Mutlu Dogus Yildirim
 
Posta Gazetesi - Programatik Reklamcilik Egitimi
Posta Gazetesi - Programatik Reklamcilik EgitimiPosta Gazetesi - Programatik Reklamcilik Egitimi
Posta Gazetesi - Programatik Reklamcilik Egitimi
Mutlu Dogus Yildirim
 
GYODER Programatik Reklamcılık Eğitimi
GYODER Programatik Reklamcılık EğitimiGYODER Programatik Reklamcılık Eğitimi
GYODER Programatik Reklamcılık Eğitimi
Mutlu Dogus Yildirim
 
Data & Programmatic Advertising - IAB
Data & Programmatic Advertising - IABData & Programmatic Advertising - IAB
Data & Programmatic Advertising - IAB
Mutlu Dogus Yildirim
 
Veri Kullanımı ve Programatik Reklamcılık- IAB
Veri Kullanımı ve Programatik Reklamcılık- IABVeri Kullanımı ve Programatik Reklamcılık- IAB
Veri Kullanımı ve Programatik Reklamcılık- IAB
Mutlu Dogus Yildirim
 
IAB Turkey - Programmatic Advertising Training by Mutlu Dogus Yildirim
IAB Turkey - Programmatic Advertising Training by Mutlu Dogus YildirimIAB Turkey - Programmatic Advertising Training by Mutlu Dogus Yildirim
IAB Turkey - Programmatic Advertising Training by Mutlu Dogus Yildirim
Mutlu Dogus Yildirim
 
IAB Turkiye Programatik Reklamcilik Egitim Sunumu
IAB Turkiye Programatik Reklamcilik Egitim SunumuIAB Turkiye Programatik Reklamcilik Egitim Sunumu
IAB Turkiye Programatik Reklamcilik Egitim Sunumu
Mutlu Dogus Yildirim
 
Digital advertising Training V2
Digital advertising Training V2Digital advertising Training V2
Digital advertising Training V2
Mutlu Dogus Yildirim
 
MedyaNet's Presentation to Turkish E-Commerce Marketers
MedyaNet's Presentation to Turkish E-Commerce MarketersMedyaNet's Presentation to Turkish E-Commerce Marketers
MedyaNet's Presentation to Turkish E-Commerce Marketers
Mutlu Dogus Yildirim
 
Real Time Bidding (RTB) Advantages
Real Time Bidding (RTB) AdvantagesReal Time Bidding (RTB) Advantages
Real Time Bidding (RTB) Advantages
Mutlu Dogus Yildirim
 
Digital Advertising Training Presentation
Digital Advertising Training PresentationDigital Advertising Training Presentation
Digital Advertising Training Presentation
Mutlu Dogus Yildirim
 
Real time bidding (DSP, SSP, DMP, ATD, ITD)
Real time bidding (DSP, SSP, DMP, ATD, ITD)Real time bidding (DSP, SSP, DMP, ATD, ITD)
Real time bidding (DSP, SSP, DMP, ATD, ITD)
Mutlu Dogus Yildirim
 
What is DSP & SSP? (Real Time Bidding)
What is DSP & SSP? (Real Time Bidding)What is DSP & SSP? (Real Time Bidding)
What is DSP & SSP? (Real Time Bidding)
Mutlu Dogus Yildirim
 

More from Mutlu Dogus Yildirim (13)

Anadolu Üniversitesi - Programatik Reklamcılık Eğitimi
Anadolu Üniversitesi - Programatik Reklamcılık EğitimiAnadolu Üniversitesi - Programatik Reklamcılık Eğitimi
Anadolu Üniversitesi - Programatik Reklamcılık Eğitimi
 
Posta Gazetesi - Programatik Reklamcilik Egitimi
Posta Gazetesi - Programatik Reklamcilik EgitimiPosta Gazetesi - Programatik Reklamcilik Egitimi
Posta Gazetesi - Programatik Reklamcilik Egitimi
 
GYODER Programatik Reklamcılık Eğitimi
GYODER Programatik Reklamcılık EğitimiGYODER Programatik Reklamcılık Eğitimi
GYODER Programatik Reklamcılık Eğitimi
 
Data & Programmatic Advertising - IAB
Data & Programmatic Advertising - IABData & Programmatic Advertising - IAB
Data & Programmatic Advertising - IAB
 
Veri Kullanımı ve Programatik Reklamcılık- IAB
Veri Kullanımı ve Programatik Reklamcılık- IABVeri Kullanımı ve Programatik Reklamcılık- IAB
Veri Kullanımı ve Programatik Reklamcılık- IAB
 
IAB Turkey - Programmatic Advertising Training by Mutlu Dogus Yildirim
IAB Turkey - Programmatic Advertising Training by Mutlu Dogus YildirimIAB Turkey - Programmatic Advertising Training by Mutlu Dogus Yildirim
IAB Turkey - Programmatic Advertising Training by Mutlu Dogus Yildirim
 
IAB Turkiye Programatik Reklamcilik Egitim Sunumu
IAB Turkiye Programatik Reklamcilik Egitim SunumuIAB Turkiye Programatik Reklamcilik Egitim Sunumu
IAB Turkiye Programatik Reklamcilik Egitim Sunumu
 
Digital advertising Training V2
Digital advertising Training V2Digital advertising Training V2
Digital advertising Training V2
 
MedyaNet's Presentation to Turkish E-Commerce Marketers
MedyaNet's Presentation to Turkish E-Commerce MarketersMedyaNet's Presentation to Turkish E-Commerce Marketers
MedyaNet's Presentation to Turkish E-Commerce Marketers
 
Real Time Bidding (RTB) Advantages
Real Time Bidding (RTB) AdvantagesReal Time Bidding (RTB) Advantages
Real Time Bidding (RTB) Advantages
 
Digital Advertising Training Presentation
Digital Advertising Training PresentationDigital Advertising Training Presentation
Digital Advertising Training Presentation
 
Real time bidding (DSP, SSP, DMP, ATD, ITD)
Real time bidding (DSP, SSP, DMP, ATD, ITD)Real time bidding (DSP, SSP, DMP, ATD, ITD)
Real time bidding (DSP, SSP, DMP, ATD, ITD)
 
What is DSP & SSP? (Real Time Bidding)
What is DSP & SSP? (Real Time Bidding)What is DSP & SSP? (Real Time Bidding)
What is DSP & SSP? (Real Time Bidding)
 

Recently uploaded

Sustainability: Balancing the Environment, Equity & Economy
Sustainability: Balancing the Environment, Equity & EconomySustainability: Balancing the Environment, Equity & Economy
Sustainability: Balancing the Environment, Equity & Economy
Operational Excellence Consulting
 
Kseniya Leshchenko: Shared development support service model as the way to ma...
Kseniya Leshchenko: Shared development support service model as the way to ma...Kseniya Leshchenko: Shared development support service model as the way to ma...
Kseniya Leshchenko: Shared development support service model as the way to ma...
Lviv Startup Club
 
Authentically Social Presented by Corey Perlman
Authentically Social Presented by Corey PerlmanAuthentically Social Presented by Corey Perlman
Authentically Social Presented by Corey Perlman
Corey Perlman, Social Media Speaker and Consultant
 
Digital Transformation and IT Strategy Toolkit and Templates
Digital Transformation and IT Strategy Toolkit and TemplatesDigital Transformation and IT Strategy Toolkit and Templates
Digital Transformation and IT Strategy Toolkit and Templates
Aurelien Domont, MBA
 
Project File Report BBA 6th semester.pdf
Project File Report BBA 6th semester.pdfProject File Report BBA 6th semester.pdf
Project File Report BBA 6th semester.pdf
RajPriye
 
Business Valuation Principles for Entrepreneurs
Business Valuation Principles for EntrepreneursBusiness Valuation Principles for Entrepreneurs
Business Valuation Principles for Entrepreneurs
Ben Wann
 
Maksym Vyshnivetskyi: PMO Quality Management (UA)
Maksym Vyshnivetskyi: PMO Quality Management (UA)Maksym Vyshnivetskyi: PMO Quality Management (UA)
Maksym Vyshnivetskyi: PMO Quality Management (UA)
Lviv Startup Club
 
ikea_woodgreen_petscharity_cat-alogue_digital.pdf
ikea_woodgreen_petscharity_cat-alogue_digital.pdfikea_woodgreen_petscharity_cat-alogue_digital.pdf
ikea_woodgreen_petscharity_cat-alogue_digital.pdf
agatadrynko
 
Improving profitability for small business
Improving profitability for small businessImproving profitability for small business
Improving profitability for small business
Ben Wann
 
amptalk_RecruitingDeck_english_2024.06.05
amptalk_RecruitingDeck_english_2024.06.05amptalk_RecruitingDeck_english_2024.06.05
amptalk_RecruitingDeck_english_2024.06.05
marketing317746
 
Discover the innovative and creative projects that highlight my journey throu...
Discover the innovative and creative projects that highlight my journey throu...Discover the innovative and creative projects that highlight my journey throu...
Discover the innovative and creative projects that highlight my journey throu...
dylandmeas
 
The Influence of Marketing Strategy and Market Competition on Business Perfor...
The Influence of Marketing Strategy and Market Competition on Business Perfor...The Influence of Marketing Strategy and Market Competition on Business Perfor...
The Influence of Marketing Strategy and Market Competition on Business Perfor...
Adam Smith
 
Bài tập - Tiếng anh 11 Global Success UNIT 1 - Bản HS.doc.pdf
Bài tập - Tiếng anh 11 Global Success UNIT 1 - Bản HS.doc.pdfBài tập - Tiếng anh 11 Global Success UNIT 1 - Bản HS.doc.pdf
Bài tập - Tiếng anh 11 Global Success UNIT 1 - Bản HS.doc.pdf
daothibichhang1
 
Set off and carry forward of losses and assessment of individuals.pptx
Set off and carry forward of losses and assessment of individuals.pptxSet off and carry forward of losses and assessment of individuals.pptx
Set off and carry forward of losses and assessment of individuals.pptx
HARSHITHV26
 
Premium MEAN Stack Development Solutions for Modern Businesses
Premium MEAN Stack Development Solutions for Modern BusinessesPremium MEAN Stack Development Solutions for Modern Businesses
Premium MEAN Stack Development Solutions for Modern Businesses
SynapseIndia
 
ModelingMarketingStrategiesMKS.CollumbiaUniversitypdf
ModelingMarketingStrategiesMKS.CollumbiaUniversitypdfModelingMarketingStrategiesMKS.CollumbiaUniversitypdf
ModelingMarketingStrategiesMKS.CollumbiaUniversitypdf
fisherameliaisabella
 
VAT Registration Outlined In UAE: Benefits and Requirements
VAT Registration Outlined In UAE: Benefits and RequirementsVAT Registration Outlined In UAE: Benefits and Requirements
VAT Registration Outlined In UAE: Benefits and Requirements
uae taxgpt
 
Recruiting in the Digital Age: A Social Media Masterclass
Recruiting in the Digital Age: A Social Media MasterclassRecruiting in the Digital Age: A Social Media Masterclass
Recruiting in the Digital Age: A Social Media Masterclass
LuanWise
 
Event Report - SAP Sapphire 2024 Orlando - lots of innovation and old challenges
Event Report - SAP Sapphire 2024 Orlando - lots of innovation and old challengesEvent Report - SAP Sapphire 2024 Orlando - lots of innovation and old challenges
Event Report - SAP Sapphire 2024 Orlando - lots of innovation and old challenges
Holger Mueller
 
Brand Analysis for an artist named Struan
Brand Analysis for an artist named StruanBrand Analysis for an artist named Struan
Brand Analysis for an artist named Struan
sarahvanessa51503
 

Recently uploaded (20)

Sustainability: Balancing the Environment, Equity & Economy
Sustainability: Balancing the Environment, Equity & EconomySustainability: Balancing the Environment, Equity & Economy
Sustainability: Balancing the Environment, Equity & Economy
 
Kseniya Leshchenko: Shared development support service model as the way to ma...
Kseniya Leshchenko: Shared development support service model as the way to ma...Kseniya Leshchenko: Shared development support service model as the way to ma...
Kseniya Leshchenko: Shared development support service model as the way to ma...
 
