Only R439 of Every R1,000 Reaches Consumers in Programmatic Advertising ANA Study Finds

Imagine handing a financial adviser R1,000 to invest on your behalf. He pockets R561 immediately for fees, platform costs, and what he calls ‘market infrastructure', then puts the remaining R439 to work. When you ask how the portfolio is performing, he produces a chart with a confident upward curve. However, your capital will take years to recoup what you put in. You would change advisers. Possibly before lunch.

This, more or less, is the current state of programmatic digital advertising.

According to the Association of National Advertisers' 2024 Programmatic Transparency Benchmark Study, only R439 of every R1,000 spent programmatically reaches a consumer as a quality impression. The remaining R561 is absorbed by a layered ecosystem of middlemen’s fees, low-quality inventory, made-for-advertising (MFA) websites, and fraudulent traffic. Meanwhile, the dashboard looks fine.

The ANA study, arguably the most credible and granular independent assessment of programmatic efficiency available, noted a 7.9 percentage-point improvement over the prior year (2023). Progress, certainly. But less than half of your budget reaching a real person is not a figure most CFOs would describe as satisfactory.

The viewability trap

For the better part of a decade, the advertising industry has relied on viewability as its primary quality signal. An ad that technically appeared on-screen for the specified minimum duration was deemed a productive impression. The logic was not entirely unreasonable as a starting position. The problem is that viewability was always a baseline, not an end goal. It confirmed that an ad was rendered. It said nothing about whether it rendered into an environment where any human could ever have the opportunity to notice it.

Research published by Lumen Research, whose eye-tracking data has become an important reference in this discussion, consistently finds that 70% of technically viewable ads are never seen at all. The human eye simply does not register them. Optimising for viewability is a little like measuring the success of a dinner party by confirming that food was placed on a table without finding out whether any guests would ever be at that table, let alone whether any ate it.

The implications become sharper still when you consider how viewability interacts with fraud. Sophisticated bots have become remarkably proficient at gaming viewability metrics: loading pixels into hidden frames, simulating scroll behaviour, and triggering every technical green light without a human being present anywhere near the screen. If your campaign optimisation is built primarily around viewability scores, you have, without quite intending to, drafted a fairly comprehensive manual on how to be defrauded.

From viewability to attentiveness

Viewability asks: Did the ad load? Attentiveness asks: Was a real human being present, and did they genuinely engage with what they saw?

At TruthsetsOnline.com, powered by FouAnalytics, we triangulate more than 300 variables to answer the second question with forensic confidence. These include live touch events, scroll patterns, mouse movements, and a constellation of behavioural signals that, taken together, constitute something no bot can convincingly replicate: the signature of an actual human being using a device. The methodology is deterministic rather than probabilistic. We do not sample and extrapolate. We examine what happened and report what the data show, including which specific sites, apps, and impressions drove waste and why.

We describe this as the ‘glass box’ approach. Legacy verification tools tend to operate as a ‘black box,’ producing a headline fraud figure without revealing the underlying data, such as which specific placements triggered the classification or why unmeasured inventory was quietly marked as clean. The glass box does the opposite: every finding is supported by granular, auditable data, and nothing is obscured.

The real cost of ‘potential reach’

There is a phrase that experienced media buyers encounter constantly: ‘potential reach.’ It sounds reassuring. It implies scale. What it actually describes is a pool of impressions for which no one has confirmed whether a human was in the vicinity.

At a time when marketing budgets are under sustained pressure, and every rand must justify its existence, potential is not an acceptable answer for a CFO. The ANA study makes clear that even after documented improvements, the majority of programmatic spend is still consumed before it reaches a person. Juniper Research projects that global losses to ad fraud will reach $172bn by 2028, suggesting that the structural incentives driving this problem are not fading of their own accord.

Brands that move from viewability-based measurement to human-verified attentiveness metrics typically recover between 20% and 60% of their media spend, redirecting it from bot-friendly inventory towards verified human audiences. For a brand allocating R5,000,000 annually to programmatic advertising, recovering 30% means R1,500,000 redirected to media that generate actual commercial outcomes. That is a material business recovery, available within a single campaign cycle.

Where to begin

The encouraging reality is that addressing this does not require dismantling your existing technology stack. TruthsetsOnline deploys a forensic analytics tag alongside your current verification tools, with no disruption to live campaigns. Within one cycle, you receive a side-by-side comparison of what your current vendor is reporting and what the deterministic data shows.

The findings are always surprising. Sometimes they are uncomfortable. They are also consistently useful because once you can see where your budget is going, you can make informed decisions about where it should go instead.

The industry has made genuine progress. The ANA data confirms it, and the people driving that improvement deserve credit. And yet, less than half of your budget reaching a human being remains, by any reasonable measure, an unfinished project.

Your R1,000 deserves considerably better than R439 worth of results.