Bots: How to Find Them &
Keep Them Out of Digital Advertising
Questions?
Attendees should ask questions by typing into
the Question box on the GoToWebinar user
interface at any time during the presentations.
– We will create a queue and answer as
many questions as possible following
the presentations.
– Additional questions should be directed
to Nicole Horsford, Nicole@iab.net
Agenda
 Setting the Stage
 Speakers
● Marvin Hidalgo Senior Product Manager, Integral Ad Science
● Vamshi Sriperumbudur VP of Platform Marketing, Yume
● Elizabeth Cowley VP Global Partnerships, Viewster
 Open Discussion / Q&A
Marvin Hidalgo
Senior Product Manager
What Are Bots?
Bots noun ˈbäts a device or piece of software that
can execute commands, reply to messages, or
perform routine tasks, or problem routine tasks with
minimum human intervention
Illegal bots noun (ˈ)i(l)-ˈlē-gəlˈbäts computers
that are compromised and whose security defenses
have been breached and control conceded to a third
party
Botnet: noun ˈbät net a collection of bots
communicating with command center in order to
perform tasks
Fraud: Why Does It Take Place?
• Supply and Demand
• Because of the way
that we buy media:
• Eyeballs (CPM)
• Action Taken
(CPC, CPA)
• Because it’s easy for
hackers
Who Are the Participants?
Profile
Hacker:
Sex: Male
Age:18-35
Location: Eastern
Europe, Asia
Background: good
computer skills
BotnetOperator:
Sex: Male
Age: 34+
Location:Eastern Europe
Characteristics:Disregard
to law, confident,driven
by money
TypicalInfected Computer
Owner:
Technologicallychallenged
Owns a dated computer and
software
Suburban,rural, household
without kids
Unlikelyto own a smart
phone/tablet
How are Bots Detected ?
First we look at behavioral patterns
We flag the following un-human signals:
Cookies that are deleted at the end of activity
cycle
Intense activity
Reoccurring activities patterns/levels
At this point: some bots are detected, others are
able to go undetected
Next – we look at each impression
• Signals that are untypical for human
• Density of page loads
• Density of page visits
• Untypical distribution of browsers
• Browser spoofing
• Conflicting measurement results
• Was the impression traded in a suspicious
way
Cross validate all of the above and determine
validity of signals and patterns
Bot
…or not
How do Publisher’s approach Fraud?
Proactive
Passive
Pretend the problem
doesn’t exist
Knowingly or unknowingly sell bot traffic
Able to eliminate
some of the bot traffic
Eliminate most to
all bot traffic
Partially address the problem:
• Use a subpar list based solution
• Run the technology only on part of
the inventory
Are serious about fraud:
• Use cutting edge
technology to vet 100% of
inventory
Vamshi Sriperumbudur
VP of Platform Marketing
Ad Industry | Market Opportunity
$517
$544 $568
$596
$117 $132 $146 $160
2013 2014 2015 2016
Global Ad Spend (in billions)
Total Ad Spend
Total Digital Spend
$171 $179 $188 $198
$42 $50 $57 $65
2013 2014 2015 2016
US Ad Spend (in billions)
Total Ad Spend
Total Digital Spend
Source: eMarketer, Dec. 2013 & Mar. 2014 reports
44% 42%
51%
61%
0%
10%
20%
30%
40%
50%
60%
70%
Q1 Q2 Q3 Q4
19%
22%
27% 25%
Q1 Q2 Q3 Q4
Ad Industry | Challenges
Non-human
Mobile Activity
Non-human
Web Activity
2013
Source: Solve Media, Mar. 2014
2013 US Non-human Traffic – Online & Mobile
Challenge: In 2013, non-human Online Activity in US grew 40% & non-human Mobile Activity grew 30%
Solution: Publishers and Advertisers should prioritize investments in anti-fraud technologies
CTV platform
needs review
Traffic Quality Solutions
Viewability
Tracking and measuring
ad positions on publisher
pages and properties
Fraud
Detection
Monitoring of non-human
traffic resulting from
traffic fraud designed to