Authentically Social Presented by Corey Perlman
Authentically Social Presented by Corey PerlmanAuthentically Social Presented by Corey Perlman
Authentically Social Presented by Corey Perlman
 
Digital Transformation and IT Strategy Toolkit and Templates
Digital Transformation and IT Strategy Toolkit and TemplatesDigital Transformation and IT Strategy Toolkit and Templates
Digital Transformation and IT Strategy Toolkit and Templates
 
Project File Report BBA 6th semester.pdf
Project File Report BBA 6th semester.pdfProject File Report BBA 6th semester.pdf
Project File Report BBA 6th semester.pdf
 
Business Valuation Principles for Entrepreneurs
Business Valuation Principles for EntrepreneursBusiness Valuation Principles for Entrepreneurs
Business Valuation Principles for Entrepreneurs
 
Maksym Vyshnivetskyi: PMO Quality Management (UA)
Maksym Vyshnivetskyi: PMO Quality Management (UA)Maksym Vyshnivetskyi: PMO Quality Management (UA)
Maksym Vyshnivetskyi: PMO Quality Management (UA)
 
ikea_woodgreen_petscharity_cat-alogue_digital.pdf
ikea_woodgreen_petscharity_cat-alogue_digital.pdfikea_woodgreen_petscharity_cat-alogue_digital.pdf
ikea_woodgreen_petscharity_cat-alogue_digital.pdf
 
Improving profitability for small business
Improving profitability for small businessImproving profitability for small business
Improving profitability for small business
 
amptalk_RecruitingDeck_english_2024.06.05
amptalk_RecruitingDeck_english_2024.06.05amptalk_RecruitingDeck_english_2024.06.05
amptalk_RecruitingDeck_english_2024.06.05
 
Discover the innovative and creative projects that highlight my journey throu...
Discover the innovative and creative projects that highlight my journey throu...Discover the innovative and creative projects that highlight my journey throu...
Discover the innovative and creative projects that highlight my journey throu...
 
The Influence of Marketing Strategy and Market Competition on Business Perfor...
The Influence of Marketing Strategy and Market Competition on Business Perfor...The Influence of Marketing Strategy and Market Competition on Business Perfor...
The Influence of Marketing Strategy and Market Competition on Business Perfor...
 
Bài tập - Tiếng anh 11 Global Success UNIT 1 - Bản HS.doc.pdf
Bài tập - Tiếng anh 11 Global Success UNIT 1 - Bản HS.doc.pdfBài tập - Tiếng anh 11 Global Success UNIT 1 - Bản HS.doc.pdf
Bài tập - Tiếng anh 11 Global Success UNIT 1 - Bản HS.doc.pdf
 
Set off and carry forward of losses and assessment of individuals.pptx
Set off and carry forward of losses and assessment of individuals.pptxSet off and carry forward of losses and assessment of individuals.pptx
Set off and carry forward of losses and assessment of individuals.pptx
 
Premium MEAN Stack Development Solutions for Modern Businesses
Premium MEAN Stack Development Solutions for Modern BusinessesPremium MEAN Stack Development Solutions for Modern Businesses
Premium MEAN Stack Development Solutions for Modern Businesses
 
ModelingMarketingStrategiesMKS.CollumbiaUniversitypdf
ModelingMarketingStrategiesMKS.CollumbiaUniversitypdfModelingMarketingStrategiesMKS.CollumbiaUniversitypdf
ModelingMarketingStrategiesMKS.CollumbiaUniversitypdf
 
VAT Registration Outlined In UAE: Benefits and Requirements
VAT Registration Outlined In UAE: Benefits and RequirementsVAT Registration Outlined In UAE: Benefits and Requirements
VAT Registration Outlined In UAE: Benefits and Requirements
 
Recruiting in the Digital Age: A Social Media Masterclass
Recruiting in the Digital Age: A Social Media MasterclassRecruiting in the Digital Age: A Social Media Masterclass
Recruiting in the Digital Age: A Social Media Masterclass
 
Event Report - SAP Sapphire 2024 Orlando - lots of innovation and old challenges
Event Report - SAP Sapphire 2024 Orlando - lots of innovation and old challengesEvent Report - SAP Sapphire 2024 Orlando - lots of innovation and old challenges
Event Report - SAP Sapphire 2024 Orlando - lots of innovation and old challenges
 
Brand Analysis for an artist named Struan
Brand Analysis for an artist named StruanBrand Analysis for an artist named Struan
Brand Analysis for an artist named Struan
 

