27. Creative Reallocate budget & Repeat Run media Attribute value Use analytics to Discover audiences RT Targeting & Site Customization Model expansion Refine creative
28. Port Your Data Everywhere. Analytics Site Customization CRM Video Email Display Mobile Offline
29. Data colors all media & channels not just cookies online Transparency will evolve into consumer models that treat anonymous targeting and PII targeting differentially DMP = salesforce.com for marketers
This is no longer an industrial economy purely driven by oil as its fuelIt is an information economy whose fuel is data. Data is very much like fuel because:Data is used by closed loop systems. It has valueIt also has to be refined to produce value – you have to separate the signal from the noiseData is unlike oil because:1. Non-rival good & Can be distributed freelyIt is growing at exponential rateToday I will discuss a few collisions which are shaping the debate on how we use data
Two trends will define the data industry in the next few yearsThey are opposing forcesThey will make some companies grow and others fail?Which way will this shake out?
Where can we learn from the past where these two forces collided in a big way? Music, movies, iTuneslots of people try to steal musicMost of the industry tried to fight the move to digital musicApple balanced making it controlled and secure with making the transactions VERY easy for the customerPricing models: subscription vs per use (netflixvs apple)
Allowed you to handle your own musicSecurity & Controlled distribution for sellersFair pricing for bothEasy, useful, usable, idiot proof for buyers
By 2005 – humanity created more data in a single year – then all data created in all history priorFor example, in 2005, experts predicted that the amount of digital information in the world was doubling every 1100 days. In 2007, that time has shortened to 11 months. Astudy by IBM predicts that sometime in the next few years the amount of information will double every 11 hours.
1. History sniffing2. Look-alike (free seeds)3. Fake rev share4 Fake cpm5. Look-alike (free targets)6. Sins of ommission
A picture maybe worth a thousand words…But proprietary data about a user is worth a thousand impressionsAt the end of the day when data causes sales, conversions or creates valuable targeted reachAdvertisers will pay
The DMP is the nexus of free data and paid for dataIt allows a marketer to indulge in all the great data they have amassedIt puts their data to workIt creates a workflow that attaches their data to useful media and targetingIt also makes it easy to show paid for data in the light of what a marketer needsPaid for data adds color
The DMP is the nexus of free data and paid for dataIt allows a marketer to indulge in all the great data they have amassedIt puts their data to workIt creates a workflow that attaches their data to useful media and targetingIt also makes it easy to show paid for data in the light of what a marketer needsPaid for data adds color
The rise of the machinesMath men versus mad menQuants versus creativesHow will this epic battle play out?Dave Morgan said “The problem is that the math parts of the applications dramatically overbalance the art”Joe Z says: TV is purchased based on rating points which are a proxy based on a proxy. That doesn’t make it correct. Being correct is far more difficult”Lets look at some pieces of the workflow that define this arm wrestling match
surveys to decide what are the top attributes of a target audience panel to decide which places have higher concentrations of people with those attributes.In the words of joe z – proxies upon proxiesGood news is that this old technique has been replaced by a far superior carbon based technique
It allows you to learn before you burn dollars.It applies analytics on all known data against your media and site conversions The human applies filtering to the machine based audiences that are discovered.Cookies/data color all media not just cookie You can improve your creatives, your products and your contextual targets all with dataLimited the use of cookie data to targeting is miopic
You can’t and shouldn’t automate the creative process. Even if you use dynamic content optimization
People have to decide how well each tactic is driving resultsModels help with cross channel attribution and what-if analysisBut the allocation decisions are driven by people
Finally – not everything is driven by an automated auctionA lot of the process of deal making is very one on one, very custom and very profitable
The world of the machine is improving, and optimizing a significant part of the display value chain
Real time bidding is radically shifting how we make media decisionsIt is driving the media market closer to the stock marketMany people fear the commoditization of they buy/sell processThis is process is inevitable
Not only can the machines execute the media, they can decide how to build up reachThis is done in two ways: look-alike modeling and social expansionBoth take a known target and increase reachThis used to be the job of the media planner
Site optimization is a case where the machine adds new value Here we can change each site to meet the needs of each personGet closer to Cheers where everyone greets Norm and Sam fixes the right drinkOnly way to do this is machine basedAt BlueKai we are seeing a lot of marketers gear up for this use case of data
Allocation decisions human, but the models that inform the decision are machine basedability to examine every impression, every page view and every past cookie attributedimensionality of the problem too large for humansattribution can pay for the whole infrastructure cost
So is this really a battle?Does one side have to win?What will the final picture look like?