manipulate ad-serving
counts
Brand
Safety
Consideration of the
content quality and
context that appears next
to the ad
Solutions | Viewability
At least 50% of the ad is visible on the users screen for at least 1 second
Above the fold
View Area
Width x
Height
Player
Width x Height
3 Player Location
3
4 View Area
4
2 Player Size
2
5 No auto-play
5
1 Minimum Player Visibility
1
Solutions | Fraud Detection
1. Diligent evaluation of every publisher before adding to
the network to make sure they meet the standards
2. SDK / Process for Placement Quality – Algorithms to
grade value of each video ad impression and assign
quality score. Fraud detection capabilities to address:
• Click Fraud (bots browse/click on bogus & legit sites)
• Impression Scams (bogus ad requests)
• Ad Stacking (stacked muted auto-play ads behind in-
view video ad)
• Ad Clutter (multiple ads in close proximity)
3. SDK & Data – Machine learning for mimicking human
behavior to downgrade fraudulent inventory
4. SDK & Surveys – First-party audience survey
response rate to gauge inventory quality
5. Partnerships with right vendors who specialize in fraud
detection
Publisher Site
Domain blacklisted
Player not in view
Safe domain, player in-view
Additionally, review every site/app of a Publisher and their syndication partners:
 Only click-to-play or user initiated video ad playback
 No skip button for 15 or 30s ads
 No muted audio
 Post-rolls after video content not acceptable
Publisher Syndicated Sites
Solutions | Brand Safety
2 Context
2
1 Content
1
Solutions | YuMe in The Press
“YuMe wants to ensure that there's a human
behind every impression. These
improvements to our fraud detection system
and PQI, combined with our newly-launched
anti-spam best practices, will provide an
essential layer of comfort for those planning
media buys or monetizing ad inventory.”
- Jayant Kadambi, CEO & Co-Founder, YuMe
According to Radar Research, up to 50
percent of all ad impressions are never
seen by the intended audience. Inclusion of
unseen impressions in campaign reporting
results in inaccurate metrics.
This industry-first capability allows YuMe to
prevent ads from running in video players
that have been embedded on inappropriate
websites, and to work with publishers to
constantly monitor and improve the list of
sites where their syndicated and user-
embeddable players are appearing.
Oct 2013 Nov 2011 Jun 2010
Elizabeth Cowley
VP Global Partnerships
What is Viewster?
Founded in 2007 & HQ: Zurich, Switzerland with offices in US, UK, & Australia
VOD Service with focus on TV and Movie content
Available globally in 10 languages on PC and 1M connected devices
# 18 US Video Content Provider by comScore reaching 18M unique visitors/month
Flavors of Bad Traffic
•Non-Human Traffic – bots
•Human but not Viewable Traffic – hidden iFrames
“<iframe src=‘http://StealBrandMoney.com style=‘visibility:hidden;position:absolute;left0;top0;’></iframe>”
•Human & Viewable Traffic - Fraudulent Arbitrage
– Browser Plug Ins Spoof URL
– In-banner sold as onsite
– Multiple Players on same page
– Ads run in pop-under pages
– Ads auto-start BTF
Buyer is sold…
But gets this
Bad Traffic Ecosystem
Fraudsters sell bad Traffic
Because they can – Buyer
metrics are fakable.
Bad traffic hits the market and real publishers
with real video [and real associated costs] are
forced to compete with fake traffic that has no
content cost.
Real publishers look to
buy affordable traffic to
compete
Bad Traffic enters the marketplace
DIRECTLY through exchanges & lower-end
adnets. It can then be laundered and sold as
premium.
INDIRECTLY through legit publishers buying
traffic to compete.
Removing Bad Traffic from the Marketplace
Stop Buying It!