The Impacts of Data on the Online Advertising Industry

  • 1. IMPERIAL COLLEGE LONDON The impacts of data on the online advertising industry by Mutlu Dogus Yildirim A Project submitted in partial fulfilment of the requirements for the MBA degree July 2016
  • 2. II SYNOPSIS Data is a game-changer in the online advertising industry. Companies that did not exist several years ago generate more revenues than The New York Times now. WPP, the largest advertising agency in the world, wants to be known as a “data” company. Google and Facebook generate their incomes mostly through online advertising by utilising data, and they are among the most valuable companies in the world now. In Germany, rival publishers are about to start a “data alliance”. Meanwhile, governments try to regulate the usage of data, and many people use ad-blockers to protect their privacies. Still, many advertising companies track users collectively, and companies like MasterCard sell user data to advertisers. One of the goals of this research is to investigate how data is used in online advertising. One of the main findings is that companies use many techniques to track users and collect many types of data like search, behavioural or location data, then this data is used for purposes like targeting, personalisation, measurement, optimisation and attribution. Another goal of this research is to address the benefits and risks for the players in the ecosystem. One important conclusion is that there are indeed many benefits as well as risks for publishers, advertisers, agencies, technology providers and users, and all parties should analyse these benefits and risks thoroughly. The final goal is to assess impacts of data on the online advertising industry. Some of the findings show that data has a huge impact in the industry, from creating the most valuable companies in the world, to data alliances among rival companies, to legislations and ad-blockers which may force the advertising companies to change their business models.
  • 3. III ACKNOWLEDGEMENTS I would like to thank my supervisor Dr John Beasley for guiding me with his valuable feedbacks. I would like to thank my associate Mr Alp Ayhan for his great help. Finally, I would like to express my deepest appreciation to my wife Maria for all of her support. Without her, I could never finish this project.
  • 4. IV LIST OF CONTENT Page 1 INTRODUCTION 1 1.1 BACKGROUND 1 1.1.1 What Does Data Mean for Advertising? 1 1.1.2 Is Data Really Important? 2 1.2 LITERATURE REVIEW 5 1.3 OBJECTIVES 6 2 APPLICATIONS OF DATA IN THE ONLINE ADVERTISING INDUSTRY 7 2.1 TYPES OF DATA 7 2.1.1 Search 7 2.1.2 Contextual 7 2.1.3 Intent 8 2.1.4 Image & Video 9 2.1.5 Ethnic 9 2.1.6 Location 10 2.1.7 Social Media 10 2.1.8 Social Sharing 11 2.1.9 Business 11 2.1.10 CRM 12 2.1.11 Offline 12 2.1.12 Behavioural 13 2.2 CONNECTING TO USER 14 2.2.1 Cookies 15 2.2.2 Flash Cookies 15 2.2.3 IP Address 15 2.2.4 Browser Fingerprinting 16 2.2.5 Local Storage 16 2.2.6 Canvas Fingerprinting 16 2.2.7 Carrier Injected Headers 16 2.2.8 Evercookie 17 2.2.9 ISP-Level Tracking 17 2.2.10 Others 17 2.3 DATA QUALITY 17 2.3.1 One Device & Multiple Users 18 2.3.2 One User & Multiple Devices 18 2.3.3 Deleted Cookies 19 2.3.4 Data Loss 19 2.3.5 Recency 19 2.3.6 Missing Data Points 20 2.3.7 Poor Integration / Implementation 20 2.3.8 Technological Challenges 20 2.4 SEGMENTATION 21 2.5 USAGE OF DATA 21 2.5.1 Traditional Advertising 21 2.5.2 Programmatic Advertising 22 2.5.3 Targeted Media Buying 24 2.5.4 Campaign Measurement & Optimisation 28 2.5.5 Attribution 29 2.5.6 Yield Optimisation 30
  • 5. V 2.5.7 Audience Optimisation 30 2.5.8 Data Sales 30 2.5.9 Frequency Optimisation 33 2.6 FRAUD DETECTION 33 2.7 VIEWABILITY 34 2.8 MOBILE-SPECIFIC CHALLENGES & BENEFITS 34 2.8.1 Publishers 35 2.8.2 Advertisers 35 2.8.3 Cookies 36 2.8.4 Other Tracking Methods 36 2.8.5 Benefits of Mobile Devices 37 3 BENEFITS AND RISKS FOR THE PLAYERS IN THE ECOSYSTEM 38 3.1 BENEFITS FOR PUBLISHERS 38 3.1.1 Data + Media Bundle Sales 39 3.1.2 Data Sales 39 3.1.3 Data Partnerships 40 3.1.4 Audience Enrichment 40 3.2 BENEFITS FOR ADVERTISERS 41 3.2.1 Targeting a Specific Audience 41 3.2.2 Banner Optimisation 42 3.2.3 Understanding Customers Better 42 3.2.4 Viewability 42 3.2.5 Fraud 43 3.2.6 Spending Budgets Efficiently 43 3.2.7 Brand Safety 43 3.2.8 Data Sales 44 3.3 BENEFITS FOR ADVERTISING AGENCIES 44 3.3.1 Optimising Media Investments 45 3.3.2 Tailoring Relevant Messages 45 3.3.3 ProvideAccountability and Knowledge to Marketers 45 3.4 BENEFITS FOR TECHNOLOGY PROVIDERS 45 3.4.1 Data at the Core 45 3.4.2 New Business Opportunities 46 3.5 BENEFITS FOR USERS 46 3.5.1 Free Content 47 3.5.2 Inevitability 48 3.5.3 Discounts & Special Offers 49 3.5.4 Additional Benefits 49 3.6 RISKS FOR PUBLISHERS 49 3.6.1 Privacy Concerns 49 3.6.2 From Publisher to Audience 49 3.6.3 Too Targeted = Too Low Spending? 50 3.6.4 Information Asymmetry 50 3.6.5 Data Leakage 50 3.6.6 Impact on Revenues 51 3.7 RISKS FOR ADVERTISERS 51 3.7.1 Being Annoying 51 3.7.2 Data Leakage 52 3.7.3 Increasing Complexity 52 3.7.4 Impacts on HR 53 3.7.5 Impacts on Advertising Costs 54
  • 6. VI 3.8 RISKS FOR ADVERTISING AGENCIES 55 3.8.1 Increasing Complexity 55 3.8.2 Low Spending 55 3.8.3 Benefits(!) of Fraud 56 3.8.4 Impacts on HR and Mindsetsof Employees 56 3.8.5 Over-relianceon Technology and Data 56 3.9 RISKS FOR TECHNOLOGY PROVIDERS 56 3.9.1 Missing Data Points 57 3.9.2 Crowded Industry 57 3.10 RISKS FOR USERS 57 3.10.1 Privacy 57 3.10.2 Increased Spending 58 4 IMPACTS ON THE INDUSTRY 60 4.1 PRIVACY CONCERNS 60 4.1.1 Legislations 61 4.1.2 Targeting Children 62 4.1.3 Other Sensitive Data 62 4.1.4 Self-Regulation 63 4.1.5 Do-Not-Track 63 4.1.6 Ad-Blockers 64 4.2 MAJOR IMPACTS OF DATA ON THE INDUSTRY 64 4.2.1 Google and Facebook 65 4.2.2 Other Companies 66 4.2.3 Publisher Partnerships & Alliances 67 4.2.4 Telecommunication Companies 67 4.2.5 Credit Card Companies 68 4.2.6 Purely Data-Focused Players 68 4.2.7 Ad Blockers 69 4.2.8 Big Data Technologies 69 4.2.9 Mergers and Acquisitions 70 4.2.10 New Business Models 70 4.3 NEAR FUTURE 71 4.3.1 TV Advertising 71 4.3.2 Internet of Things (IOT) 72 4.3.3 Virtual/Augmented Reality 72 4.3.4 Outdoor & Radio Advertising 72 4.3.5 Cross-Device Attribution / Targeting 73 4.3.6 Converging Technologies 73 4.3.7 Improving Technologies & Decreasing Costs 73 4.3.8 Change in Employee Skills 74 4.3.9 Privacy Concerns & Legislations 74 5 CONCLUSIONS 75 6 FUTURE WORK 77 7 REFERENCES 78
  • 7. 1 1 INTRODUCTION 1.1 Background In the last few years, people have created more data than in the entire human history (Marr, 2015). This data is analysed and used in many different ways in many industries, but for one of them, the advertising industry, it has a special value. John Wanamaker famously said, “Half my advertising is wasted, I just don't know which half.” (AdvertisingAge, 1999). Data helps advertisers to understand how their advertising budgets are spent, who has seen their ads and how many times, did it result in any action and so on. This information gives advertisers great insights regarding their advertising campaigns and helps to reduce the money they “waste”. In fact, the advertising industry, online advertising, in particular, seems to be “have a crush” on data (Marshall, 2013a). It affected the industry so much, that the largest advertising agency company in the world, WPP, does not want to be known as an “advertising” company anymore, but instead wants to be seen as a “data” company (Marshall, 2013b). 1.1.1 What Does Data Mean for Advertising? In mid-2009, McKinsey suggested that consumers are changing the ways they research and buy products. They suggested that one of the things marketers should do is to “customise advertising by viewing the context and the consumer” (Court et al., 2009). The good news is; current technologies allow advertisers to do just that - by using data. In the online advertising world, data can be collected from many sources like what product the user is viewing on the advertiser’s website, what s/he reads on a news publisher’s website, his/her location and so on. Mostly users generate this data, so this research will mainly be concerned with user data.
  • 8. 2 However, there are other types of data sources as well, like the weather condition, which can be useful to advertise an umbrella for instance. Hence, some of these sources will also be discussed in this project. It is possible to change the messages in real-time considering the audience or context - as suggested by McKinsey, which is possible only through data. It will be discussed later in more detail, but in the online advertising world, data can especially be helpful with targeting, personalisation, measurement, optimisation and attribution. By using data from various sources, it is possible to show the ads to a specific audience like “women aged between 18-34” or “men who searched for Audi cars recently”. The messages can even be tailored at the individual level, and a 10% discount can be offered only to those who visited advertiser’s website and spent more than 3 minutes exploring a product for instance. Thus, the term “data-driven advertising” became popular in the industry (GlobalDMA, 2014). 1.1.2 Is Data Really Important? Two examples; Google and Facebook, prove the importance of data in online advertising. The famous search engine, Google, was born in 1998 (Google, n.d.g). In 2000, Google introduced AdWords, a self-service tool for creating online advertising campaigns. In 2004, after going public, Google was valued at $23 billion (Ritter, 2014). Today, Google’s market capitalisation is $469.05 billion (Yahoo, 2016a) and it is one of the most valuable brands in the world (Forbes, 2016). Around 90% of Google’s revenue comes from advertising (Statista, 2016a), which amounted to $67,39 billion in 2015. The fast growth of Google in years can be seen in Figure 1.1.
  • 9. 3 Figure 1.1 Google advertising revenues (Statista, 2016b) Google achieved this big success thanks to AdWords – a platform which allows advertisers to reach people as they search for some specific words or phrases and who are in specific geographical areas. It also enables measurement and optimisation of campaigns in real-time (Google, n.d.a). These capabilities, which are possible thanks to search and geographical data of the users, are very precious to advertisers (Rampton, 2014). Facebook is another good example of how data helped a company grow very quickly and become one of the biggest companies in the world. It was founded in 2004 (Facebook, n.d.a) and held its initial public offering (IPO) in 2012. During its IPO, it was valued at $104 billion, which was the largest ever valuation of a U.S. company at the time of its offering. Today, Facebook’s market capitalisation is $290.35 billion (Yahoo, 2016b). Facebook’s success in advertising is related directly to data. Facebook knows where its users live, what they like, who their friends are, their age and so on, which are extremely valuable to the advertisers (Martin, 2012). Hence, advertisers shifted much of their budgets to Facebook advertising campaigns, and Facebook generated $17.93 billion revenue in 2015, from which $17.08 billion comes from advertising, as can be seen in Figure 1.2 (Statista, 2016c).
  • 10. 4 Figure 1.2 Facebook advertising revenue (Statista, 2016c) Facebook is expected to grow even further thanks to the continuously advancing technologies in targeted advertising, which is helping them to be way ahead of their competitors (DiChristopher, 2015). Obviously, data is not the only reason why Google and Facebook became so successful; there are many other factors too (e.g. product quality, the number of users, etc.). However, the shift in advertising budgets so quickly and so substantially can only be explained by data and its benefits to the advertisers. Data does not only help Google and Facebook; it also helps to the whole online advertising industry and the players in it. In 2017, online advertising spending is expected to surpass TV spending in the US for the first time in history (Nicolaou, 2016). Comcast - a broadcasting and cable television giant, is planning to give advertisers some targeting capabilities by utilising data (Levy, 2016b), like their “purely digital” rivals do. Criteo, a company that offers personalised advertising products by utilising data, has a market capitalisation of $2.8 billion (Bloomberg, n.d.). Amazon.com generated around $850 million from advertising in the US in 2015 (Statista, 2016e), whereas The New York Times generated only $638 million (The New York
  • 11. 5 Times Company, 2015). Hence, traditional publishers around the world like Financial Times, The Guardian and Reuters join their forces and create partnerships to combine their data and offer advertisers better products (Bilton, 2016). Meanwhile, companies like MasterCard or Visa sell data to advertisers to allow them to target a specific audience with specific purchasing behaviours. As can be seen, data has been shaping an entire industry quite a bit. As Economist puts it, data is crucial to the online advertising economy (Koen, 2014). 1.2 Literature Review There are many studies on data and online advertising in the literature. Evans (2009) discusses the economic benefits of online advertising, which can be communicated to the most relevant people, and how it disrupts traditional advertising methods. Fulgoni (2013) discusses different ways marketers can take advantage of (big) data, where Chen et al. (2014) discuss how to maximise revenues for publishers. Gironda (2014) tries to analyse consumer attitudes towards personalised ads. Langheinrich et al. (1999) show that the usage of data started at the early ages of the internet, and they suggest ways of “customisation” of ads without being intrusive. Englehardt et al. (2016) identify many different techniques of tracking users and show how different companies share data with each other. Karlsson et al. (2015) discuss why marketers should be transparent about how and why they collect personal information. Goldfarb et al. (2010) analyse how privacy regulations affect campaign effectiveness negatively – around 65% reduction where EU privacy laws are applied for instance. Toubiana et al. (2010) come up with a new approach, which allows behavioural tracking while respecting user privacy.
  • 12. 6 Actually, behavioural advertising and privacy related issues are researched quite well by many others (Mayer and Mitchell (2012), Ur et al. (2012), Hoofnagle et al. (2012), Corniere and de Nijs (2013)). 1.3 Objectives The purpose of this research is to extend the overall knowledge of data and its usage in the online advertising industry, and mainly answer the questions below: 1- How is data used in online advertising? 2- What are the benefits and risks of data for the major players in the online advertising industry? 3- What is the impact of data to the online advertising industry?
  • 13. 7 2 APPLICATIONS OF DATA IN THE ONLINE ADVERTISING INDUSTRY 2.1 Types of Data There are many types of data that are useful to advertisers. Some of them are briefly explained below. 2.1.1 Search One of the important data points for advertisers is search data, which shows what the user is actively looking for. It is the type of data that made Google so big and powerful, as Google shows ads to people who look for specific information (e.g. audi a3 price), and this is much more efficient than showing ads to people who are not interested in anything specific (Rampton, 2014). Search data is not limited to queries made on search engines like Google. People also search within websites. For instance, someone can visit Amazon.com and search for a product. This data, along with others, can be used by Amazon to show its visitors some personalised ads, and it is one of the reasons why Amazon generates more revenue than The New York Times from advertising business (Liyakasa, 2015). The same logic can be applied to a newspaper or to any other website, where the search queries are used to segment users and for ad optimisation purposes. Chango, a company who specialised in this area, was acquired by Rubicon Project for $122 million (Rubicon Project, 2015). 2.1.2 Contextual Contextual data is about the content a user views (Zhang & Katona, 2011). For instance, if someone visits Theguardian.com and starts reading an article called “Zurich city guide”, this might indicate that s/he could potentially be interested in visiting Zurich. Therefore, it may be a good idea to show this person a Zurich hotel ad.
