You bought some excel spreadsheets for millions of dollars
They keep sending you excel spreadsheets that tell you how much you spent last month, how many impressions you bought, and how many clicks you got. They tell you that your campaigns got great “performance” or “engagement” — which means they got a lot of clicks. They tell you that “spend efficiency” was up — which means they got you more impressions for lower average CPM (cost per thousand, unit prices). They never talk about your total spend and never try to help you save money on overall spend because they want you to spend it all with them. They only tell you about unit prices – CPMs. But if you buy ten times the quantity at one-tenth the unit price, you are still spending the same total amount. That’s the flawed logic of reporting on CPM prices. It leads marketers to buy ads as if they were shopping at Costco — larger quantities at lower unit prices, even if you don’t need all that toilet paper.
Reporting on unit prices and number of clicks on ads have predisposed marketers for the last decade to shift more and more budgets to programmatic channels — thinking they are getting “better deals” and more “engagement.” But the only parties that benefitted from that shift were the ad tech vendors and media agencies, certainly not the marketers themselves. The media agencies got a decade-long wave of shiny objects to sell to marketers. The ad tech vendors made money hand-over-fist for themselves and their investors, even fulfilling the hockey-stick revenue projections they made up for their powerpoints, despite the fact that the human population and their media consumption habits remained roughly at a plateau for the same decade. The only way ad tech could sell you continuously more ad impressions at lower unit prices was through ad fraud — using bots and other means to generate more and more impressions at lower and lower costs. It was super helpful that black-box fraud detection tech couldn’t detect anything “invalid” about this traffic; that kept advertisers in a spending mood.
Humans are not reliable; bots are super-reliable for ad revenue generation
Good publishers, who don’t buy traffic and do other shenanigans on their sites, struggled to survive. They had to wait for real humans to come to their sites, so that webpages would load and ad inventory would be created. Sometimes not enough humans showed up on their sites and they missed their revenue projections for the month, the quarter, and for the last 10 years. It’s just not reliable enough to wait for humans to show up on sites and use mobile apps for enough hours to generate enough ad impressions to survive. Quibi shut down, after enormous ambitions and enormous investments . T-Mobile shut down its streaming service after only a few months . There simply aren’t enough humans to watch all those streaming services.
What is reliable is “sourcing traffic” — a euphemism for ad fraud. Bot traffic is very reliable for growing ad revenues. It’s a simple economic equation — for every dollar you spend buying traffic, you can get ten dollars of ad revenue out. It’s so reliable why would anyone not put in the next dollar and get ten more dollars? Even if the fraud detection companies couldn’t detect anything wrong or “invalid” about this traffic, common sense would tell you there aren’t a whole bunch of humans sitting around with nothing to do but to go to specific websites when you tell them to. You try forcing a bunch of humans to go to your site; you will quickly find out that’s not possible. What is possible, and downright easy is a single line of code that instructs a vast botnet to visit a collection of sites X million times, or to pretend to be “Roku sticks” to generate millions of fake CTV ad impressions for sale.
Real humans’ streaming versus bots pretending to be Roku sticks
Humans are streaming more; but bots are streaming even more than that. Some of these scams get caught, like the 7 below; when the details are outed, advertisers pretend to be surprised at their scale: “StreamScam generated traffic appearing to come from nearly 30 million U.S. households, spoofing 3,600 CTV apps, and impersonating 3,400 different device types” and “ParrotTerra generated fake CTV inventory across a large number of apps, IPs and devices, spoofing 3.7 million devices and 2.7 million IP addresses per day.”
Data from TVision Insights, a “television attention measurement” specialist, shows that daytime TV viewing did nearly double (+77%) in 2020 compared to pre-Covid levels (because people and their kids are all at home during the daytime now). But most TV viewing by humans still remain in the evening hours. Furthermore, in the second chart below, the vast majority of the time spent streaming was on subscription services like Netflix (no ads), Amazon Prime Video, Hulu, and Disney+ and on YouTube and YouTubeTV. Note how small a percentage of human viewing time is spent on ad supported CTV streaming channels. Then what are all those CTV ad impressions you are buying when not buying directly from media sources?
Ask them to prove it’s real; they won’t be able to
If you’re a marketer who actually cares about the real performance of your digital campaigns, you should ask your ad tech vendors and media agencies to prove to you that what you are buying is NOT fraud. That’s right. We’re not asking what percentage IS fraud. We’re asking what percentage IS NOT fraud and for them to prove it to you.
We already know that fraud detection tech companies are not good at catching most things, because they are not even looking for most things — other forms of fraud that are not IVT (“invalid traffic”). Ask your agency to prove the impressions were actually served, when they were served (hourly reports), what sites/apps they were shown on (detailed placement reports), and were actually shown to humans (percentage human, not just not invalid). They won’t be able to do so. They will simply argue that the ad tech vendors they use are all accredited, the fraud reports don’t show much fraud, and everything must be real because you’re getting “engagement” — i.e. clicks — on the ads. But you already knew that bots click on ads, so having clicks doesn’t prove it’s real, and none of their typical explanations or excuses hold water, if you actually think about it.
Take matters into your own hands
The best thing you can do is to run “turn off” experiments yourself to prove whether the digital ads were real or useful in producing business outcomes for you. P&G turned off $200 million in digital spend and there was no change in business outcomes. Chase reduced showing their ads on 400,000 sites to just 5,000 sites (a 99% reduction) and saw no change in business outcomes. Uber turned off 80% of its paid mobile app install ad spending, and saw app installs continue unabated. Ebay turned off paid search ads and saw no change in traffic or sales on its site. And Airbnb cut $800 million at the start of the pandemic in 2020 and saw “traffic levels back to 95% of the traffic levels of 2019 without any marketing spend” .
Have your media agencies and ad tech vendors prove to you the ad impressions you are buying are real — ads shown to humans. If you don’t, then it’s been April Fool’s on you for the last decade.