Ad Fraud / Ad Blockingand Polluted Analytics
December 2015Augustine Fou, [email protected] 212.203.7239
November 2015 / Page 2marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Brief Agenda
• Ad Fraud Background
• What is Ad Fraud
• Impact of Ad Blocking
• How Fraud Pollutes Analytics
• Low Hanging Fruit – You Can Do NOW!
Ad Fraud Background
November 2015 / Page 4marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Percent of digital ad spend in programmatic: 70 - 75%
1995Hundreds of major sites.
2005Thousands of mainstream blogs.
2015Millions of “long-tail” websites.
November 2015 / Page 5marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Fraud continues upward as digital ad spend goes up
Digital ad fraudHigh / Low Estimates
plus best-guess
Published estimates
Digital ad spendSource: IAB 2014 FY Report
$ billions
E
November 2015 / Page 6marketing.scienceconsulting group, inc.
Dr. Augustine Fou
UPDATED: Full Year 2014 Digital Ad Spend – $50B
Impressions(CPM/CPV)
Clicks(CPC)
Leads(CPL)
Sales(CPA)
Search 38%$18.8B
Video 7%$3.5B
Lead Gen 4%$2.0B
10% Other$5.0B
Source: IAB, FY 2014 Internet Advertising Report, May 2015$42.5B
Display 16%$7.9B
Mobile 25%$6.2B$6.2B
CPM Performance
• classifieds• sponsorship• rich media
$7.0B
November 2015 / Page 7marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Bad guys go where the money is – impressions/clicks
Impressions(CPM/CPV)
Clicks(CPC)
Search$18.8B
86% digital spend
Display$7.9B
Video$3.5B Mobile
$6.2B$6.2B
Leads(CPL)
Sales(CPA)
Lead Gen$2.0B
Other$5.0B
• classifieds• sponsorship• rich media
estimated fraud
not at risk
(up from 84% in 2013)
November 2015 / Page 8marketing.scienceconsulting group, inc.
Dr. Augustine Fou
retail finance automotive telecom CPG entertainment pharma travel cons. electronics0
10
20
30
40
50
60
70
80
90
100
indexed spend share
indexed fraud rate
Ad fraud impacts every industry vertical
High CPC industries
Source: Ad spend share data from IAB, May 2015 | Fraud rate data from Integral Ad Science Q2 2014 Fraud Report
What is Ad Fraud?
November 2015 / Page 10marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Two main types of fraud and how each is generated
Impression (CPM) Fraud
(includes mobile display, video ads)
1. Put up fake websites and load tons of ads on the pages
Search Click (CPC) Fraud
(includes mobile search ads)
2. Use bots to repeatedly load pages to generate fake ad impressions (launder the origins of the ads to avoid detection)
1. Put up fake websites and participate in search networks
2a. Use bots to type keywords to cause search ads to load
2b. Use bots to click on the ad to generate the CPC revenue
November 2015 / Page 11marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Bots are the cause of all automated ad fraud
Headless BrowsersSeleniumPhantomJSZombie.jsSlimerJS
Mobile Simulators35 listed
Bots are made from malware compromised PCs or headless browsers (no screen) in datacenters.
November 2015 / Page 12marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Bots range from simple to advanced; do different tasks
Malware (on PCs)Botnets (from datacenters)
Toolbars (in-browser)Javascript (on webpages)
November 2015 / Page 13marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Bot fraud observed as high as 100%
Source: ANA / White Ops Study Published December 2014 [PDF]
display ads
11%
25%
video ads
23%
50%
sourced traffic
52%
100%
ANA/WhiteOps Study
What We’ve Seen
Case 1 Case 2
Why are bad bots so hard to identify?
November 2015 / Page 15marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Bad guy’s advanced bots are not on any industry list
10,000bots observed
in the wild
user-agents.org
bad guys’ bots3%
Dstillery, Oct 9, 2014_“findings from two independent third parties,
Integral Ad Science and White Ops”
3.7%Rocket Fuel, Sep 22, 2014
“Forensiq results confirmed that ... only 3.72% of impressions categorized as high risk.”