  • 14. 8 Contextual data can also be used to segment users based on the content they consume. For instance, people who read articles about “Samsung smartphones” can be grouped into a segment called “Samsung-smart-phone- prospects”, and they can be targeted with relevant ads, either while they read Samsung smartphone related content, or somewhere else. This can be done through advanced technologies like data management platforms (DMP), which will be covered later. Companies like Grapeshot collect contextual data from various sources, segment users, and sell these segments to the advertisers (GrapeShot, n.d). When an advertiser buys a user segment from a company like Grapeshot, the advertiser knows that this segment consumed relevant content; hence, they have a potential interest in the advertiser’s product or service. There are also ad networks like Clicksor that allow their advertisers do contextual targeting on the publishers. 2.1.3 Intent Intent data indicates if a person is in-market for a product or service (Sharp, 2015). This is an important type of data since it can connect the advertiser directly to the potential buyer. One of the most powerful types of intent data is “search” data. When someone searches for “Audi A4 prices” on Google, it is a signal that s/he is potentially interested in buying that car. Intent data can also be classified as internal and external intent data: ● Internal intent data is captured by the actions of the user made on company’s own properties ● External intent data is captured by other companies and enabled through either third parties or directly (Singh & Grennev, 2015)
  • 15. 9 For instance, if a user views different colours of Audi A4 on Audi’s website and spends more than ten minutes to review the features of the car, s/he is giving a strong signal regarding his/her intention, which can be considered as “internal intent” data. On the other hand, if a person goes to cars.com and compares Audi A4 with Volkswagen Passat, it is also a signal regarding his/her intention. Audi (or Volkswagen in this case) can obtain this “external intent” data either by partnering with cars.com or through third party companies and use it to target the relevant person. 2.1.4 Image & Video An interesting data point, which is hard to extract, is the content of the images. Understanding that a user is looking at a beach photo, for instance, would allow an advertiser to target him/her with an advertisement of a hotel. Companies like Visenze or Cortica are working on this problem and try to convert images to texts, which can be used for targeting purposes. Video content is another data point that can be used by advertisers to understand their customers better or show an ad at a particular time of a video. Picturesafe, a company that tries to recognise video speech and converting to text is one of the companies working on this problem. 2.1.5 Ethnic Some advertisers believe ethnicity is the key to people’s identity (The Economist, 2011). Dublis (2001) recommends that advertisers should have an understanding of the linguistic acculturation, ethno-cultural stereotypes, intensity of ethnic identification and language differentiations. Advertising agencies like Ethnicity work on how to communicate to different cultures even if they are in the same country. Companies like Ethnic Online Network allow advertisers to target a particular ethnic group. Facebook is one of the effective channels where advertisers are able to target people from different cultures (Boykiv, 2015).
  • 16. 10 2.1.6 Location An advertiser offering its products or services in only one country would not want to show its ads in another country. In the online advertising world, it is quite possible to show ads to irrelevant audiences from other countries. For instance, CNN.com can be visited from many countries - so an advertiser selling products only to people live in U.S. would have to target U.S. visitors on CNN.com specifically, otherwise, a portion of the advertising budget would be wasted. Nowadays especially with the high usage of mobile phones, advertising technologies can understand the location of a user and allow advertisers do “hyper-local targeting” (Improve Digital, n.d.; Kusek, 2014; Andrews: 2015). Facebook is working on using location data in new forms like understanding which stores the users visit after seeing the ads (Constine, 2016). Companies like Digital Element and Maxmind are providing data about the location of the users. 2.1.7 Social Media Social media advertising proved to be very useful for the marketers (DeMers, 2014). The market size is estimated to be around $24 billion worldwide, around 65% of this market belongs to the Facebook and it is expected to grow further in the next years (eMarketer, 2015b). The reason is simple: not only people spend a lot of time on social media channels, but they also provide a lot of valuable data about themselves. On Facebook, it is very easy to target women who are between 18-24, engaged and like music. Facebook knows this data because people give it to them. Actually, Facebook collects a lot of data from everyone, things people do, content they provide, their location and so on (Facebook, n.d.b). It is important to note that some social media channels like Snapchat do not allow targeting at an advanced level - yet. Moreover, Snapchat CEO Evan
  • 17. 11 Spiegel claims targeting belongs to the early days of internet advertising (Snapchat, 2015). Ironically, the first question Snapchat answers in the F.A.Q. section of their website is “What type of ad targeting does Snapchat offer?” (SnapChat, n.d.), and some industry experts think that they work on a new advertising technology allowing more targeting options to the advertisers (Sloane, 2016). 2.1.8 Social Sharing When people see something they like, dislike, important news, etc. they tend to share it with their friends or families who they think would also be interested. They can do it through email, social media channels like Twitter or SMS. It gives advertisers an idea about people’s intentions or interests. For instance, if a person shares news about dropping smartphone prices with his/her friends, the advertiser could assume that the person who shares the news, and the person who receives the shared message would be interested in buying a phone. Thus, showing smartphone ads to these people would make sense. Companies like RadiumOne use algorithms to analyse this social sharing data and make it actionable for the advertisers (RadiumOne, n.d.). Other companies like AddThis collect this data from the publishers and sell it to advertisers (AddThis, n.d.). 2.1.9 Business Unlike B2C companies, B2B companies would not have many potential customers, and they would want to show their ads only to a specific audience. Moreover, they would want to be able to tailor their messages considering the industry, position and seniority level of the person who sees their ads. For instance, a company called SAP serves in 26 industries; hence, when they run an advertising campaign, they want their messages to be tailored accordingly (Anon, 2015).
  • 18. 12 LinkedIn provides advertisers with different types of targeting options like job title and function, industry, and seniority level of the person who sees the ad (LinkedIn, n.d.). Other companies like Bombora and Multiview also provide different types of targeting options for B2B advertisers. 2.1.10 CRM A Customer Relationship Management (CRM) is a software that gives a holistic view of each customer’s interactions with the organisation (Cross, 2004). To be able to do that, companies collect different types of data from their existing/potential clients like their names, phone numbers, last time they made a purchase, how much they spent and so on. This information is very precious to advertisers as they collect it directly from channels like their own website, their call-centre or their store. Since the data is collected directly from their own sources, its probability to be more reliable and accurate is higher than others. It means, tailoring the advertising message considering this data would be very beneficial, and one of the examples show 390% ROI (Ravesencio, 2015). Companies like Adroll and Retargeter help with CRM targeting. It is also possible to use data-management-platforms (DMP) to activate CRM data, which will be discussed later. 2.1.11 Offline Many people still use offline channels to buy something and interact with the brands. This valuable “offline” data is becoming more actionable for “online” targeting (Vegliante, 2015). Connecting offline data to the online world is very attractive, and advertisers start to put more focus on that. Companies like Datalogix and Liveramp help
  • 19. 13 with this process, which is also known as “on-boarding” in the industry (Kaye, 2014). 2.1.12 Behavioural Current technologies allow advertisers and publishers to collect data from visitors about their online activities like how much time they spend on a particular page, what type of content they read, what kind of news they share with their friends and so on. This data is used to profile the visitors, and these profiles can be used to show relevant ads to different types of people (Ur at al., 2012). Behavioural data is related to the most of the topics mentioned previously. For instance, if a user searches for “iphone” on the internet, reads content about “smartphones”, likes or shares an iPhone-related news or clicks an iPhone ad, it could potentially mean that s/he is interested in buying an iPhone. Advertisers use this behavioural data to find and target potential customers. Likewise, publishers can use this data to show relevant ads to their visitors. It is also known as “interest-based advertising” (Internet Advertising Bureau UK, n.d.). Behavioural data can be collected from many different sources. News sites, price comparison sites, responses to the existing ads, time spent on a product page, etc. can be used to collect data from a user and create a profile about him/her. Behavioural advertising helps advertisers greatly, as it results in higher conversion to sales. In some cases, it even modifies consumer behaviours by changing users’ self-perceptions (Walker Reczek, Summers & Smith, 2016). In one study, it is observed that when people see an advertisement of an environmental product, they feel “greener” and they even tend to donate
  • 20. 14 money to an environmental charity (Summers, Smith & Walker Reczek, 2016). Moreover, behavioural data can be used to “predict” consumer behaviour and target them accordingly. Some people believe it is the future of advertising (Goodwin, 2014), and companies like Affectv are focusing in this area. Behavioural data can also be used to find similar audiences, which are discovered algorithmically. This is also called “look-alike targeting” (Rajeck, 2014). Some advertisers who have tried this targeting model achieved very good results (Goodfellow, 2014). Two companies, in particular, have access to a very rich consumer data through many different channels: Google and Facebook. Thanks to the tools or services like Google.com, Youtube.com, Google Analytics, Chrome, Gmail or Android, Google can collect a massive amount of data about its users. This data enables Google to be very powerful in online advertising space (Team, 2015). Similarly, thanks to the tools or services like Facebook.com, WhatsApp, Instagram or “Like” buttons, Facebook also collects a vast amount of data from its users and use it to personalise ads (Facebook, n.d.b). Behavioural data is crucial in the online advertising industry, and it can be expected to be more important in the future. 2.2 Connecting to User As discussed previously, there are many types of data that can be collected. But this data needs to be “connected” to its owner so that it would be meaningful. It also needs to be up-to-date to be more useful; hence, its owner should constantly be tracked.
  • 21. 15 There are many techniques to do these; some major ones are explained below. 2.2.1 Cookies Cookies are the most popular method to track people online. It is so popular that European Commission has a dedicated page about cookies describing them as a piece of data websites ask browsers to store on people’s computers or mobile devices. This data can be about many things like remembering users’ preferences, identify them, helping them complete tasks without re-entering data and so on. There are different types of cookies like the session cookie which is deleted when the user closes the browser, the permanent cookie which is not deleted until a pre-determined date, the first-party cookie which is set by the domain user visits, and the third-party cookie which is set by a different domain than the visited one (European Commission, n.d.). Many premium publishers like The Economist or The Guardian notifies users about their cookie policies when people visit them the first time. 2.2.2 Flash Cookies Adobe’s Flash software can also be used to store information - if installed by the user. It has advantages like not being easy to be deleted and cross- browser tracking capabilities (Ullrich, 2015). Since Flash is expected to be “dead” very soon (Singleton, 2016), this type of cookie may not be used in the future. 2.2.3 IP Address An Internet Protocol (IP) address is an identifier every computer uses when connecting each other and share data (Guthrie Weissman, 2015). IP address can be used to track the online behaviour of users. However, it is not very
  • 22. 16 reliable, since users can change their IP addresses or multiple users can share the same IP address (Ullrich, 2015). 2.2.4 Browser Fingerprinting Browser fingerprinting is done through configuration and settings made in the browsers (Panopticlick, n.d.), which are likely to be unique. Hence, it can be used to track users (Ullrich, 2015). 2.2.5 Local Storage HTML5 is a web language promising to launch the next era of websites, smartphone apps and online videos (Goldman, 2012). It has introduced some mechanisms like session storage and local storage, which can help with storing data on the browser of the user; hence, helping with tracking their behaviours (Anon., 2016). 2.2.6 Canvas Fingerprinting Mowery and Shacham (2012) introduced canvas fingerprinting by utilising a feature of HTML5, which became popular in time (Wagstaff, 2014). It is a very persistent way of tracking users, and there is not an easy way of stopping it (Ullrich, 2015). 2.2.7 Carrier Injected Headers Some mobile carriers track users using some methods like altering the internet (HTTP) requests (Ullrich, 2015). Verizon uses this technique and calls it “Unique Identifier Header”, which they claim to be used for more personalised advertising (Verizon, n.d). It is discovered that Verizon has been doing this practice since 2012 but did not declare it until 2014, which resulted in a fine of $1,35 million (Ha, 2016a). Carriers like AT&T also use unique trackers, and they also claim that it is used to show relevant ads (AT&T, n.d.).
  • 23. 17 2.2.8 Evercookie By combining many tracking methods, Samy Kamkar developed a software which can track users persistently. Even if the user deletes cookies, the Evercookie script could use canvas fingerprinting or another technique to track him/her, and deleted cookies would be recovered (Kamkar, 2010). This method became very popular and even featured in The New York Times (Vega, 2010). On the other hand, there are some privacy and ethical issues surrounding this method (Cleere, 2015). 2.2.9 ISP-Level Tracking Some companies like Phorm track virtually all online activities through their integrations at the internet service provider (ISP) level. Obviously, this is potentially the best way of tracking users, as it is possible to analyse every action a user does online. However, it is also the most harmful way from user privacy point of view, and it led to some scandals in the past (Williams, 2011). 2.2.10 Others There are many other ways of tracking. Some companies even started to use more innovative ways like AudioContext fingerprinting (Englehardt & Narayanan, 2016). However, there is no need to go into details of all of them. Many companies track users on the web and collect data about them. Online tools like Ghostery and Disconnect help users understand which companies track them. 2.3 Data Quality As discussed before, data can be collected from many sources; from online behaviours to offline actions. If it is not done properly, there might be some issues like poor data quality. In an interview, Bob Rupczynski, VP of media, data and CRM at Kraft Foods, gave an example regarding one of their campaigns, where they wanted to target females but noticed that female data
