Nico Neumann from Melbourne Business School has conducted some of the most critical investigations into digital advertising’s core promises. His team compared third-party consumer data to verified ground-truth datasets. The findings are alarming. In the best-case scenario, the most premium data sources misidentified one in ten consumers. In lower-tier datasets, error rates reached 85 percent.
This means that for every 100 ad impressions delivered using third-party targeting, as few as 15 may actually reach the correct person. The rest are either miscategorized, mistimed, or misaligned. Imagine a GPS that gets you to the wrong location four out of five times but still charges you full fare.
This is not a theoretical flaw. It’s a foundational problem. MIT’s Sinan Aral has shown that once ads are exposed to clean A/B experimentation, most of their reported effects disappear. Garrett Johnson at Northwestern found in a meta-analysis of 432 campaigns that the median incremental sales lift from programmatic display ads was statistically zero. Zero.
The World Federation of Advertisers estimates that only 15 cents of every programmatic dollar actually reaches the consumer. The remaining 85 percent is absorbed by intermediaries, fraudulent activity, and technical leakage. In no other industry would this be acceptable.
And yet, the global digital ad market grew to 740 billion dollars in 2024, up 8.9 percent from the previous year. Meta and Google captured nearly half of that. This is not because advertisers believe it works. It’s because they cannot afford for it not to.
The modern internet economy is structured around advertising. Over 80 percent of Alphabet’s revenue and 97 percent of Meta’s comes from selling attention. News outlets, creators, platforms, influencers, streamers—all depend on monetized engagement. In this system, ads are not a channel. They are the infrastructure.
That makes the stakes much higher. Just as the 2008 housing collapse was driven by a widespread mispricing of mortgage risk, today’s online ecosystem is fueled by a mispricing of attention.
We treat a click as intent
We treat time-on-page as value
We treat engagement as causality
But these are synthetic signals. They look predictive in dashboards but fail to map to actual business outcomes. They are the digital equivalent of subprime mortgage tranches—bundled, labeled, and sold as premium while quietly rotting underneath.
Neumann’s field tests reveal that many digital ads don’t just fail. They may actively harm performance. In one controlled experiment, showing auto insurance ads to likely buyers reduced conversions. The presence of the ad introduced friction rather than persuasion.
He proposes a return to mass marketing. Not because it is nostalgic. But because it is resilient. Without relying on flawed personal data, mass campaigns reach broadly, ride on repetition, and are priced against real outcomes like sales lift and brand recall. They assume less and prove more.
Imagine two campaigns. One uses hyper-targeted data from a vendor with 70 percent profile error. The other uses national TV reach at fixed frequency with zero targeting. Which is more likely to deliver business impact? The answer is increasingly clear.
This is not a call to abandon digital. But it is a call to audit its assumptions. A system that absorbs hundreds of billions annually must be grounded in causal impact, not theatrical dashboards.
We are not in an age of precision. We are in a subprime attention crisis.