Presentation on Brand Safety measures undertaken by UM London.
It's our view that agencies need to lead the charge against ad fraud. We use brand safety software as standard to protect clients' interests.
7. Ad fraud has become more prevalent as the
supply chain complicates
8. A fraudulent impression is one that has
no potential to be seen by a human user
Impression served to a bot
Invisible impression served to a human user
Defining ad fraud today?
9. Ad Stacking
Placing multiple ads
on top of one other in
a single ad placement,
with only the top ad in
view
Examples of ad fraud
AD
Illegal bots
Non human user
agents producing
HTTP web traffic
Domain Spoofing
Pixel Stuffing
Stuffing an entire ad-
supported site into a
1x1 pixel
Make advertisers think
fake sites are those of
reputable publishers
10. How do ad fraud bots work
Exchanges /
Networks
Monetisation3
Premium
sites
Profile
creation2
Bot infection1
Source: Integral Ad Science
11. Not all bots are bad
Trading Bot Media Bot Spider Bot
12. How do we know they are bots?
Bot clicks vs humans
16. We partner with & support industry leaders
Industry Tools Industry Bodies
Global and UK Relationships
17. We use Integral Ad Science to block all
fraudulent traffic*
98% visibility of the web
*UM London
18. Ad fraud detection technology identifies fraudulent impressions, and
thus infected machines
By subscribing to an Anti-Targeting Pixel Solution, UM is alerted in real-
time of infected machines and can avoid serving impressions to bots
We use machine level detection
18
Source: Integral Ad Science
19. TRAQ score for programmatic media:
Bid on the highest quality inventory
Pre-bid Integrations
• Exclude fraudulent
impressions from pre-bid
targeting.
• Apply anti-targeting segment
or supply ad server pixel to
eliminate fraud in real-time.
20. Brandsafety:Amulti-layeredapproachtoensureadsrunin
appropriatesitesadjacenttoappropriatecontent
Practice Method
Inventory Validation Human curation; all Inventory is vetted by
Cadreon team for adherence to IAB standard
content and maturity classifications
Cadreon Universal Ban List Cadreon’s Universal Ban List (UBL) is applied
across all buys. Sites are banned for inappropriate
content, fraudulent activity and other non-
compliant activities. The list is updated regularly
3rd Party Data Validation 3rd party platform integrated to the buying of
quality media to ensure safe environments and
quality exposures
21. Practice Method
Ad blocking against
inappropriate content
Ad safety partners have C. 98% visibility of the web allowing
UM to set content appropriateness thresholds by client
Viewability Optimisation Optimise towards inventory in view according to industry
standard metrics
Ad collision UM Can block more than one campaign ad appearing on a
page
Brand safety: A multi-layered approach to ensure
ads run in appropriate sites adjacent to
appropriate content
22. Ad fraud is prevalent & has been for a long time
The best example of a fraudulent ad impression is one that can never be viewed by a human
Bots are estimated to account for 12% of all ad impressions
Not all bots are malicious
It is the responsibility of agencies to use best in class ad fraud software to block them or work with industry accredited media partners
Brand safety is wider than fraud so a multi layered approach is needed
In Summary
Bad practices have been around for a while.
The first spam email sent in 1978
400 of the 2600 people on ARPAnet, “first Internet.”
Was 15%
Now 70%
http://www.themarysue.com/first-spam-email/
Big Headlines
1/3 Traffic
FT 72% programmatic impressions on exchanges fraudulent
Everyone used to buy pay per click
Googles ad model
11 milliseconds
Many opportunities for bad practice
Many definitions but this is the best
3 stage process:
BOT INFECTION
1 - User download
2 – Trojan horse, or malicious code, is delivered as part of download
3 – Laptop/machine is now infected with bot
PROFILE CREATION
1 – The bot now only needs an internet connection
2 – Bots premium sites Telegraph; high cost product Amazon
3 – Bot profiles are intelligently created, often mimicking normal user behavior
MONETISATION
1 – Bots are then directed to ‘front’ websites which have been created specifically to allow the laundering of impressions through them. Typically female lifestyle or sports sites that are full of content ripped from other sites or content written specifically to drive traffic through SEO. These sites are then flooded with bot traffic.
2 – Some bots will not be instructed to build cookie profile and are only used to generate fake traffic which is then sold to networks seeking to extend audience/reach.
3 – Impressions generated by bots with cookie profile are then sold through exchanges with the URL’s of the ‘front’ sites often masked or spoofed to avoid detection.