  • 24. 18 was only half-accurate. In another campaign, they targeted some people because of outdated data (Morrisey, 2015). There are several factors which could cause poor data quality. 2.3.1 One Device & Multiple Users In the United Kingdom, the average number of people per household is 2.4 (Knipe, 2015), and on average there are 1.3 laptop and 1.2 tablets per household (Press Association, 2015). Therefore, it is fair to assume that more than one family member might use the same computer. When different people use the same computer, it is difficult to understand who is behind the computer and which ad would be interesting for him/her. Since smartphones are more personal devices, the case might be different for them, though mobile devices have their own complications, which will be explained later. 2.3.2 One User & Multiple Devices The average household in the United Kingdom owns 7.4 internet enabled devices (Press Association, 2015), and the average number of people in a household is 2.4 (Knipe, 2015). Hence, it would be fair to assume that one user can use multiple devices to connect to the internet. When one user uses multiple devices, it is difficult to connect his/her data across devices. However, advertisers want to be able to target the same person on multiple digital devices, which is called cross-device or cross- screen targeting (Tan, 2014). There are two main approaches to cross-device targeting: deterministic and probabilistic. The deterministic method requires a user to use a unique identifier, which would help to recognise him/her across devices (Schiff, 2015). Facebook, for instance, is one of the “lucky” companies that is widely used with login data, and this login information helps Facebook to recognise
  • 25. 19 the same person across devices. Hence, one of the Facebook products, Atlas, is claiming to help to the advertisers in delivering and measuring ads across “everything” (Atlas Solutions, n.d.). Probabilistic method, on the other hand, relies on some algorithms that are analysing many data points to understand if a person is the same across devices (Schiff, 2015). Tapad and Drawbridge are two companies that apply this method. Both methods are not working perfectly and there is still improvement needed in this area (eMarketer, n.d.a). 2.3.3 Deleted Cookies Some people delete their cookies from time to time, and as soon as cookies are deleted, the data that was kept in cookies and connection to the user is lost. A 2014 research shows that 28% of users in Europe delete their cookies four times a month (comScore, 2014). It affects data quality negatively and may result in poor campaign performance. 2.3.4 Data Loss Different technology players in the online advertising ecosystem synchronise cookies with each other (Bonder, 2014), so that they can all track the same user in a more efficient way. A recent study shows that most of the companies apply this method (Englehardt & Narayanan, 2016). However, the more players synchronise data with each other, the more problematic it becomes, as the number of network requests increases significantly (Mitchell, 2014). It may also cause discrepancies and data loss during this process, which can be up to 30% (Ad Ops Insider, 2011a). 2.3.5 Recency It is important to show users timely ads. It is an obvious situation, but as mentioned before, some people can be targeted due to outdated data.
  • 26. 20 Therefore, it is important to understand user’s “recent” behaviours and show ads accordingly, which is called “recency” in the industry. 2.3.6 Missing Data Points Unfortunately, it is not always possible to gather all necessary data or to fully merge different data points of a user. Some of the reasons are: ● Users visit various websites, and not all these websites have a centralised way of collecting and sharing data. For instance, Google does not share search data with Facebook. ● One user may use multiple devices - as discussed before. ● Users have “offline” behaviours as well. For instance, someone might buy a car from a car dealer after doing some online research and seeing some online ads, but this offline data point might not be synchronised with the online world. 2.3.7 Poor Integration / Implementation As will be explained later, using “pixels” is a popular way of tracking users, which are small codes placed in the web pages of the publishers or advertisers. If this technical implementation is not done properly, it might cause poor data quality. For instance, if pixel implementation is problematic, it will not be possible to understand whether the user made a purchase; hence, will be continued to show ads to the user, which could ruin the user experience (Rosenbaum, 2014). 2.3.8 Technological Challenges As mentioned before, cookies are the most popular way of tracking users. However, mobile devices do not fully support cookies, and this causes some technical difficulties (Internet Advertising Bureau, n.d.). Another problem is with cross-device targeting. It is already mentioned that existing methods are not functioning perfectly yet.
  • 27. 21 Finally, one computer can be used by multiple users, and existing technologies are not advanced enough to immediately detect who uses it. 2.4 Segmentation Once data is collected, publishers or advertisers create segments of people that have similar interests and behaviours. For instance, people who visited a car company’s (e.g. Audi) page and checked a particular car model (e.g. A4) several times in the last few weeks would fall into the same segment. In this example, Audi would be able to tailor its message to this specific A4- interested-segment. Hence, all people under this segment would see the same banner. Similarly, publishers can segment their users based on their behaviours on their website. For instance, people who read some car-related news in the last few weeks are potentially interested in cars, and they can be potential customers of car advertisers. It is important to note that the smaller a segment is, the less attractive it becomes for publishers due to economic reasons. A segment of 450 people would not help with increasing revenues, especially in a world where a publisher is paid per one thousand ad impressions (CPM). If the publisher is paid per click (CPC), even worse. 2.5 Usage of Data Data can be used in many areas of online advertising. The major ways will be explained below. 2.5.1 Traditional Advertising One of the potential use cases for data usage is with traditional ways of selling media. Publishers can create audience segments from visitors who visit their websites. They should be using an ad-server (a software managing advertising campaigns) which has support for audience segmentation and
  • 28. 22 targeting. DFP, Google’s ad-server for publishers, can create audience segments and target them (Google, n.d.c). Alternatively, the publisher can use a DMP, which is a data-focused technology that will be discussed later. In each case, when an advertiser asks the publisher to target a specific audience, the publisher can create these segments if they do not exist yet, and then run an advertising campaign targeting that specific audience. 2.5.2 Programmatic Advertising Programmatic advertising is a relatively new and growing way of showing ads, where machines calculate many variables and decide whether to show an ad, with a minimal human involvement in the process (Rayport, 2015). Data is vital for programmatic advertising (Olenski, 2015), as will be explained shortly. In programmatic advertising, the most important players in the value chain are demand-side platforms (DSP), supply side platforms (SSP) and ad exchanges. For simplicity purposes, the term SSP will be used for both SSPs and ad exchanges as they are doing a similar job nowadays. A DSP is a technology mainly used by advertisers or agencies to buy advertising spaces on publishers. Data is a crucial part of this buying process since DSPs usually try to target specific users (Marshall, 2014a). As there are billions of transactions happening in a day (Rayport, 2015), many DSPs use machine learning to optimise their campaigns. Similarly, SSPs help publishers to sell their inventory in an automated way (Marshall, 2014a). Again, data is critical here since advertisers require publishers to send them information about users, the more data, the better. These technologies communicate with each other mostly through IAB’s OpenRTB protocol (Internet Advertising Bureau, 2016). This protocol enables
  • 29. 23 the sellers and buyers transact many data about users like location, gender, content they read, segment they belong to and so on (Internet Advertising Bureau, n.d.) 2.5.2.1 Cookie Syncing Since a DSP gives decisions based on data, it is important to understand who the user is to make decisions like which ad to show, how much to pay and so on. On the other hand, users visit publishers and publishers usually work with SSPs, which are enabling DSPs to show ads to the users. However, DSPs and SSPs are different systems, they both have their ways to track users, hence, the same user would probably have different IDs in these systems. There are many different SSPs and DSPs, and a user is likely to have a different ID in each system. Through a process called “cookie syncing”, DSPs and SSPs share user IDs with each other so that they can recognise the same person. Cookie syncing does not happen only between DSPs and SSPs, but also between other programmatic (and non-programmatic) players. A recent study shows that it is a quite popular method used by many companies (Englehardt & Narayanan, 2016) 2.5.2.2 DSP & SSP A DSP uses data mainly through the methods below: ● Cookies: When a DSP shows an ad to a user, it can put a cookie on his/her browser or read data from it. Through cookie syncing with other platforms, DSP can track users. ● Web beacons: Web beacons are also known as tracking pixels (or simply “pixels”). They can be placed on advertiser’s pages to see which pages a user is visiting and how long s/he stays there, which helps to understand the intentions of the user. When a user purchases a product or fills a form, s/he is taken to a “Thank You” page, where another pixel (a.k.a. conversion pixel) is placed. It is also an important piece of data since the DSP can analyse the profile of the users who
  • 30. 24 “convert”, and optimise the campaign accordingly (Internet Advertising Bureau, 2014). ● DMPs: A DMP is used to collect data about audiences, segment them and pass these segments to DSPs to show relevant ads (Marshall, 2014a). DMPs and DSPs also synchronise cookies so that they can identify the same person (Mediagician, 2013). ● Third Parties: A DSP can be integrated to many other third party companies to measure how much of the ad is seen, whether the page is “safe” for the brand (i.e. there is no content damaging the brand), whether the user activated the mobile app s/he downloaded and so on. A DSP can also collect some of this information and use it for optimisation or reporting purposes. For instance, if the ads are not seen by anyone on a particular website, DSP can stop targeting that site. Similarly, an SSP can utilise data through: ● Cookies: Using cookies, an SSP can track users, collect user information and operate the technology (Rubicon Project, 2016). One of the primary purposes for cookie usage in SSPs is to match their IDs with other third parties like DSPs. ● Web beacons (Pixels): As explained before, pixels can be used to track browsing behaviour. ● Web Logs: Web-logs also can be used to collect information regarding how a user interacts with the ads s/he sees (Pubmatic, 2015). ● DMPs: Publishers can use DMPs to segment their users and offer these segments to the advertisers to show relevant ads. 2.5.3 Targeted Media Buying Data can be used to target audiences. Instead of showing ads to everyone, only people who would potentially be interested in the product or service can be targeted, which would result in efficient budget usage.
  • 31. 25 Targeted media buying can be done in various ways. The major models are described below. 2.5.3.1 Behavioural As discussed before, depending on the information like what a user reads, searches or buys, his/her intentions can be analysed and used for advertising purposes. An ad-server or DSP can utilise this behavioural data, and as soon as the related user is detected, an ad can be shown to him/her. 2.5.3.2 Predictive The concept of "big data" is mostly about human behaviours, and these past behaviours can help with predicting the person's future actions (Pentland, 2014). When this approach is applied to online advertising, it can be predicted what the user will buy, when is the right time to show the ad, what is the chance of conversion and so on. Companies like Adtheorent and Trusignal use data to make predictions and to optimise the advertising campaigns. 2.5.3.3 Lookalike (Audience Extension) By analysing behavioural data of some users, it is possible to find similar audiences who acted in similar ways and who would likely be interested in the same product or service (Facebook, n.d.c), which is a form of predictive targeting (Sukornyk, 2013). Another name for look-alike targeting is “audience extension”. DSPs like Adgear and Fiksu have this capability. 2.5.3.4 Retargeting In simple terms, retargeting is tracking a person who visited an advertiser’s website and showing him/her an ad on another site (AdRoll, n.d.). For instance, when a person checks a Samsung Galaxy phone on Samsung’s website, this product’s ads start to appear to him/her on many other
  • 32. 26 websites. Advertisers use retargeting to reconnect with their prospects or customers and pull them back to their websites. There are several types of retargeting; some major ones are explained below: ● An advertiser can retarget people who visited the advertiser’s website, which is called site-retargeting. ● Search targeting is the natural capability of search engines like Google. Using Google’s advertising technologies like AdWords, it is possible to target people who searched for specific keywords. People do searches on other websites as well, and their search keywords can be used to retarget them. Moreover, when people do a search on a search engine and land to a website, it may be possible to retrieve the search keyword made on the search engine, and retarget people accordingly. ● Social retargeting is retargeting people on social media channels. Facebook allows advertisers to retarget people through Facebook Exchange (FBX) (Smith, 2014), though this service may be shut down soon (Shields, 2016). Twitter also has similar capabilities (Shrivastava, 2014). ● Sequential retargeting is showing ads to a person in a sequence (Gaudio, 2014). For instance, after seeing the first ad of a car, the person can be retargeted with another ad giving more details about the car or showing the interior design of it. ● CRM retargeting, which is a type of retargeting, will be explained separately. Based on the banner type, there are two types of retargeting: static and dynamic. A static retargeting is simply retargeting people with a non- personalised banner, whereas dynamic retargeting is personalising the banner and show the audience the most relevant products or services (Perfect Audience, n.d.). For instance, Samsung could retarget its website’s visitors with the same banner, which is an example of static retargeting.