Publishers, AdNets and AdExchanges need to
stop buying bad traffic.
Brands Can Help Too
•Brands should expect to pay a decent price for
decent inventory and a high price for premium
inventory.
•Buyers should consider campaign metrics tied
to human activity.
– Increased brand recognition
– Increased sales
– Hijack device camera to confirm eyeball on ad? [just kidding…]
How to Spot Bad Traffic
• Before Buying
– Is the price unreasonably low?
– Check Alexa Rank – does the site have real traffic?
– Check referring URLs in Alexa – lets you know where the traffic’s traffic is
coming from?
– Check against known offenders list
• Run Test Campaign
– Manual checks to verify placement
– 3rd Party Traffic Verification such as IAS, Adometry or similar
– 3rd Party Viewability Tool such as OpenVV
• After Buying
– Don’t fall for a bait & switch. Continue to monitor manually and via 3rd party
technologies.
Do not underestimate the value of a relationship between
buyer and seller.
Publisher Challenges with 3rd Party
Verification Tools
Publisher Traffic GOOD
BAD
Publishers need actionable data not just a verdict of good or bad.
A good step would be if 3rd party verification technologies would
work with publishers to determine where their algorithms are
producing data useful to buyers and where it is not.
There is no standard for “good” or “viewable” traffic therefore technologies do not
agree with one another.
IAB 3MS Initiative
“Making Measurement Make Sense” – intended standardize digital media metrics and promote
shift from gross “served” impressions to audience-based “viewed” impressions (50% viewable for
a minimum of 1 second).
IAB Safe Frame 1.0 Initiative
Open source iframe code that removes browser-type restrictions on the geometric approach and
enables the same level of verification coverage as browser optimization approach.
TOGI Task Force
“Traffic of Good Intent” task force created to address bad traffic.
Industry Best Practice Initiatives
Brendan Riordan-Butterworth
Director, Technical Standards
Marvin Hidalgo
Senior Product Manager
Open Discussion / Q&A
Elizabeth Cowley
VP Global Partnerships
Vamshi Sriperumbudur
VP of Platform Marketing

Iab bots how to_find_them_webinar_2014_03_27

  • 1.
    Bots: How toFind Them & Keep Them Out of Digital Advertising
  • 2.
    Questions? Attendees should askquestions by typing into the Question box on the GoToWebinar user interface at any time during the presentations. – We will create a queue and answer as many questions as possible following the presentations. – Additional questions should be directed to Nicole Horsford, Nicole@iab.net
  • 3.
    Agenda  Setting theStage  Speakers ● Marvin Hidalgo Senior Product Manager, Integral Ad Science ● Vamshi Sriperumbudur VP of Platform Marketing, Yume ● Elizabeth Cowley VP Global Partnerships, Viewster  Open Discussion / Q&A
  • 4.
  • 5.
    What Are Bots? Botsnoun ˈbäts a device or piece of software that can execute commands, reply to messages, or perform routine tasks, or problem routine tasks with minimum human intervention Illegal bots noun (ˈ)i(l)-ˈlē-gəlˈbäts computers that are compromised and whose security defenses have been breached and control conceded to a third party Botnet: noun ˈbät net a collection of bots communicating with command center in order to perform tasks
  • 6.
    Fraud: Why DoesIt Take Place? • Supply and Demand • Because of the way that we buy media: • Eyeballs (CPM) • Action Taken (CPC, CPA) • Because it’s easy for hackers
  • 7.
    Who Are theParticipants? Profile Hacker: Sex: Male Age:18-35 Location: Eastern Europe, Asia Background: good computer skills BotnetOperator: Sex: Male Age: 34+ Location:Eastern Europe Characteristics:Disregard to law, confident,driven by money TypicalInfected Computer Owner: Technologicallychallenged Owns a dated computer and software Suburban,rural, household without kids Unlikelyto own a smart phone/tablet
  • 8.
