Exploring pseudo-anonymized data performance woes and plotting a path ahead
There’ve been experiments where monkeys are just as likely to make a good or bad stock pick as the experts.
Could the same scenario be playing out when it comes to identifying a “targeted” audience for serving up digital ads?
The answer is an emphatic YES. That’s according to researcher Dr. Augustine Fou, who is one of the foremost experts in identifying fraud in digital ad campaigns. Dr. Fou has taught at New York University and Rutgers, with market experience dating back to “client side” work at institutions like American Express and “agency side” work at firms like Omnicom and McCann Worldgroup/MRM, where he served in chief digital officer and senior digital strategy roles.
Dr. Fou minces no words when he says a significant contingent of digital marketers has perpetrated a massive fraud.
This is not a new phenomenon. Fake traffic. Fake clicks. Even fake sales and metrics too. Plenty of light has been shone on some of the most unscrupulous tactics that have tarnished the digital ad industry.
But here’s the thing…
Even the “legit” stuff is now being called into question. Dr. Fou traces it back to the common practice of inferring the interests or characteristics of someone based on signals like browsing and purchase history. In other words, relying on cookies.
He gives the example of assuming someone browsing sportsillustrated.com and shopping for beard trimming kit probably being a man, and someone browsing victoriassecret.com and shopping for feminine hygiene products probably being a woman. It’s a common practice among many data providers.
But Dr. Fou’s data has shown time and again that this approach just doesn’t hold water:
…when academics studied the targeting parameters purchased from ad tech data brokers, they found that even a single parameter – gender – was only accurate 42% of the time (green highlight in the table below). To put that in perspective, that is LESS than the natural population gender split of 50%.
Things don’t get better when inferring two characteristics, like gender plus age. Accuracy dropped to an average of just 24% in these studies.
You may as well just have the monkeys pick an audience for you. To make matters worse, these audiences typically aren’t even permissioned.
Worthless doesn’t begin to describe it. Yet, marketers end up paying MORE for these CPMs based on layered targeting parameters.
Back to basics
At BRIDGE, we put a premium on the truth. On accuracy. On what’s right.
There are more layers to peel back here. We help digital marketers understand how to take back control and make sure they’re actually putting digital ad spend to good use.
This includes seeing how the numbers game changes when introducing permissioned, curated data based on actual people.
Stay tuned here for more on this soon.