Fou's Notes: Good Ideas vs In Real Life

As opposed to Cliff's Notes (which are awesome), here are my notes in bullet point format, since everything else is TL;DR (too long; didn't read). I'll say why something is a good idea, but then why it doesn't work as expected in real life (IRL).

TAG 100 million IP address block list

• good idea - to block IP addresses of known bots
• IRL - bad guys swap out blocked IP addresses within minutes and use new ones to make money again; to IRL, TAG's 100 million IP address list is completely useless in blocking any fraud.

IAB bots and spiders list

• good idea - to block bots from being served ads
• IRL - this only blocks the honest bots that say their name in the user agent (e.g. googlebot, bingbot, facebookbot, etc.). It does not block any bot that lies about their name by pretending to be Chrome, Safari, Firefox, etc.

MRC accreditation

• good idea - to have a third party validate what you cannot validate yourself with respect to a black box fraud detection company (because they won't tell you or show you any data)
• IRL - MRC can only validate that the vendor measures what they said they would measure, NOT whether they are measuring accurately. And if the vendor showed you the details of why they marked something fraudulent, or not fraudulent, so you can understand why, then 3rd party accreditation would not be necessary in the first place. read more: https://www.dhirubhai.net/pulse/why-i-dont-offer-certification-accreditation-ad-fraud-investigator/

Ads.txt

• good idea - to have publishers say which exchanges are authorized to sell their inventory, so buyers can check
• IRL - buyers don't check. Fake sites simply put plagiarized ads.txt files on their sites and game it in many other ways. read more: https://www.dhirubhai.net/pulse/adstxt-zero-day-exploit-wild-brief-history-fraud-ad-fraud-historian?trk=portfolio_article-card_title

Fraud detection

• good idea - we should detect fraud so we can avoid paying for it in campaigns
• IRL - bad guys have A-B tested their bots and those bots can get by every fraud detection technology already. They even sell specific traffic like "IAS-compliant" traffic or "DV-compliant" traffic -- i.e. gets by undetected. Fraud detection tech can be blocked by bots or easily tricked. read more: 1) https://www.slideshare.net/augustinefou/why-fraud-detection-doesnt-work and 2) https://www.dhirubhai.net/pulse/assuming-fraud-detection-works-its-still-useless-?

Viewability detection

• good idea - we only want to pay for ads that are viewable
• IRL - the only sites that have 100% viewable ads 100% of the time are fraudulent ones; good publishers have to array the ads on the page, some above the fold and others below the fold. Buying only viewable ads means good publishers lost a third to a half of their inventory; but bad guys who trick the viewability measurement can sell everything.

Brand safety detection

• good idea - we don't want our ads on bad sites and next to unsavory content
• IRL - "safe" is subjective and detection tech is not very good or doesn't work at all for very simple reasons. "For example, “shooting” is one of the most common blacklist terms. While it may identify some content about violence, that term will also block content by astronomy buffs (shooting stars), sports fans (shooting hoops), technology users (troubleshooting), photographers (shooting a photo)." (Source:?AdAge)?Read more: https://www.dhirubhai.net/pulse/brand-safety-gone-mad-bad-just-plain-wrong-ad-fraud-historian/

Industry wide estimates

• good idea - to give the industry a sense of the size of the ad fraud problem
• IRL - all industry-wide estimates are extrapolated from a data set that is only a slice of the universe. Further, by only focusing on IVT (bots) the estimates are missing many other forms of fraud or hidden costs. So the numbers are always under-estimating the problem and creating a false sense of security. read more: https://www.slideshare.net/augustinefou/hidden-costs-in-digital-media-supply-path

Prebid filters / Real time blocking

• good idea - block the fraud before we even bid on it
• IRL - the bid requests are all declared variables so bad guys will always declare something that is desirable and in-demand (e.g. device is an iPhone, to earn higher CPMs). You can block the most obvious kinds of fraud but in most cases you don't have enough information or detected data points to know for sure it is fraud or not; so the default action is to let it through.

Whitelists

• good idea - buy only from good publisher domains
• IRL - bad guys pretend to be those good domains so they can get bids; if your fraud detection vendor or your DSP or your agency doesn't provide you with placement reports that show both the domain and the seller ID associated with that domain then you can't tell whether there was any fake inventory mixed with the real. Be sure to use analytics to measure your ads yourself, so you can see where they actually went.

About Me:??“I consult for advertisers and publishers who actually want to know the truth and who have the courage to do something when they find ad fraud. I am not a fraud detection tech company that relies on fraud to continue. I show my clients the supporting data so they can understand and verify for themselves what is fraud and what is not fraud. If they agree, they can take the necessary actions to eliminate the fraud while campaigns are still running, rather than post-mortem fraud reports and trying to get their money back.”?

Follow me here on?LinkedIn (click)?and on?Twitter @acfou?(click)

Further reading:?https://www.slideshare.net/augustinefou/presentations