  • 33. 27 Alternatively, it could show to each of them the product they have viewed, so one visitor could see a TV ad whereas the other could see a smartphone ad, which is an example of dynamic retargeting. In any case, utilisation of data is crucial. Especially for dynamic retargeting, a deeper integration between retargeting technology and the advertiser is necessary. Retargeting has proven to be one of the most effective targeting methods (Koetsier, 2014). Companies like Criteo, Adroll and Retargeter are focusing on this area. 2.5.3.5 CRM Customer relationship management (CRM) allows businesses to manage their relationships with their clients and prospects (Burnham, 2013). Customer and prospect data can be integrated with advertising technologies and make customers and prospects available for (re)targeting. As an example, a bank can target customers who spend more than $5,000 in the last two months and offer them a premium credit card. Since CRM data is about a company’s existing clients or prospects, it is very precious. Existing customer and prospect data can be activated in many different combinations (e.g. a bank can target people who spent more than $5,000 monthly and have a loan less than $100,000). It is also possible to add offline data into the mix, which is obtained from stores or call centres. For instance, a retailer can target customers who spent more than $500 in its stores in the last few weeks, and offer them a special discount. Companies like Adroll, Retargeter and Liveramp can use CRM data to target audiences. DMP companies like Bluekai, Lotame or CXense are also able to utilise CRM data (online and offline). 2.5.3.6 Cross Device To be able to target the same person across devices, s/he needs to be identified on different devices. It has been discussed before that this can be achieved through two methods: deterministic, which uses unique identifiers
  • 34. 28 to understand who the user is, or probabilistic, which uses an algorithmic approach to “guess” who the user is. To be able to use Facebook, a user should login first - on any device. It helps Facebook to identify the same user on different devices. Google tries to use login data as much as possible, asking its users to login on Chrome, on Android and so on. Adobe is also working on a technology which uses a “pool” of login data from different parties, so that if any of the parties recognises the user, it can share this information with others (Ha, 2016b). These are all examples of deterministic approach. Companies like Bluecava, Drawbridge, Tapad and Crosswise uses probabilistic models to identify the users. Cross-device targeting is one of the important challenges for the advertisers, and the existing technologies are not very useful yet (eMarketer, n.d.a). 2.5.4 Campaign Measurement & Optimisation Data can be used to measure and optimise advertising campaigns. 2.5.4.1 Campaign Measurement One of the obvious applications of data is campaign measurement. How many times an ad is shown and clicked can be measured easily, but it is also possible to measure how much of the banner stayed in the viewing area of users, whether a user made a purchase after seeing the ad and so on. 2.5.4.2 Banner Optimisation Usually, advertisers create a few different versions of their banners and show them to users in a rotation. Some technologies can optimise the advertiser's campaign after a while by checking which banner is clicked more by the users, then show only that banner to the users and drop the rest.
  • 35. 29 2.5.4.3 Dynamic Creative Optimisation (DCO) Dynamic creative optimisation (DCO) is creating banners automatically in real time, based on interest or action data of the user (myThings, n.d.). This would help the same banner to be tailored for each user considering his/her data and show him/her the latest product s/he viewed for instance, which can also be called "personalised banner". DCO results in much better campaign results than static banners (Sizmek, 2014; Ad Ops Insider, 2011). 2.5.4.4 Bidding Optimisation Real-time bidding (RTB) is the most popular way of trading media in the programmatic advertising ecosystem. It is buying or selling ad impressions in an auction which happens in milliseconds. When a user loads the page, publisher’s SSP auctions him/her off to the advertiser’s DSP that pays the highest price (Marshall, 2014a). DSP decides the bid value by analysing data. If a user has viewed a book and a house recently, for instance, a real estate company would be willing to pay a very high price to show an ad to that user. On the other hand, the bookstore can calculate how much they would make from selling a book to this person, and calculate the value of showing an ad to him/her accordingly. 2.5.5 Attribution Before making a purchase, a customer goes through some decision-making stages. Some call these stages “purchase funnel”, some “consumer decision journey” (Court et al., 2009). People are exposed to many marketing touch- points during the purchasing process, and marketers should determine what affected the consumers before purchasing their product or service, which is also called “attribution” or “attribution modelling” (Convertro, n.d.). Many marketers do not have any attribution modelling in place (eMarketer, 2015c), and there is also offline actions of the users which should be considered, but still, most marketers who use attribution modelling see great benefits in it (Kehrer, 2013).
  • 36. 30 Companies like Convertro and VisualIQ work on attribution modelling. 2.5.6 Yield Optimisation Many publishers start to understand how valuable their audience data is (Stenberg, 2013) and although some of the publishers are still sensitive about their data (Reagan, 2013), they sell more and more audiences to their clients (AdExchanger, 2010). By understanding which audience segments are more demanded by the advertisers, publishers can adjust prices and maximise revenues. 2.5.7 Audience Optimisation By targeting different audience segments in advertising campaigns, advertisers can understand who is more interested in their products or services, the cost of these segments, their contribution to the revenues, and make optimisations accordingly. For instance, if an advertiser tries to sell cars to women, and if young women click the ads and book test drives more than the others, the advertiser may start targeting young women more aggressively and spend the budget more efficiently. 2.5.8 Data Sales There are three different types of data from ownership perspective: first, second and third party data. First-party data is the data owned directly by publisher or advertiser, and created from sources they own, as demonstrated in Figure 2.1. CRM data, or behavioural data, which is obtained by the actions made on the publisher’s or advertiser’s website is first-party data for instance. Publisher or advertiser can be sure about the quality of the first-party data since it is generated from their own sources.
  • 37. 31 Second-party data is the data obtained from a partner company. Two (or more) different companies can make an agreement and share their first-party data with each other, as demonstrated in Figure 2.1, in which case each company’s data would be second-party to the other. Third-party data is the data owned by other companies, aggregated from different sources and usually sold on marketplaces to any buyer (Lotame, 2013), as presented in Figure 2.1. In this case, quality of the data may be questioned since the source of the data would not be fully visible to the buyer. Figure 2.1 Types of Data (Muir, 2016) All three types of data can be beneficial.
  • 38. 32 ● First-party data can be very precious, especially to those who own it. An airline company can offer a discount to its customers who have flown to Japan in the last three months for instance. It is also possible to sell the first-party data. For example, many advertisers would want to know if a specific person is reading Financial Times. Thus, Financial Times could potentially sell this data to the advertisers. ● Second-party data can be very helpful in case existing first-party data is not enough or can be enriched by additional data from partners. For instance, Rolex and Mercedes could partner with each other and share their website visitor cookies since a Rolex owner is likely to be interested in having a prestigious car and vice versa. ● Third-party data sales will be explained along with data marketplaces (data exchanges) below. 2.5.8.1 Data Marketplace (Data Exchange) A data marketplace helps companies to buy and sell audience data (Techopedia, n.d.). Third-party data is traded in these marketplaces. For instance, owner of a blog on prestigious cars can put its visitor data to these marketplaces and any buyer who is interested in this segment could buy and show ads to those people. MasterCard is one of the companies that sells data on data marketplaces in a similar fashion (BlueKai, 2012). Data in the marketplaces can be branded or anonymous (Exelate, n.d.). If the buyer can see the owner of the data, it is a branded data. If the seller name is not declared, the data is anonymous. One of the question marks about data marketplaces is who contributes to the anonymous data segments, how much they contribute and what is the quality of them. Some of the data management platforms (DMP) like Bluekai, Exelate and Lotame also have a data marketplace solution. DMPs will be explained later in detail, but typically a DMP’s focus is helping publishers or advertisers to analyse and segment their data and make this data actionable, whereas a data marketplace helps them to trade it. Having an anonymous data selling
  • 39. 33 option in a marketplace may raise questions by the DMP owners regarding how much of their data is contributing to the quality or quantity of data sold in the marketplaces. 2.5.9 Frequency Optimisation Frequency capping is limiting the number of times a person sees the same ad (Will Marlow Agency, n.d.). It is important to use frequency capping in advertising campaigns and deciding optimal frequency correctly since a user might receive campaign message when s/he is exposed to the creative five times, and targeting him/her more would be a waste of money. Moreover, being exposed to the same banner over and over may disturb user, so frequency capping can protect the brand from being “annoying” (Kessler, 2011). Research shows that frequency capping can decrease user acquisition costs up to 30% (Chandler-Pepelnjak & Song, 2009). Nowadays many people use more than one device, so setting a frequency capping across devices is becoming more important (BlueCava, n.d.). Data can help to understand and limit how many times a user has seen the same ad, not only on the same device but also across devices. Usually, frequency capping is set at campaign level, so all users are targeted at the same level of frequency. Finding optimal frequency capping is a challenging task (Buchbinder et al., 2014) and it may be difficult to decide this number at the user level. On the other hand, advanced technologies might make it possible in the future. 2.6 Fraud Detection Ad fraud is one of the biggest problems in the online advertising industry. It is mostly caused by non-human traffic, also known as “bots” (Heine, 2016), and it is expected to cost $7.2 billion to the industry in 2016 (Stefo, 2016).
  • 40. 34 Software, acting like a human, visits the publishers’ website and “see” the ads that are paid by the advertisers, and sometimes the software “clicks” the ads, visits the advertiser’s site and spend some time there. Thus, they consume advertisers’ budgets and harm the industry. Companies like Forensiq and Integral Ads offer solutions to analyse data and detect fraudulent traffic generated by bots. 2.7 Viewability Viewability is another important topic in the online advertising industry. Many publishers have more than one banner space on their pages, and not all of them enter the viewing area of their visitors; hence, the advertisers question if their budgets are spent wisely. A banner ad displayed on a desktop device is considered to be “viewable” if 50% of it is shown at least 1 second to the user (Internet Advertising Bureau, 2014b). On the other hand, companies like Facebook have different approaches to the definition of viewability (Marshall, 2015b). So, there is a debate about what a “viewable ad” means (Marshall, 2015c) and different vendors provide different results under same test conditions (Internet Advertising Bureau, 2014b), but still, it is an important subject for the industry. Also, the “person” who views the ad should be real, not fraud, so this topic is intersecting with fraud detection at some points. Companies like Meetrics and Moat help advertisers solve this problem. 2.8 Mobile-Specific Challenges & Benefits Smartphones become more and more popular, helping mainly to communicate and to access information among other things. Hence, mobile internet penetration is increasing rapidly - it is forecasted to reach 71% by
  • 41. 35 2019 (Internet Society, 2015), and in 2015, Google announced that global mobile searches surpassed PC searches first time (Macmillan, 2015). Although it may not mean that internet access through PCs will die (Marshall, 2015a), it shows how dominant mobile devices started to be in peoples’ lives. High usage of mobile devices bring some challenges to publishers and advertisers, since tracking a user on mobile devices is not easy (Internet Advertising Bureau, n.d.). 2.8.1 Publishers An ordinary premium publisher would have a regular website, a mobile website and mobile apps for both iOS and Android supporting phones. As a result, running a cross-platform campaign might be challenging. Doubleclick for Publishers (DFP), the widely used ad-server of Google, provides some mechanisms to track users on mobile devices, one of them being publisher provided identifiers - PPID (Google, n.d.e). But not all ad-servers have similar functionalities, and not all publishers can provide such unique identifiers. Another challenge is about frequency capping, which is limiting the number of times a person sees the same ad in a given period. Since it is hard to track the same person across devices, it is also difficult to prevent the same person seeing the same ad more than desired, which can potentially impact campaign results and reputation of the publisher negatively. 2.8.2 Advertisers Tracking a user on a single mobile device is hard, but it is not the only problem for advertisers. Many people do not use a single device to connect to the internet. 90% of people shift between devices to complete a task (Taube, 2013). When British people are at home, they switch devices 21 times an hour (Ridley, 2014). It creates further complications for the advertisers.
  • 42. 36 On the other hand, advertisers would want to collect data about who sees their ads and how many times, how people react to these ads, on which channels they react positively or negatively and so on. Existing technologies do not fully help with this need due to difficulties with tracking the same person across devices. 2.8.3 Cookies Using cookies is still a popular method to track users on mobile devices. On the other hand, there are different mobile devices, web browsers and applications. First party cookies are allowed by almost all web browsers. However, third party cookies are not allowed by some of them. Moreover, there is only a limited allowance on mobile apps (Internet Advertising Bureau, 2013). 2.8.4 Other Tracking Methods There are also other methods of tracking mobile behaviours. Major ones are briefly described below. ● Client / Device Generated ID: These are unique identifiers provided by mobile operating systems. Both iOS and Android offers identifiers to marketers (Google, n.d.f). These IDs mainly help with tracking users across mobile applications within the same device (Tune, 2016). ● Statistical ID: By analysing various data points and applying some probabilistic algorithms, the uniqueness of a user can be inferred (Internet Advertising Bureau, 2013), which can help with identifying the same user across apps, websites and devices. Companies like Drawbridge and Tapad use this approach. ● HTML5 Cookie Tracking: HTML5’s local storage can be used to store information and can help with tracking. It can be used only when the browser is open (Internet Advertising Bureau, 2013). ● Universal Login: If a user logs into the mobile application or website, this login data can be used to track the user. Some companies like Facebook require their users to be always logged in, which is very helpful for tracking purposes (Internet Advertising Bureau, 2013). This
  • 43. 37 identifier can be used to track the same user across browsers, apps and devices. Sometimes tracking users is done at mobile operator level; Verizon and AT&T use special techniques to track users within their network (Verizon Wireless, 2016; Ha, 2016a). Cross-device targeting, which is tracking the same person across devices (Bilton, 2015), is a complicated problem and mainly done through two models: deterministic and probabilistic. The deterministic model relies on data of logged-in users - as in the universal login method explained earlier, and probabilistic model relies on analysing different data points to identify the same user - as in the statistical ID method discussed before (Heine, 2016). Although it is in high-demand, existing technologies are still underdeveloped (eMarketer, n.d.a). 2.8.5 Benefits of Mobile Devices One of the biggest benefits of advertising on mobile devices is being able to target people in a particular geographic location since people usually carry their mobile devices with themselves (Internet Advertising Bureau, 2013). The location information can be obtained through various methods; hence, the advertisers can target people in a particular shopping mall for instance.
