Watermarking test shows path to unified TV ad counting, but adoption hurdles remain

The television advertising industry has long operated with an acknowledged but unresolved problem: not all ads that run can be verified as having run.

Fragmented distribution paths, inconsistent tagging and siloed measurement datasets have left buyers and sellers without a common way to count impressions across linear and streaming television.

A new report from the Coalition for Innovative Media Measurement details a proof of concept test designed to determine whether open-standard watermarking can serve as the technical foundation for a more unified counting system.

The report, “Counting What Counts: The Future of TV Advertising Measurement,” was published in March 2026, developed by Deloitte and documenting a test led by CTV1 Media Services.

The core idea is straightforward. Embed a common identifier, called a Universal Ad ID, into an ad creative once, then detect it wherever the ad plays. The standard used to accomplish this is ATSC A/334 audio watermarking, paired with the IAB Tech Lab’s Ad Creative ID Framework, known as ACIF. Together, they form the backbone of what CTV1 calls the TV Registered Ad Counting framework, or TRAC.

The distinction between open and proprietary matters here.

Most current watermarking implementations rely on closed systems controlled by individual measurement vendors or device manufacturers. Those systems can function within a given ecosystem but tend to reinforce the fragmentation the industry is trying to solve.

“Combining two open standards, ATSC A/334 watermarking and IAB ACIF, is exactly the kind of progress the industry needs as it pursues reliable cross-platform ad delivery data,” said Madeleine Noland, president of ATSC.

Advertisement

What the test measured

CTV1 distributed watermarked ads across 17 TV stations in seven U.S. designated market areas between September and December 2025, spanning Portland, Salt Lake City, Denver, Nashville, Phoenix, Washington D.C. and Baltimore. The test used a dual reader approach: watermark detectors embedded in smart TV firmware for deterministic, device-level counts, and cloud-based readers placed in broadcast monitoring infrastructure, with detections correlated against automatic content recognition, or ACR, data to estimate household exposure.

The test confirmed watermarks were detected and counted across both pathways.

Station-reported airings were independently verified against watermark detections. The POC also demonstrated that ad exposure data could be reported at the ZIP code level rather than the designated market area level. Ad-count signals were aggregated across multiple TV original equipment manufacturers, or OEMs, into a single unified view of delivery.

The local market gap

The gap between watermark detections and ACR fingerprinting results, however, produced one of the test’s more telling data points.

One hypothesis tested was that ACR-based datasets systematically miss local small-business advertising. Local ad creatives turn over frequently, come from a large number of small buyers and are rarely submitted to the fingerprint reference libraries ACR systems rely on to identify ads. National campaigns fare considerably better.

The data supported that hypothesis.

Across local stations in the test markets, roughly 85 to 90 percent of watermark-detected ad occurrences had no corresponding ACR ad match, even when those same stations showed high channel tune-in match rates. Watermarking sidesteps this gap because detection does not depend on a pre-populated reference library, an ad is identified every time it plays, regardless of whether it was ever submitted for fingerprinting.

What stakeholders said

Deloitte conducted interviews with executives across broadcasters, OEMs, advertisers, measurement companies and standards bodies. Those conversations surfaced consistent agreement on the potential value of the approach and equally consistent concern about what operational adoption would require.

“Watermarking has the potential to bring more transparency, accountability, and unified measurement across digital and linear platforms,” said Megan Halscheid, vice president of global data intelligence at Publicis Media.

But interviewees drew a clear line between technical feasibility and operational reality.

The most frequently cited obstacles were not engineering problems: they were organizational and commercial: friction around Universal Ad ID adoption, the absence of a neutral industry steward, concerns about incremental costs and the challenge of aligning incentives across a supply chain with competing interests.

“True industry progress depends on aligning these incentives so that advertisers drive measurement innovation and adoption,” said Michael Vinson, chief research officer at Comscore.

Advertisement

Business model and what comes next

CTV1 outlined a revenue-sharing model in which costs are distributed across advertisers, registries, watermark vendors, broadcasters and OEMs based on usage. A nominal CPM-based fee would grow as a larger share of TV ad impressions becomes measurable via watermarking.

The report acknowledges this model has not yet been validated in practice and that a follow-on pilot is planned to test its financial assumptions.

CTV1 stated a goal of expanding watermark ad counting to more than 100 million TV screens by mid-2026, contingent on broader OEM and broadcaster participation. The report frames the next phase as dependent on active participation from multiple industry sectors, not endorsement alone, but hands-on testing in real operating conditions.

The framework is not designed to replace existing measurement systems.

Panel-based reporting and fingerprinting are expected to coexist with watermarking for several years. The intended role of TRAC is to feed deterministic ad-count data into existing workflows, providing a verifiable signal that buyers and sellers can reconcile independently of any single measurement provider.

Whether the industry moves from a validated proof of concept to coordinated adoption will depend less on the technology and more on the commercial and organizational alignment the test was not designed to measure.