    How are BotsDetected ? First we look at behavioral patterns We flag the following un-human signals: Cookies that are deleted at the end of activity cycle Intense activity Reoccurring activities patterns/levels At this point: some bots are detected, others are able to go undetected Next – we look at each impression • Signals that are untypical for human • Density of page loads • Density of page visits • Untypical distribution of browsers • Browser spoofing • Conflicting measurement results • Was the impression traded in a suspicious way Cross validate all of the above and determine validity of signals and patterns Bot …or not
  • 9.
    How do Publisher’sapproach Fraud? Proactive Passive Pretend the problem doesn’t exist Knowingly or unknowingly sell bot traffic Able to eliminate some of the bot traffic Eliminate most to all bot traffic Partially address the problem: • Use a subpar list based solution • Run the technology only on part of the inventory Are serious about fraud: • Use cutting edge technology to vet 100% of inventory
  • 10.
    Vamshi Sriperumbudur VP ofPlatform Marketing
  • 11.
    Ad Industry |Market Opportunity $517 $544 $568 $596 $117 $132 $146 $160 2013 2014 2015 2016 Global Ad Spend (in billions) Total Ad Spend Total Digital Spend $171 $179 $188 $198 $42 $50 $57 $65 2013 2014 2015 2016 US Ad Spend (in billions) Total Ad Spend Total Digital Spend Source: eMarketer, Dec. 2013 & Mar. 2014 reports
  • 12.
    44% 42% 51% 61% 0% 10% 20% 30% 40% 50% 60% 70% Q1 Q2Q3 Q4 19% 22% 27% 25% Q1 Q2 Q3 Q4 Ad Industry | Challenges Non-human Mobile Activity Non-human Web Activity 2013 Source: Solve Media, Mar. 2014 2013 US Non-human Traffic – Online & Mobile Challenge: In 2013, non-human Online Activity in US grew 40% & non-human Mobile Activity grew 30% Solution: Publishers and Advertisers should prioritize investments in anti-fraud technologies CTV platform needs review
  • 13.
    Traffic Quality Solutions Viewability Trackingand measuring ad positions on publisher pages and properties Fraud Detection Monitoring of non-human traffic resulting from traffic fraud designed to manipulate ad-serving counts Brand Safety Consideration of the content quality and context that appears next to the ad
  • 14.
    Solutions | Viewability Atleast 50% of the ad is visible on the users screen for at least 1 second Above the fold View Area Width x Height Player Width x Height 3 Player Location 3 4 View Area 4 2 Player Size 2 5 No auto-play 5 1 Minimum Player Visibility 1
  • 15.
    Solutions | FraudDetection 1. Diligent evaluation of every publisher before adding to the network to make sure they meet the standards 2. SDK / Process for Placement Quality – Algorithms to grade value of each video ad impression and assign quality score. Fraud detection capabilities to address: • Click Fraud (bots browse/click on bogus & legit sites) • Impression Scams (bogus ad requests) • Ad Stacking (stacked muted auto-play ads behind in- view video ad) • Ad Clutter (multiple ads in close proximity) 3. SDK & Data – Machine learning for mimicking human behavior to downgrade fraudulent inventory 4. SDK & Surveys – First-party audience survey response rate to gauge inventory quality 5. Partnerships with right vendors who specialize in fraud detection
  • 16.
    Publisher Site Domain blacklisted Playernot in view Safe domain, player in-view Additionally, review every site/app of a Publisher and their syndication partners:  Only click-to-play or user initiated video ad playback  No skip button for 15 or 30s ads  No muted audio  Post-rolls after video content not acceptable Publisher Syndicated Sites Solutions | Brand Safety 2 Context 2 1 Content 1
  • 17.