2 - 3%comScore, Sep 26, 2014
“most campaigns have far less; more in the 2% to 3% range.”
detect based on industry bot list
“not on any list”disguised as normal browsers –
Internet Explorer; constantly adapting to avoid detection
November 2015 / Page 16marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Example of filtering using bot lists – good, but not enough
Google Analytics filters visits using official bot lists
Bad guy bots are not on those lists and don’t declare themselves honestly; they pretend to be browsers like Internet Explorer, Safari, etc.
“bad guy bots are not on industry lists”
November 2015 / Page 17marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Humans vs “honest” bots vs fraudulent bots
Confirmed humans• found page via search• observed events (mouse
click with coordinates)
“Honest” bots• search engine crawlers• declare user agent honestly• observed to be 1 – 5% of
websites’ traffic
Fraud bots• come from data centers• malware compromised PCs• deliberately disguise user
agent as human users
November 2015 / Page 18marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Mitigation does not require developers or statisticians
Sites or ad networks that have high percentage of confirmed bots are blacklisted from ad-serving or ad spend to those sites is reduced
In-ad (display ads served)On-site (clients’ websites)
Sources of traffic that have high incidence of bots are added to ad-serving blacklists and filtered in analytics reports
Impact of Ad Blocking
November 2015 / Page 20marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Ad blocking user growth continues to soar
Source: PageFair / Adobe Aug 2015
November 2015 / Page 21marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Ad blocking as a percent of users
Source: PageFair / Adobe Aug 2015
Europe: 8% - 38%U.S.: 8% - 17%
November 2015 / Page 22marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Estimated economic impact of ad blocking
Source: PageFair / Adobe Aug 2015
Global economic impact: $41BU.S. economic impact: $20B
November 2015 / Page 23marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Directly measured ad blocking rate
Non-mobile Mobile
Ad Block
No Ad Block 53.6%
15.4%
25.6%
5.4%
29% 21%
Overall Average
79.2%
20.8%
26% Ad Blocking Rate
* percentages represent portion of data from N = 10 million sample
69.0% 31.0%Column Totals
Pollution of Analytics
November 2015 / Page 25marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Bot activity pollutes quantity metrics
• Bots can be programmed to send as much traffic and generate as many impresisons and clicks as the advertiser wants
By systematically reducing ad spend to ad networks and sites that had the highest bots, and increasing allocation to premium publishers, the advertiser increased ad impressions served to humans, and lowered those served to bots.
November 2015 / Page 26marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Bot activity pollutes quality metrics
• Bots can manipulate bounce rates, click through rates, time on site, pages per visit; These engagement metrics appear to be tuned to 47 – 63%; pages per session averaged 2.03; and time on site was 1 – 2 minutes.
November 2015 / Page 27marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Bot activity pollutes conversion metrics
378411
357 361
512 495525
409
595
536 552596
437 452
380425
532489
592 584
403 416 415
587 570
490463
516
400 389418
Avg. 475 conversions /day
Avg. 3,526 sessions /dayAvg. 6,636 sessions /day
24% confirmed humans
-47%
Avg. 473 conversions /day
40% confirmed humans
0%
5%
10%
15%
20%
25%
Avg. 7.1% conversion rate
Avg. 13.5% conversion rate
“doubling humans, doubles conversion rates”
November 2015 / Page 28marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Bad guys hide fraud by passing fake parameters
Click thru URL passing fake source “utm_source=msn”
fake campaign“utm_campaign=Olay_Search”
http://www.olay.com/skin-care-products/OlayPro-X?utm_source=msn&utm_medium=cpc&utm_campaign=Olay_Search_Desktop_Category+Interest+Product.Phrase&utm_term=eye%20cream&utm_content=TZsrSzFz_eye%20cream_p_2990456911
November 2015 / Page 29marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Bad guys fake KPIs, trick measurement systemsBad guys have higher CTR Bad guys have higher viewability
AD
Bad guys stack ads above the fold to fake 100% viewability
Good guys have to array ads on the page – e.g. lower average viewability.