  • 44. 38 3 BENEFITS AND RISKS FOR THE PLAYERS IN THE ECOSYSTEM So far it has been discussed how data is collected and used in various contexts. This section focuses on potential benefits and risks of data for major players in the online advertising industry; namely publishers, advertisers, advertising agencies, technology providers and of course users themselves. The section will start with the discussion of benefits first, and then will continue with risks. 3.1 Benefits for Publishers Publishers, especially large ones like newspapers, are visited by many people every day. For instance, Nytimes.com is visited by 65.8 million people (The New York Times Media Kit, 2016) and Forbes.com is visited by 46 million people monthly (Forbes Media, n.d.). It means a company’s existing and potential customers would likely to be among those millions of people. Also, the publishers are one of the main channels to show ads to people in the online advertising world. The only question is; how can an advertiser show its ads to the relevant people, and not waste their money by showing ads to others? In the past, advertisers were using a model called “spray and pray”, showing their ads to as many people as possible and hoping some of them would become their customers (Pomerantz, 2014). However, those days are coming to an end since the consumer tolerance for seeing irrelevant ads are very low nowadays (Roman, 2015), and obviously, it is not an efficient way of spending advertising budgets. Therefore, publishers offer different advertising models to the advertisers, so that they can reach to the audiences they target. One of the major ways publishers offer is categorical targeting, so a technology company could
  • 45. 39 show its ads on the -related news, whereas a car company could show its ads on the car-related news for instance. The other major and trending way is audience targeting, where the publisher would show technology-related ads to the technology related people, even if they read other types of news at that moment. The major ways data helps to the publishers are explained below. 3.1.1 Data + Media Bundle Sales Publishers can bundle their data with their media and offer it to the advertisers. For example, Forbes.com could offer Volkswagen an advertising campaign on its main page, where ads are shown only to people who are interested in cars. This way of showing ads would be beneficial for publishers in various ways: ● It would help organising the inventory and use different pricing models for different products. For instance, showing ads to the car-lovers on the main page could be priced differently than showing ads to the book-lovers. ● Showing ads to relevant people would help to achieve better campaign results, which could increase publisher’s reputation and repeat sales. ● Showing relevant ads to the visitors would be less disturbing for them, which could again increase publisher’s reputation and traffic. 3.1.2 Data Sales Publishers can offer their data (audience segments) to buyers, without bundling them to their media, which would allow advertisers target these audiences on any publisher they want. For instance, Forbes.com could sell cookies of car-lovers to Volkswagen, and Volkswagen could target these people on TheGuardian.com. Publishers can do this through different ways like:
  • 46. 40 ● They could allow a buyer to collect data directly from their websites by putting advertiser’s pixels into their website. ● They could allow an advertiser to put a cookie into user’s browser each time an ad is shown. ● They could use a DMP (a technology which will be explained later) and pass segment cookies to the advertiser’s ad-server or DSP. 3.1.3 Data Partnerships One of the ways publishers can benefit from data is through partnerships with other companies. For instance, Telegraph.co.uk has a section dedicated to cars (The Telegraph, 2016), but they can partner with Autoblog.com to access to more people who are interested in cars. Thus, even if a visitor of Telegraph.co.uk does not read car related news on its site, s/he may read on Autoblog.com, and The Telegraph would know this through its partnership with Autoblog.com. As explained before, this is called second-party data, and it can be very beneficial. Data partnerships can help publishers in the ways below: ● It can help contribute to understanding visitors better, providing a complete view by enhancing existing or integrating non-existing data points. ● When selling data, directly or bundled with media, additional data points would contribute to data quality, and in some cases, it would increase revenues further thanks to the new audiences discovered in partner’s website and higher unit prices. ● Data partnerships with prominent companies may contribute to the brand image of a publisher. 3.1.4 Audience Enrichment A user usually visits more than one publisher in a day; so a publisher can have only a partial understanding of its users. Buying data from other websites through data marketplaces (a.k.a. third party data) and enriching publisher’s own data points about the user (a.k.a. first party data) is called
  • 47. 41 “audience enrichment” (Lotame, 2016). Actually, it is possible to combine first party, second party and third party data and enrich existing data points. Some publishers offer “guaranteed” audiences, combining advertiser’s first party data with other data points, aiming to make sure that only people that advertiser targets see the ads (Friedman, 2016). The benefits of audience enrichment are similar to the data partnerships explained before. 3.2 Benefits for advertisers Data is crucial to advertisers. It can help mainly in the ways below. 3.2.1 Targeting a Specific Audience A car company would usually want to show its ads to the potential car buyers. To be able to “find” this audience, they can do contextual targeting - showing ads in car related websites or car related news. However, this would help only partially, because a potential car buyer would not spend his/her whole time on a car-related website or page, s/he would read other news or visit other websites as well. Missing him/her on those pages, or worse, allowing a competitor to target him/her on other sites would reduce the chances of selling him/her a car. So, the advertisers would want to show their ads to potential clients wherever they are. Data can help to achieve this goal. ● An advertiser can use first-party data to (re)target potential customers. Thus, whenever a potential client is “spotted” on a publisher, advertiser would show an ad to him/her. ● An advertiser can partner with another brand to share data with each other. It would help them to use second party data to target potential customers. For example, Rolex and Mercedes could share data with each other, and whenever a Mercedes client is spotted on a publisher, Rolex can show an ad to him/her. On the other hand, Mercedes may
  • 48. 42 also want to show an ad to this particular user, which should be taken into consideration by both parties. ● An advertiser can use third party data to target potential customers. For example, MasterCard sells audiences based on their spending behaviour; hence, a company like Rolex could buy data from MasterCard and target people who have high monthly spending. 3.2.2 Banner Optimisation Showing dynamic banners with relevant products to the users can result in huge benefits (DoubleClick Advertiser Blog, n.d.) - as discussed before in dynamic creative optimisation (DCO) section. By using data, it is possible to show personalised banners that may attract customers, either by retargeting or predicting what could be interesting for them. 3.2.3 Understanding Customers Better By analysing different data points about users, advertisers can understand what type of banners or channels drive their customers to sales (or another action they want), what is the optimal price to show them ads, what product or service their customers are interested in and so on. This understanding would result in optimisations in campaign budgets and channels, also would potentially increase revenues by selling more products or services. 3.2.4 Viewability Data allows advertisers to understand how many and how much of their ads are actually “viewed” by users, which is called “viewability” (Marshall, 2014b), and make optimisations accordingly.
  • 49. 43 As discussed before, the definition of “viewability” is not entirely standardised yet, but it will definitely evolve in time, and eventually help advertisers spend their budgets on viewable ads only. 3.2.5 Fraud It is estimated that ad fraud will cost advertisers $7.2 billion in 2016 (Bond, 2016), which is a big concern for the advertisers. One of the important players in programmatic advertising space, AppNexus, lost around 90% of its traffic after removing its fraudulent traffic (Wiessbrot, 2016). Data helps advertisers to detect fraudulent activities and stop them as much as possible. 3.2.6 Spending Budgets Efficiently Showing ads only to target audience helps with spending the campaign budgets more efficiently as well. ● Using data, instead of showing ads to everyone and pay for it, advertisers can show ads only to existing & potential customers. ● After a user sees an ad a predetermined number of times, s/he can be excluded from the targeting group (a.k.a. frequency capping). ● By understanding who responds positively to ads, shifting budgets towards that audience would help with budget optimisation. ● Programmatic advertising technologies use sophisticated algorithms to evaluate many data points and decide when and where to show an ad to an audience, and optimal price to show the ads. 3.2.7 Brand Safety Data can help advertisers to show their ads only on websites or pages where it is “safe” for them. For instance, an airline company would not want to show its ads on a page where there is news about a plane crash.
  • 50. 44 3.2.8 Data Sales Some brands that look for new revenue streams can decide to sell their data. Companies like MasterCard and American Express are two examples that use this business model (Kaye, 2013). 3.3 Benefits for Advertising Agencies Usually, advertisers work with an advertising agency to run their advertising campaigns. Therefore, an agency would have access to data of its clients. As explained before, this data can be very useful for the advertiser; hence, the agency should also be able to utilise it. The programmatic specialist agency Xaxis, owned by the largest advertising and marketing agency WPP, has invested $25 million in developing a data management platform (DMP) to help them gain leverage and make a more efficient buying, and it is considered part of their technology stack (Kantrowitz, 2014). Annalect, owned by the second largest advertising and marketing agency Omnicom, provides solutions and helps marketers to make data actionable (Omnicom Media Group, 2016). Another programmatic specialist agency, Vivaki, owned by the third largest advertising agency Publicis, also owns a DMP to manage data (Vivaki, 2015). So, all of the large agencies have created their own solutions to utilise data. Chief Strategy Officer of Accuen, Murphy (2016), thinks data would help with optimising media investments, tailoring relevant messages to audiences and provide accountability and knowledge to marketers.
  • 51. 45 3.3.1 Optimising Media Investments Some advertising agencies are specialised in media buying. Depending on campaign goals of their clients, they plan where to show the ads, and how much to spend on each media channel. Data would help greatly with this planning and buying process since the agency would know which channels are more “viewable”, less fraudulent, how much they contributed to sales and so on. If the agency benefits from programmatic advertising methods, data can help further by helping with real-time optimisation. 3.3.2 Tailoring Relevant Messages As explained before, data helps with targeting a specific audience and personalise the ads considering context or audience data. 3.3.3 Provide Accountability and Knowledge to Marketers An agency can use data to justify its media spending. It would help to agency’s client to understand the rationale behind agency’s actions. Moreover, data can help with analysing the types of banners that lead to sales, the profiles of people who respond positively to the campaign, what could improve the results and so on. This information is very important for the advertisers, and agencies can use data to gain and pass this knowledge to their clients. 3.4 Benefits for Technology Providers Data helps advertising technology providers in many ways. 3.4.1 Data at the Core By their nature, almost all technologies depend on data to be able to function properly. An ad-server system like DFP (Doubleclick for Publishers) uses
  • 52. 46 data to decide which campaign to prioritise or which banner to show and when. A DSP like DataXu uses data to determine the optimal bid price to show an ad to a specific audience. An SSP like Pubmatic uses data to decide the winner of an auction. A fraud detection system like Forensiq uses data to determine whether an activity is fraudulent. A DMP like Cxense directly promises to manage “data”. It is possible to give more examples, but it should be obvious that data is at the core of almost all technologies in the online advertising ecosystem. 3.4.2 New Business Opportunities Different types of data can be used to solve different problems. In the past, there was just a simple interaction between media seller and media buyer, and an ad-server was enough to run an advertising campaign. Then some start-ups wanted to automate this process with real-time auctions. Some companies wanted to understand if the ads were “really” viewed. Some companies wanted to expose fraudulent activities. Some companies wanted to know if it is the same person on different devices. It is possible to give more examples, but as it can be seen, data helped many start-ups to be borne, which are specialised in analysing different types of data and helping advertisers or publishers to do their jobs better. Some of these start-ups like Criteo now worth billions of dollars. 3.5 Benefits for Users Users have different attitudes against the fact that their data is used to show them relevant ads. Some people desire to see relevant ads (Sasaki, n.d.) and some of the research support this approach (eMarketer, 2016a), whereas some people find seeing targeted ads “creepy” (eMarketer, 2015b). Still, there are some benefits in seeing relevant ads.
  • 53. 47 3.5.1 Free Content One of the most important advantages of seeing ads is that they translate to free content. Not many people would pay to avoid seeing ads (eMarketer, 2015d). Many large publishers block their content to those who use ad-blockers. Figure 3.1 demonstrates how Forbes.com warns a user who uses an ad- blocker. Figure 3.1 Forbes.com main page with ad-blocker Not using or disabling an ad-blocker and visiting a publisher means giving consent to the publisher to collect data and showing relevant ads whenever possible. Figure 3.2 demonstrates how The Telegraph notifies its visitors about the usage of cookies.
  • 54. 48 Figure 3.2 The Telegraph main page with cookie usage notification As a result, users can either choose not to access to free content or access it knowing the fact that their data is used to show them relevant ads. 3.5.2 Inevitability Related to the topic above, if seeing an ad is inevitable, why not seeing a relevant one? Consumer tolerance for irrelevant ads is very low nowadays (Roman, 2015). Users may have realised that there is no escape of seeing ads; hence, prefer seeing a relevant one. On the other hand, with the rise of ad-blockers, seeing an ad is not entirely inevitable anymore. Ad-blockers are used by around 200 million people (Richards, 2015). To fight the ad-blockers, publishers use technologies that detect if a user is using an ad-blocker and “forces” him/her to see an ad by bypassing the ad-blocker, or offering other options like paid subscriptions. Some of the publishers do not allow their content to be viewed by a user if s/he is using an ad-blocker.
