Data from FouAnalytics, on-site measurement and in-ad measurement was compared to DBM exchange data for 26 exchanges, 7.5 trillion impressions (30 day period) to analyze browser market share -- specifically Safari/iOS.
Findings include: 1) bots pretending to be Safari/iOS outnumber real Safari users 5 to 1, and 2) there is a 1.5X average surplus of Safari impressions available on exchanges compared to unique cookies.
Unraveling the Mystery of The Circleville Letters.pptx
Fake Safari vs Real Safari: 5X More Bots Pretend To Be Safari Than Actual Users
1. December 2019 / Page 0marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
GROUND TRUTH
Fake Safari vs Real Safari
December 2019
Augustine Fou, PhD.
acfou [at] mktsci.com
212. 203 .7239
2. December 2019 / Page 1marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Key Takeaways
• There are 5X more bots pretending to be Safari than actual human
users of Safari browser – confirmed humans’ Safari share was 3% while
confirmed bots’ Safari share was 16%
• There is also a surplus of Safari impressions on exchanges - Safari share
based on impressions (35%) was 1.5X the Safari share based on unique
cookies (24%)
• The two problems above – 5X more fake Safari users and 50% more fake
safari impressions – compound each other.
SO WHAT?
• If you see greater than 3% share of Safari in your digital programmatic
campaigns, you should investigate for ad fraud.
3. December 2019 / Page 2marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Browser share stats
by other public sources
4. December 2019 / Page 3marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Reported Safari Share Varies
Source: caniuse.com: 4.1%
Source: netmarketshare: 3.9%
Source: statcounter: 16.7%
Source: w3counter: 14.6%
Which is right?
5. December 2019 / Page 4marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Ground truth measurement
• Must be measured directly, on-site (code on page)
• Must be confirmed humans only, excluding bots
(bots significantly skew the results)
• Must be U.S. only - variations in browser share by
geographic location (country) are significant
• Exchange data combines Safari (desktop) and iOS
(mobile) into a single browser “Safari”
• Browser share represents overall percentages
without separation of desktop versus mobile
6. December 2019 / Page 5marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
On-Site Measurement
Humans vs Bots
Data Source: FouAnalytics
10. December 2019 / Page 9marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Bots pretending to be Safari browsers are
5X higher than humans using Safari,
based on data from 15 publishers (50
sites), directly measured by FouAnalytics.
11. December 2019 / Page 10marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Exchange Browser Share
Data Source: DBM, Last 30 Days, U.S. Only
12. December 2019 / Page 11marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Share by Unique Cookie
66%
22%
10%
50%
26%
43%
2%
41%
34%
12%
21%
35%
3%
34%
31%
47%
30%
23%
10%
17%
5%
34%
15%
25%
8%
22%
14%
24%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
By Unique Cookie
Safari
Unknown
Opera
Mobile
IE
Firefox
Edge
Chrome
Android
Safari + iOS
Data Source: DBM, Last 30 Days, U.S. Only
13. December 2019 / Page 12marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Share by Impressions
47%
46%
32%
70%
33%
26%
32%
47%
15%
25%
38%
62%
15%
39%
22%
39%
45%
17%
32%
23%
33%
51%
15%
47%
52%
15%
28%
35%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
By Impressions
Safari
Unknown
Opera
Mobile
IE
Firefox
Edge
Chrome
Android
Safari + iOS
Data Source: DBM, Last 30 Days, U.S. Only
15. December 2019 / Page 14marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Analysis of data from 26 exchanges (7.5
trillion potential impressions, 30 day
period) shows 1.5X more impressions
share than unique cookies share.
16. December 2019 / Page 15marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Corroboration
How do we know the data is reliable?
17. December 2019 / Page 16marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
In-Ad and On-Site vs Exchange
52%
42% 43% 46% 50%
63%
68%
35%
41%
47%
41%
35%
28% 24%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
IN-AD 1 IN-AD 2 IN-AD 3 IN-AD AVERAGE AVERAGE by
Impressions
ON-SITE AVERAGE AVERAGE By Cookie
Safari
Unknown
Samsung Browser
IE
Firefox
Edge
Chrome
On-SiteIn-Ad
• In-Ad measurement corroborates average by impressions
• On-Site measurement corroborates avg by unique cookie
18. December 2019 / Page 17marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Measurement Notes
• Browser share by our in-ad measurement
correlates well to exchange data browser
share by impressions (measures ads)
• Browser share by our on-site measurement
correlates well to exchange data browser
share by unique cookies (measures users)
• This gives us confidence that the data is
reliable and usable because it is corroborated
across different data sets.
19. December 2019 / Page 18marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Recap
• There are 5X more Safari bots than actual
human users of Safari
• There is a 1.5X average surplus of Safari
impressions on exchanges
• Bots have a motive to pretend to be Safari
and iOS because they can get higher bids
(buyers believe iOS/Safari users are affluent,
desirable target audience)
20. December 2019 / Page 19marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Who am I?
• I am a digital marketer of 23+ years
• I look at analytics data every day
• I audit campaigns for fraud and other
bad sh*t that lowers performance
• I teach clients how to spot fraud and
make optimizations to improve
outcomes
21. December 2019 / Page 20marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
FouAnalytics
Site
Analytics
Media
Analytics
“see fou yourself”
• alternative to Google Analytics
• secure, hardened against attack
• shows all details, no black box
• innovated w/ practitioners
• verify your own media/ads
• fraud detection baked-in
• shows details for decisioning
• recommended optimizations
for #publishers for #marketers