What you can do NOW
November 2015 / Page 31marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Recommendations
1. Don’t panic; but also don’t be complacent – directly measure the amount of fraud that is impacting your digital ad spend and continuously mitigate.
2. Focus on the details – don’t assume someone else has taken care of the problem; take small, simple steps at low to no cost – e.g. look in analytics for referring sites that have 100% bounce and 0:00 time on site.
3. Update KPIs to focus on things that are not easily faked (i.e. don’t focus on number of impressions, clicks, or visits); focus on “conversion events” like purchases or other human actions.
November 2015 / Page 32marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Normal Weekday vs Weekend Traffic Patterns
weekends weekends weekends weekends
weekdays weekdays weekdays weekdays
Natural website pattern is weekends have lower traffic
November 2015 / Page 33marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Typical Hour-of-Day Pattern
humans sleeping humans awake; visiting websites
November 2015 / Page 34marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Humans Sleep At Night
Hourly traffic charts show lower traffic at night (as expected because humans sleep at night)
Unusual traffic patterns with no normal night time trends visible, likely due to bot activity
November 2015 / Page 35marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Humans Visit via Search
humans find sites via search, during waking hours
Bot traffic adds anomalous spikes to pattern
November 2015 / Page 36marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Search vs Non-Human Traffic
notice the timing
hour-of-day pattern
November 2015 / Page 37marketing.scienceconsulting group, inc.
Dr. Augustine Fou
Closeup by Hour of Day
6 am5 am 2 am 3 am 3 am 2 am 3 am18396 sessions 162 184 178 159 156
Sunday
85% avg bounce rate; 100% peak bounce rate
November 2015 / Page 38marketing.scienceconsulting group, inc.
CONFIDENTIAL
These advanced bots also faked some Goal EventsGoal events that are based on page visits and video plays, could be (and were) faked.
page visit goal page visit goal video play goal
November 2015 / Page 39marketing.scienceconsulting group, inc.
CONFIDENTIAL
But, there was no motive to fake other goals – e.g. pledges
Other goals like pledges and downloads were not faked (faking downloads would cost them server resources).
make a pledge curriculum download
“Bots don’t make donations!”
November 2015 / Page 40marketing.scienceconsulting group, inc.
CONFIDENTIAL
Despite traffic loss, real human goals did not changeDespite losing all of the traffic from these fake/fraud sites, there was no change to the number of pledges and downloads, during the same period of time.
102,231 sessions
0 sessions
Conversion event 1
Conversion event 2
About the Author
November 2015 / Page 42marketing.scienceconsulting group, inc.
CONFIDENTIAL
Dr. Augustine Fou – Recognized Expert on Ad Fraud
2013
2014
2015SPEAKING ENGAGEMENTS / PANELS4A’s Webinar on Ad Fraud – October Digital Ad Fraud Podcast – JanuaryProgrammatic Ad Fraud Webinar – MarchAdCouncil Webinar on Ad Fraud - AprilTelX Marketplace Live – JuneARF Audience Measurement – JuneIAB Webinar on Ad Fraud / Botnets - September [email protected] | 212.203.7239
November 2015 / Page 43marketing.scienceconsulting group, inc.
CONFIDENTIAL
Harvard Business Review – October 2015
Excerpt:
Hunting the Bots
Fou, a prodigy who earned a Ph.D. from MIT at 23, belongs to the generation that witnessed the rise of digital marketers, having crafted his trade at American Express, one of the most successful American consumer brands, and at Omnicom, one of the largest global advertising agencies. Eventually stepping away from corporate life, Fou started his own practice, focusing on digital marketing fraud investigation.
Fou’s experiment proved that fake traffic is unproductive traffic. The fake visitors inflated the traffic statistics but contributed nothing to conversions, which stayed steady even after the traffic plummeted (bottom chart). Fake traffic is generated by “bad-guy bots.” A bot is computer code that runs automated tasks.