    Solutions | YuMein The Press “YuMe wants to ensure that there's a human behind every impression. These improvements to our fraud detection system and PQI, combined with our newly-launched anti-spam best practices, will provide an essential layer of comfort for those planning media buys or monetizing ad inventory.” - Jayant Kadambi, CEO & Co-Founder, YuMe According to Radar Research, up to 50 percent of all ad impressions are never seen by the intended audience. Inclusion of unseen impressions in campaign reporting results in inaccurate metrics. This industry-first capability allows YuMe to prevent ads from running in video players that have been embedded on inappropriate websites, and to work with publishers to constantly monitor and improve the list of sites where their syndicated and user- embeddable players are appearing. Oct 2013 Nov 2011 Jun 2010
  • 18.
  • 19.
    What is Viewster? Foundedin 2007 & HQ: Zurich, Switzerland with offices in US, UK, & Australia VOD Service with focus on TV and Movie content Available globally in 10 languages on PC and 1M connected devices # 18 US Video Content Provider by comScore reaching 18M unique visitors/month
  • 20.
    Flavors of BadTraffic •Non-Human Traffic – bots •Human but not Viewable Traffic – hidden iFrames “<iframe src=‘http://StealBrandMoney.com style=‘visibility:hidden;position:absolute;left0;top0;’></iframe>” •Human & Viewable Traffic - Fraudulent Arbitrage – Browser Plug Ins Spoof URL – In-banner sold as onsite – Multiple Players on same page – Ads run in pop-under pages – Ads auto-start BTF Buyer is sold… But gets this
  • 21.
    Bad Traffic Ecosystem Fraudsterssell bad Traffic Because they can – Buyer metrics are fakable. Bad traffic hits the market and real publishers with real video [and real associated costs] are forced to compete with fake traffic that has no content cost. Real publishers look to buy affordable traffic to compete Bad Traffic enters the marketplace DIRECTLY through exchanges & lower-end adnets. It can then be laundered and sold as premium. INDIRECTLY through legit publishers buying traffic to compete.
  • 22.
    Removing Bad Trafficfrom the Marketplace Stop Buying It! Publishers, AdNets and AdExchanges need to stop buying bad traffic. Brands Can Help Too •Brands should expect to pay a decent price for decent inventory and a high price for premium inventory. •Buyers should consider campaign metrics tied to human activity. – Increased brand recognition – Increased sales – Hijack device camera to confirm eyeball on ad? [just kidding…]
  • 23.
    How to SpotBad Traffic • Before Buying – Is the price unreasonably low? – Check Alexa Rank – does the site have real traffic? – Check referring URLs in Alexa – lets you know where the traffic’s traffic is coming from? – Check against known offenders list • Run Test Campaign – Manual checks to verify placement – 3rd Party Traffic Verification such as IAS, Adometry or similar – 3rd Party Viewability Tool such as OpenVV • After Buying – Don’t fall for a bait & switch. Continue to monitor manually and via 3rd party technologies. Do not underestimate the value of a relationship between buyer and seller.
  • 24.
    Publisher Challenges with3rd Party Verification Tools Publisher Traffic GOOD BAD Publishers need actionable data not just a verdict of good or bad. A good step would be if 3rd party verification technologies would work with publishers to determine where their algorithms are producing data useful to buyers and where it is not. There is no standard for “good” or “viewable” traffic therefore technologies do not agree with one another.
  • 25.
    IAB 3MS Initiative “MakingMeasurement Make Sense” – intended standardize digital media metrics and promote shift from gross “served” impressions to audience-based “viewed” impressions (50% viewable for a minimum of 1 second). IAB Safe Frame 1.0 Initiative Open source iframe code that removes browser-type restrictions on the geometric approach and enables the same level of verification coverage as browser optimization approach. TOGI Task Force “Traffic of Good Intent” task force created to address bad traffic. Industry Best Practice Initiatives
  • 26.
    Brendan Riordan-Butterworth Director, TechnicalStandards Marvin Hidalgo Senior Product Manager Open Discussion / Q&A Elizabeth Cowley VP Global Partnerships Vamshi Sriperumbudur VP of Platform Marketing