The out-of-home media industry has been selling the same product for three decades: estimated impressions. The number of people likely to have passed a panel, multiplied by a visibility probability derived from panel size, orientation, and traffic speed. Cost-per-thousand calculated on an estimate of an estimate.
Digital advertising moved beyond this model fifteen years ago. Online impressions are verified — the ad loaded, the browser rendered it, the session lasted long enough for it to be seen. OOH has not made the equivalent move, not because the technology does not exist, but because the industry built its commercial infrastructure around estimates and has not had a strong enough push to change.
What vision-verified attention actually measures
Canopy's OOH attention measurement uses the same camera infrastructure that media owners install for security and operational purposes. The cameras observe the space in front of each panel. The inference model estimates, on a per-frame basis, how many people are oriented toward the panel and for how long.
The output is not a count of people who walked past. It is a count of verified attention-seconds per creative rotation — the number of person-seconds during which a human was demonstrably oriented toward a panel and at a distance and angle where the creative was visible. That is a meaningfully different metric from CPM.
No PII is involved. The model estimates head pose from body geometry — it does not use facial recognition, does not extract biometric data, and does not produce any output that could identify an individual. The output is a time-series of attention counts aggregated per panel per hour.
Why this changes the commercial conversation
Verified attention data changes three things in the OOH buying process.
First, it makes panel performance comparable. Estimated CPM varies by panel location, size, and traffic model, but those models are opaque and not independently verifiable. Attention-second data is directly comparable across any panel that carries Canopy measurement. A highway billboard and a mall corridor panel can be evaluated on the same axis.
Second, it enables real creative optimisation. When attention is measured per creative rotation on a digital panel, media owners and advertisers can see which creatives capture more attention in a given context. That feedback loop does not exist in an estimated-impression world. It creates genuine value for brand advertisers who want to know whether their creative is working, not just whether it was technically displayed.
Third, it creates a defensible pricing basis. Media owners who can show verified attention data can justify premium pricing for panels that outperform their estimated CPM. That is a revenue opportunity, not just a compliance cost.
The infrastructure is already there
The camera infrastructure to measure OOH attention already exists in most environments where panels are deployed. Malls have security cameras covering every corridor. Transport hubs have coverage at every concourse. The compute to run the inference model is modest — a compact edge appliance connects to the existing camera LAN and the data never leaves the site.
The barrier to adoption is not technology. It is the commercial inertia of an industry that has priced on estimates for thirty years and has not yet faced a buyer-side demand for verification. That demand is building. Digital-native advertisers who are accustomed to impression verification are increasingly buying OOH as part of integrated campaigns. They are asking for the same accountability standard. The OOH industry will need to meet it. The operators who build the measurement infrastructure now will be positioned to price on it first.