Turns out that the black box of programmatic advertising is not the only place where IMP’s are, um, not counted correctly.
Measurement is the bedrock of this entire industry and has always employed a bit of fuzzy math. Starting around 1954, Nielson was able to automatically measure TV viewership levels in the top 30 markets and by 1973 was able to report them on an overnight basis. But most other TV markets relied on the Diary method. As such, Nielson would ask people to take note (literally with pen and pad) of the programming they watched during the ‘sweeps’; designated time frames about a month long, 4 times a year. Gaps in memory, environmental factors and other stimuli made this methodology suspect at best, but nonetheless was a standard which everyone agreed upon.
To bore you with the details of how this has evolved through the years would be blog suicide. We now have Code Readers, Local People Meters, Portable People Meters, Return Path Data and Set top Meters. The markets by measurement type as well as the release dates for the data collected vary substantially. Toss in Broadband Only homes (“BBO”) which are now folded into the Nielsen measurement set to count vMPVD subs that carry local programming or streamers that have access to local signals and locating TV lovers starts getting a bit arduous.
Finding eyeballs is only half the equation. Verifying that an advertisement was actually placed where it was supposed to be is just as crucial. In the “set it, and forget it” world of programmatic display advertising, you can pay $10,000 at a $10 CPM and “reach” 1,000,000 people with the press of a button. But A) how do I know my ad was targeted correctly and B) how do I know it got delivered? The answer comes more or less down to, “trust me, it was and it did”. Fake sites, Fake apps, ad stacking, headless browsers, sourced traffic, bots, viewability and an entire rabbit hole’s worth of question marks make for the extreme likelihood that a good amount of inflation is going on in online campaign reporting. In 2018, Jupiter Research estimated global online ad fraud at $19 billion. You read that right… NINETEEN.
Television may have the exact opposite problem. MRC found that total usage of TV in February by people ages 18-49 — the audience most desired by advertisers in traditional TV entertainment — may have been understated by 2% to 6%, Variety reported. In a statement, Nielsen acknowledged the discrepancies. “As a result of some of the COVID measures we implemented, we found that there was some understatement of audience estimates,” the company said. The Video Advertising Bureau estimates that a 1% undercounting of audiences may represent $39 million in ad dollars, while a 6% undercounting could represent $234 million… for the month of February alone. Extrapolated over 12 months, the organization projected those figures could mean as much as $468 million to $2.8 billion!
So to recapture and summarize, advertisers may be paying for digital audiences that aren’t there and not paying for TV audiences that are.
