How to Measure DOOH Advertising: Metrics, Attribution, and Reporting

Every DOOH measurement conversation starts by separating two numbers that look similar and mean very different things. A play is a delivery event: your creative aired on screen 4471 at 8:14 AM for ten seconds. It is a hard log record. An impression is a model: an estimate of how many people had a reasonable opportunity to see that play.

The impression model has three layers. The first is circulation, meaning how many vehicles or pedestrians pass the location in a given hour, drawn from transportation department counts, mobile location panels, and venue traffic data. The second is a visibility adjustment that accounts for screen size, viewing angle, distance from the audience path, obstruction, and travel speed. A bulletin set back from a highway at a sharp angle converts far less circulation into opportunity to see than one facing traffic head on. The third is an occupancy multiplier: a vehicle passing a roadside screen carries more than one potential viewer, and the model applies an average load factor per market and road type.

In the United States, Geopath is the audience measurement body that standardizes this methodology across publishers. When a plan quotes impressions, ask whether the numbers are Geopath-rated or publisher-supplied. Publisher-supplied numbers are not always wrong, but they are not comparable across networks unless they use the same methodology.

One practical consequence: never compare a DOOH CPM to a digital CPM without knowing the impression methodology behind each. A modeled opportunity to see is not the same unit as a rendered browser impression, and the DOOH definition is generally the more conservative of the two because it requires physical proximity.

Exposure measurement is what turns a modeled impression into an addressable audience you can study. A measurement partner draws a geometry around each screen, sized to the realistic viewing area rather than a lazy radius, and collects anonymized mobile location signals observed inside that geometry during the exact minutes your creative played. The result is an exposed device set.

Two qualifiers matter. First, the panel is a sample, not a census. Only a fraction of passersby produce a usable location signal, so the exposed set is scaled up statistically rather than counted directly. Second, geometry quality decides accuracy. A one mile radius around an urban panel captures thousands of people who never faced the screen, which inflates the exposed set and dilutes any measured lift. Ask any partner how they draw their viewsheds before you accept their numbers.

The exposed set is the input to everything downstream: visit studies, brand lift panels, and DOOH to mobile retargeting. It is also the reason precise geofencing around screens and around store locations is a measurement decision, not just a targeting decision.

Visit lift is the most requested outcome study in DOOH, and the most frequently misread. The method builds two groups: an exposed panel of devices seen near your screens during the flight, and a control panel of devices matched on geography, mobility patterns, and behavioral profile that were not exposed. Both groups are then observed for visits to your locations during and after the flight window.

The headline output is a lift percentage: exposed visit rate divided by control visit rate, minus one. A three to fifteen percent lift is a normal range for a well-targeted retail campaign. Anything reported above thirty percent deserves scrutiny of the control construction rather than celebration.

Read visit lift with three caveats. Panel size drives confidence, so short flights and small budgets often produce results that are directionally interesting and statistically weak. Attribution windows matter, because a grocery brand should measure days while an auto dealer should measure weeks. And proximity confounds everything: screens near your stores will always show higher raw visit rates, which is exactly why the control panel must be matched on location behavior.

Brand lift measures what happened in the audience's head rather than their feet. Survey panels are served to exposed and control device sets, asking about aided awareness, message association, favorability, and purchase intent. The difference between the two groups is the lift.

Awareness moves first and moves most, typically four to twelve points on a campaign with real weight. Consideration and intent move less and slower. Message association is the most useful diagnostic of the four because it tells you whether the creative communicated the intended idea at a glance, which is the specific thing out-of-home creative either does or fails to do.

Brand lift requires scale. Below roughly two million impressions in a market, the survey cell sizes get too small to separate signal from noise, and the study becomes an expensive way to produce an uncertain number.

Screens move people to phones. The exposed device set can be pushed into a mobile retargeting audience, which produces two things at once: a second touch in a channel with direct response tracking, and a measurable path from physical exposure to online action. Click-through and conversion rates on retargeted exposed audiences typically run well above prospecting benchmarks because the audience has already seen the brand in the physical world.

Beyond retargeting, three signals are worth instrumenting. Direct and organic site traffic in exposed markets, compared against holdout markets over the same period. Branded search query volume, which reliably lifts five to twenty percent in markets carrying meaningful out-of-home weight. And market-level conversion rate, which is where a media mix model eventually picks up the channel contribution.

We do not recommend driving that behavior with scannable codes on screens. The distance, dwell, and motion conditions of most out-of-home environments make scanning impractical, and a memorable brand name plus a simple message outperforms it consistently.

Proof of play is the delivery audit. Log-level records show screen ID, timestamp, creative ID, and duration for every airing. Reconciling those logs against the plan is how you confirm that delivery matched the venue mix and daypart schedule you bought, rather than concentrating in whatever inventory happened to be cheapest.

Photo-verified proof of play goes one step further by capturing dated photographs of the creative running on the physical screen. That is the only evidence that the file rendered at the right aspect ratio, that the screen was actually on, and that the environment looks the way the plan claimed. It is also the asset clients and stakeholders respond to most. Read our detail page on photo-verified proof of play for how the process works in the field.

Every credible measurement in this channel rests on a comparison, because there is no click to count. The question is always what would have happened without the campaign, and the answer requires a control.

Device-level control panels are the fastest to run and the most sensitive to matching quality. Market-level holdouts are stronger: hold out three matched DMAs, run everywhere else, and compare sales, visits, or search between the groups. Store-cluster holdouts are the strongest of all for retail, matching pairs of stores on baseline volume and trade area, exposing one of each pair, and reading point of sale data directly.

Holdouts feel like wasted budget. They are the cheapest insurance you can buy on a channel decision, because the alternative is renewing or cancelling a six-figure plan on a number nobody can defend.