  • 55. 49 3.5.3 Discounts & Special Offers Users may benefit from discounts and special offers that are tailored considering their data. 52% of Canadian internet users would like to see such ads (eMarketer, 2015d) 3.5.4 Additional Benefits Some companies offer various benefits to users when they share their data for online advertising purposes. Datacoup, for instance, offers users money in return of their data and sells this data to the advertisers (Van Allen, 2014) along with other potential buyers (Datacoup, n.d.). Similarly, People.io offers users credits in return of their data, which can be used in the exchange of products, services, offers and so on (People, n.d.). 3.6 Risks for Publishers Data has its disadvantages for publishers as well. Some of them are explained below. 3.6.1 Privacy Concerns Tracking users and collecting data about their behaviours triggered some privacy concerns among users - many people do not like the fact that a “big brother” is watching them constantly. 31% of British people who use ad-blockers use them due to privacy concerns (Internet Advertising Bureau UK, 2015). Similarly, 50% of people in the US who use ad-blockers feel that their personal data is misused (Richards, 2015). It shows how serious the problem is, especially for the publishers since they are the ones who cannot sell their traffic. 3.6.2 From Publisher to Audience Especially with the rise of programmatic advertising, advertisers started to be able to target specific people wherever they are (Battelle, 2014). It means
  • 56. 50 that the audience might be on a premium website or might be reading a small blog; in any case, the advertiser can target him/her. This shift in the mindsets of the advertisers, from publisher to audience, might affect premium publishers negatively. The New York Times mentions this in its 2015 annual report; regards this problem as “audience fragmentation” and explains how it affects their business negatively (The New York Times Company, 2015). 3.6.3 Too Targeted = Too Low Spending? If advertisers target only specific audiences, publishers would not be able to sell enough inventory. If there are only 1000 potential real estate buyers out of millions of visitors, showing ads only to these people would result in very little revenue. 3.6.4 Information Asymmetry In some cases, a user might be too valuable to an advertiser and publisher may not be aware of this. For instance, a car company may know that a user is a potential car-buyer thanks to its first-party data, while the publisher may not be aware of this. Therefore, even if the car company would be willing to pay a high price to show an ad to that user, lack of data and knowledge would not allow the publisher to increase the price for that user, which would result in leaving money on the table. 3.6.5 Data Leakage Some publishers want to sell their data only bundled with their own media. Data leakage happens when a third party company collects data about publisher’s users and use this data elsewhere (Moses, 2015). Some of the companies that work with publishers directly might cause this issue, “inviting” other companies to drop pixels on users who visit the publisher’s website. Sometimes even companies that work directly with the
  • 57. 51 publisher might not be aware that another company is using “them” to reach to the users of the publisher. Although data leakage is a big concern of publishers, many of them use Google Analytics, Facebook buttons, or other technologies like AddThis. These companies explicitly announce that they use their tools to collect data and use it either for internal purposes or sell it to other companies. 3.6.6 Impact on Revenues Many different companies in the online advertising industry use and analyse data in a variety of ways; some use it for brand safety purposes, some use it for viewability analysis, some use it for fraud analysis and so on. All these different players get a share from each advertising dollar spent. Research shows that publishers receive around 65% of the advertising spending in EU - including agency commissions (IHS Inc., 2015), while some think they receive only 30% (Polisetti, 2014). So, it seems like the more intermediaries analyse and interpret different parts of data, the more revenues publishers lose. 3.7 Risks for Advertisers Data may create some disadvantages for advertisers in the ways explained below. 3.7.1 Being Annoying Particularly with the retargeting method, there is a risk of being annoying to users and damage the brand’s reputation. It happens when the user feels like s/he is being “stalked”, particularly when s/he has already purchased the product, but still seeing the same ad again and again (Rosenbaum, 2014).
  • 58. 52 3.7.2 Data Leakage As with publishers, there is a data leakage risk for advertisers as well. It happens when the advertiser works with a company to retarget people for instance, and willingly or unwillingly, this company calls “other” companies to track the customers of the advertiser, or use data of one advertiser for the campaign of another. Similarly, advertising agencies usually work with multiple advertisers, and they might use one advertisers’ data in another one’s campaign. 3.7.3 Increasing Complexity One of the advantages online advertising brought to advertisers was being able to measure how many times an ad was shown and how many times it was clicked. It was a big improvement in comparison to conventional advertising channels like TV or newspaper, where there was no way of measuring how many times an ad was shown. Being able to measure impressions and clicks required further data like who saw the ads, how much of a banner stayed in the viewing area of a user and for how long, what banner size contributed to sales more, was there any external factors contributing to the decision process of the user and so on. With the popularity of smartphones, things started to get more complicated since one user was using multiple devices, so the need of unifying data points from different devices increased a lot. It seems like the more data advertisers get, the more data they need to analyse further and make sense of the data at hand. This is not necessarily a bad thing, and obviously, it is better than having no data at all, but in comparison to the old days where there were some simple media plans and simple reports to make sense of, the current way of online advertising can be a bit “confusing” for some advertisers.
  • 59. 53 Another challenge is about the players in the industry. Some of them claim they do fraud detection very well, some of them focus on viewability, some of them are pure video DSP, some of them are pure mobile DSP, some of them are pure DMP, some of them are both DSP and DMP and so on. It is not easy for an advertiser to keep up with new companies that are potentially adding another type of value and test all the claims in an easy way. Thus, there is an increasing complexity in the market, and not many advertisers can stay up-to-date with all the changes in the industry. 3.7.4 Impacts on HR While data and technologies around it evolve, the needed skills and qualifications of people working in the industry are also evolving. WPP’s Sir Martin Sorrell referred to this change as “from mad men to maths men”, and many companies in the industry started to hire people with skills in math, engineering and analytics (Cookson, 2015). It is predicted that by 2017, Chief Marketing Officer (CMO) of a company will spend more time on technology than Chief Information Officer (CIO), therefore, new roles like Chief Marketing Technologist (CMT) started to appear and companies like Kimberly-Clark have roles like “Global Head of Marketing Technology” (Brinker & McLellan, 2014). As a result, the workforce in the industry is in change and this change may affect companies concerning training costs, employee costs, hiring procedures and so on. Programmatic advertising brings further complications. Its market size will grow from 60% in 2015 to 80-90% in 2019 in the UK (Internet Advertising Bureau UK, 2016), and it almost entirely depends on data. The difficulty of hiring and training people with the right skill set is seen to be the biggest obstacle in the adaptation of programmatic advertising models (IHS Inc., 2015).
  • 60. 54 3.7.5 Impacts on Advertising Costs As discussed before, there are many intermediary companies analysing a different set of data and getting a cut from advertising budgets in return. As a result, out of €40.3 billion advertising spend in 2014, publishers received €30,7 billion - including agency commissions (IHS Inc., 2015), and this figure should be even lower when agency commissions and publisher-side technology fees are included into calculation. Figure 3.3 Display Advertising Ecosystem (Luma Partners, n.d.) As it can be seen in Figure 3.3, there are many players between a marketer and a publisher. Hence, advertising costs are impacted negatively. On the other hand, advertisers would not want to show their ads in brand-damaging pages, meaning benefits of working with these intermediary companies should also be considered carefully.
  • 61. 55 3.8 Risks for Advertising Agencies Adverting agencies may also suffer from the usage of data, mainly in the ways explained below. 3.8.1 Increasing Complexity As mentioned before, in the earlier days of online advertising it was possible to measure how many times an ad was shown and how many times it was clicked. This data required more data like who saw the ads, how much of banners stayed in viewing areas of users and for how long, what banner sizes contributed more to sales and so on. Now the industry needs further data about smartphones, cross-channel attributions and so on. To make sense of all these data is not easy - if not impossible in some cases, thus, daily tasks of agencies are getting more challenging and confusing. Similarly, it is becoming harder to keep up with the industry since there are many technology players in the ecosystem, and each of them claims something else like doing a task better or specialising in another area. 3.8.2 Low Spending The business model of advertising agencies is mainly taking a share from the advertising budgets of the advertisers. If an advertiser spends $100K for instance, the agency may make around $20K (naturally this rate can vary). However, data may limit this spending in the ways explained below, which would affect agency revenues negatively. 3.8.2.1 Too Targeted = Low Spending? Data helps with targeted campaigns; however, spotting advertiser’s target audience on the web and showing ads to him/her may not be easy. This would cause the advertising budgets to be spent slower than usual; hence, it would affect agency’s revenues negatively. 3.8.2.2 Brand Safety = Low Spending? Data also helps with showing banners in websites or pages where there is no harmful content for the advertiser. But if a campaign’s safety rules are too
  • 62. 56 strict, spending may be slow as there will be few pages where the ads will be able to appear. 3.8.3 Benefits(!) of Fraud Although ad-fraud is a big problem for the online advertising industry, it may not be necessarily so for all the players in it. Some agencies benefit from low costs of fraudulent activities and their reports look “excellent” to advertisers when costs are low. In one example, an agency’s client was so happy with a campaign, where 90% of activity was fraudulent, that the client renewed contract with the agency (Davies, 2015). Thus, the more advertisers utilise data to detect fraudulent activities, the more agencies may suffer from “client happiness”. 3.8.4 Impacts on HR and Mindsets of Employees As discussed before, needed skills and qualifications from people who work in the online advertising industry have been changing. It is even more challenging from an agency point of view since there is a risk of losing a client if results are not satisfying. Hence, some agency employees ignore fraudulent activities, or falsify data, or do other “shady practices” (Davies, 2015). These practices may have some unexpected consequences for the industry in the future. 3.8.5 Over-reliance on Technology and Data As mentioned before, poor data quality is one of the problems in the industry. Also, not all technologies are perfect yet. Murphy (2016) believes agencies' over-reliance on technology and poor quality 3rd party data may paralyse and neglect core strategy. 3.9 Risks for Technology Providers Data creates some challenges for technology providers, mainly in the ways explained below.
  • 63. 57 3.9.1 Missing Data Points Data is at the core of many technology providers in the online advertising space. However, it is not always possible to capture all data points, which may result in poor decisions by technologies. For instance, when a user views a product, s/he might be retargeted. When s/he purchases that product, retargeting should be stopped immediately. However, this is not always the case, since either technical integration is not done properly or it is not possible to capture some data points at all. 3.9.2 Crowded Industry Apart from major companies like Google and Facebook, there are dozens of other companies (if not hundreds worldwide) in the advertising technology industry, and this number is growing constantly (Display AdTech, n.d.). It results not only in intense competition, but also getting “noticed” among so many players is becoming hard - even if the technology company’s proposition is unique. Therefore the industry has been consolidating (Ledbetter, 2015) and small players either go out of business or acquired by bigger ones. 3.10 Risks for Users There are two major ways data impacts users negatively. 3.10.1 Privacy Almost all companies in the online advertising industry are after data of users. Every bit of data like what they read, search, share or intend to buy is important. Therefore, many companies track the users collectively. However, not everyone welcomes being tracked. As discussed before, one of the top reasons why ad-blockers are used is because of concerns about privacy (Internet Advertising Bureau UK, 2015; Richards, 2015)
  • 64. 58 3.10.2 Increased Spending One of the ways data-driven marketing activities help to companies is; they drive sales and profitability (Forbes, 2015). For instance, retargeted users are 70% more likely to “convert” on advertisers website (Abramovich, 2013). It also means users spend more money online (and probably offline as well). Figure 3.4 and Figure 3.5 show how online spending increased in Europe and US. Figure 3.4 Online Spending in Europe (Statista, 2016d) Figure 3.5 Online Spending in the US (Statista, 2016e)
  • 65. 59 The increase in spending may not be purely due to the usage of data in online advertising, but it is clearly one of the factors. However, spending more and saving less is not necessarily a good thing for some users.
  • 66. 60 4 IMPACTS ON THE INDUSTRY 4.1 Privacy Concerns One of the biggest challenges the online advertising industry faces is privacy concerns of users and related regulations, which may impact the industry negatively. Many advertising technology companies, including the largest ones like Google and Facebook, base their algorithms mostly on user data. This data is collected and related to a user mostly through cookies (European Commission, n.d.) and in a way that keeps the user anonymous. However, some companies like Facebook work on other types of tracking methods (Jakubowski & Johnson, 2014) since a cookie is dependent on the browser of the user and it is difficult to identify the same user across browsers or devices. It might also mean a threat to the privacy of users. As a result, many users have some concerns about their online privacy. A recent study shows that 91% of US consumers do not want companies to collect information about them without their knowledge, even if they get discounts in return (Turow, Hennesy & Draper, 2015). Concerns about online privacy are rising in the UK as well (Ofcom, 2015). Google collect data from many sources, including videos watched on Youtube, content of emails, location and so on, which are used for various reasons including showing tailored ads (Google, 2016). Similarly, Facebook collects data from many sources, including what users like, their location, what they do on other Facebook services like WhatsApp and so on (Facebook, n.d.b.).