This week's billion-dollar discovery of waste in programmatic media buying

This week's latest adtech scandal -- or "big reveal" as it were -- was from Cognitiv, which found another $6.6 billion lost to "bid shading". Don't know what bid shading is or how to describe it if someone asked you? Don't worry. Most of the marketers they talked to didn't know either. Years ago bid shading was pitched as something that can help advertisers "gauge CPM prices" and optimize their own bids, so they don't overpay. But no one really knew how it worked and never had any way to actually check. So it became yet another "black box" or "secret algorithm" thing that advertisers paid for, thinking it gave them an advantage. Like most other things in ad tech that started as a good idea, bid shading was exploited by baddies to increase their own profit. Most of this came in the form of hidden arbitrage, where some middleman did the bid shading to maximize their own profits, and only passed along some of the savings to the client, to make it appear bid shading was working.

Other more malicious behavior took the form of "bid flooding" where bad actors or the publishers themselves submitted many simultaneous bids to artificially inflate the CPM prices they got (slide above was from 2019). When done across billions of bids, this technique increased their own profits.

More recently in March 2022, we again documented DELIBERATE duplicated bid requests -- for example 30 to 80 simultaneous duplicate bid requests to game the system to increase profits for fraudulent sites/publishers.

Advertisers that don't turn off the multiple dozens of exchanges in a media buy are essentially bidding against themselves and costing themselves more money, to the sheer joy of all the ad tech middlemen who take cuts of the higher CPM bids. How many of you advertisers knew you were bidding against yourself, or have a way to check if you are or are not?

OK, OK, enough of me telling you I told you so for years. Let me summarize by saying "You have to know how ad fraud is perpetrated in order to know how to catch it." I would challenge any and all of the engineers working on the detection code and algorithms of legacy fraud verification vendors on their knowledge of this issue and about a hundred others, in order to properly detect these types of fraud. Some forms of fraud cannot be detected using javascript tags. Did they know that?

Let me finish with two examples from a few hours ago, today. I posted that I was "So proud of my client. They turned off programmatic display and video ads, fired the agency that was wasting their money, and never looked back." I included the chart below and asked if anyone could tell when we turned off all programmatic display and video ads. (Can you tell)?

Many wrote to me to ask where the "drop" was. They were expecting to see a lot of dark red, and then a reduction in the dark red when bad sites and apps were added to the block list. I had to post another example where we could see such a drop. You can clearly see the sites and apps were added to the block list on April 26. The dark red dropped in half from 48% to 23%. The latter case was measured in-ad, where we can see which bad sites and apps your ads went to. So we can determine which to add to the block list. The above example was measured on-site, because the client was a b2b advertiser and favored Google paid search ads (which we cannot measure). By putting the FouAnalytics on-site tag on the site, we can see the quality of the clicks arriving from various paid channels. In this example, the display ads worked so poorly they got hardly any clicks (it was less of a fraud or not issue). So turning off all programmatic display and video ads led to cost savings and no visible difference in traffic to the site.

Note also that the green volume bars in the in-ad chart dropped significantly. The large amounts of fraudulent impression volume dropped significantly (that is what is supposed to happen). Experienced practitioners understand that by cutting out bad sites and apps, CPM prices will go up (on average). But because you are buying far smaller quantities, even at higher CPM prices, you will still save money. Your costs will go down -- higher CPM prices multiplied by far smaller quantities = lower overall costs. Don't let your agencies trick you by saying "lower CPMs" means more "cost efficiency." CPMs are prices, not costs.

Steady state

Finally, we are at a stage of maturity with experienced practitioners using FouAnalytics for large advertiser clients, I was asked today, "why would we still need analytics if the client is buying well and we are at a steady state of good media buying?" My reply was that you need analytics to continue to monitor that bad guys don't creep back into your campaigns. Remember the breitbart example? You added the domain to a block list, but FouAnalytics in-ad tags still showed ads going to breitbart.com. You troubleshoot and additionally block the sellerIDs of breitbart.com. But FouAnalytics data still shows your ads going to breitbart. None of the placement reports show the domain -- breitbart.com -- or their sellerID anywhere. Turns out, breitbart borrowed the sellerIDs of cooperative partners to get around the sellerID blocks. The cooperative partners take payment, take a small "handling fee" and pass the net around under the table back to breitbart. Ok, you are more advanced now, and you start using an allow-list of domains. Breibart still shows up in your FouAnalytics in-ad measurement. How is that possible? There's actually multiple ways -- breitbart lies about the domain passed in the bid request and the faked domain is one that is in your inclusion list, or a publisher you are buying from is doing audience extension without telling you, and your ads are ending up on breitbart. Anyway. I hope this story illustrates there are SO many different ways things can go wrong buying through programmatic platforms that no amount of tech can protect you 100%. What's more important and useful is the experience of practitioners, and the fact that those practitioners have better tools to use to monitor and control the campaigns for you.

Referrals

If you are looking for alternatives to your current agencies or vendors, ask me. I am happy to refer better agencies, better media buyers, and better analytics practitioners to help you. I know they will be happy to show you all the details so you can understand what they are doing for you and why they are doing it. After all, "there's nothing more transparent than if you can "see Fou yourself."

I am very happy to acknowledge experienced practitioners -- Michael M. M. Thomas Koch Keri Thomas Cassie Stox .Marilois Snowman. Jim Gordon Dominic T. Michael McCarver Michelle Strom Camelot Strategic Marketing & Media

I am only recognizing the most vocal and public ones here, so others can continue their work behind the scenes.