Programmatic has spent the better part of fifteen years building an industry to measure itself, and the headline finding has not meaningfully moved.
ISBA and PwC published their first supply chain transparency study in 2020. For every $1 of advertiser spend, 51 cents reached the publisher and 15 cents was simply unattributable. The 2022 follow-up was kinder: 65 cents to publishers, 3 cents unaccounted for, after two years of cross-industry effort to develop a toolkit specifically designed to make the audit possible. The ANA's 2023 study in the US looked at the open web and found that only 36 cents of every dollar entering a DSP reached a consumer at all, with annual waste estimated at $22 billion.
Three studies. Three methodologies. Three geographies. Same shape of finding. The money does not arrive where the advertiser thought it was buying, and a meaningful proportion of it cannot be properly accounted for after the fact.
The industry's response to this has been, almost without exception, to commission more measurement. More verification vendors. Bigger dashboards. Log-level data initiatives. SPO frameworks. Curation layers. And the shortfall persists, because almost none of what the industry calls measurement actually qualifies as verification.
The Self-Graded Essay Problem
A useful way to think about programmatic measurement is this. Every participant in the transaction also writes the report on the transaction.
The DSP reports on the bids it made. The SSP reports on the impressions it sold. The verification vendor reports on viewability and IVT, while being paid by one of the sides it is reporting on. The agency reports on its own performance against KPIs the agency helped to set. The publisher reports the inventory it served. None of these reports is independent of the party producing it.
That is the self-graded essay problem. You can ask a student to mark their own coursework, then again at GCSE, then again at A-Level, and produce a tidy upward trend. The grades are not necessarily wrong. They are structurally compromised, because the entity producing them has a position in what they say.
ISBA's first study made the point clearly without using the phrase. Of 267 million impressions in scope, only 31 million could be matched across the supply chain. The 88 per cent unmatched rate was not a failure of analytical technique. It was a consequence of data not being formatted, permissioned, or shared in a way that allowed an independent party to reconstruct what happened. Pulling half a story together required a year of cooperation between PwC, advertisers, publishers and tech vendors. Even after the cross-industry toolkit launched, PwC noted that audits should be achievable within five months. The 2022 study still took nine.
That is the real gap. Not between transparency and opacity. Between self-reported data and verified data.
Why Each Fix Hits the Same Wall
The industry has tried four broad responses, and each one runs into the same structural issue.
The first was viewability verification. Useful, but the verification vendor sits inside the commercial relationship it is measuring. It is contracted by the buyer, paid through media spend, and dependent on the continued participation of the platforms it grades. Whatever its intent, it is not an independent third party in any meaningful sense of that phrase.
The second was SPO. Supply-path optimisation was supposed to be verification by procurement. Cut the SSPs, route through fewer paths, get cleaner data. In practice SPO turned into a commercial negotiation. Advertisers used it as leverage on rates. SSPs used it as a marketing position. The exercise became a story about cost rather than a story about verification, and the underlying data problem stayed exactly where it was.
The third was log-level data. The ANA report was explicit about this: log-level data is the closest thing the industry has to a primary source. And yet 69 per cent of the companies that volunteered for the ANA study dropped out because they could not actually obtain the log-level data they had committed to providing. The data exists. Getting it into a form that allows independent reconstruction is, in practice, a months-long political project rather than a technical one.
The fourth has been curation and inclusion lists. Decide upfront which inventory is acceptable, buy only that, and treat verification as implied by the curated nature of the supply. This works better than open-auction optimism, but it solves an inventory quality problem rather than a verification problem. You can curate a perfectly respectable list and still have no independent record of what was bought, at what price, from whom, with what take-rate at each hop.
Measurement Is What You Can See. Verification Is What You Can Prove.
The distinction is not pedantic.
A measurement is a description of an event by a party with a position in that event. A verification is a description by a party with no position in it. Programmatic has spent fifteen years asking the first thing to do the work of the second, and the studies that took the trouble to test the difference have all reported back the same result.
If your post-campaign report is produced by a vendor that profits from the campaign continuing, you do not have a report. You have a sales pitch with charts.
The verification gap is also why every transparency conversation eventually loops back on itself. The advertiser asks for clarity on take-rates. The DSP and SSP respond with self-reported numbers. The agency layers in its own reconciliation. A verification vendor adds another dataset, also self-reported by definition, because the vendor has a commercial relationship with one or more parties in the chain. The advertiser ends up with a great deal of data and no answer to the actual question, which is the only one that matters: what was bought, from whom, at what price, and where did the difference between the cheque written and the cheque cashed actually go.
The Layer That Has To Exist
The structural fix is not more measurement. It is verification that does not sit inside the transaction it is verifying.
A layer that records what happened. That none of the parties in the trade can edit, manipulate or quietly reformat. That produces the same record for everyone looking at it, on the same timetable, in the same units.
This is not really a technology argument. It is a market-structure argument. Every industry that runs on trust between buyers and sellers eventually develops an independent verification layer, because the alternative is an arms race of self-reporting that nobody wins. Financial markets have settlement and reporting infrastructure that does not belong to either side of the trade. Public companies have audited accounts produced under rules the company itself cannot rewrite. Insurance markets have loss adjusters. Commodity markets have inspection regimes. Programmatic has none of this. It has reports.
Whether the verification layer in programmatic arrives as a regulatory requirement, an industry consortium with actual enforcement powers, a standards body with teeth, or a piece of infrastructure that simply makes self-grading impossible, is a separate conversation worth having. But the diagnostic phase of this should be done by now. Fifteen years of studies have established, repeatedly and from independent angles, that the industry cannot measure its way out of its measurement problem.
The next part of the story is not what else programmatic can measure.
It is what it agrees to verify, and who gets to do the verifying.
---
Related reading from Alkimi Research
What Your Clients Are Actually Asking When They Ask About Agentic. The agency-side complement to the brand-readiness conversation above.
Is your Brand ready for Agentic Advertising? The brand-side complement to the agency reposition above.