How DOOH Audience Measurement Works

DOOH audience measurement is the set of processes that quantify how many people pass or look at a digital out-of-home screen, and what those people are like. This guide explains the three layers of measurement, the difference between a modeled and a measured audience, which metrics are actually captured, and how the resulting data feeds programmatic trading.

Why is DOOH audience measurement a hard problem?

In most of digital advertising, an impression is relatively clean: a browser loads an ad and a counter ticks up by one. Out of home, the screen is fixed, the audience is moving, and nobody sits down to click “I saw this.” A billboard or a transit screen may have thousands of people pass in front of it, but how many actually looked at the screen, for how long, and who they were is not directly visible.

For that reason, the DOOH audience was estimated for a long time through indirect methods. The industry’s recent direction has been to make that estimate both more precise and more compatible with programmatic trading. The clearest way to understand it is to break measurement into three layers.

Three layers: measurement, currency, and activation

The journey of DOOH audience data typically runs through three successive layers.

1. The measurement layer produces the raw ground truth: anonymous people counts, impressions, dwell time, attention, and — under suitable conditions — coarse age bands and a gender split. This layer tries to answer who was in front of the screen, when, and with how much attention.

2. The currency layer turns measurement into trade. Raw counts cannot be bought and sold on their own; they need to be converted into a unit the industry agrees on, a standardized impression value. That conversion ties measurement to an IAB- and OpenRTB-compatible impression multiplier, which is what unlocks programmatic buying and selling.

3. The activation layer turns data into a real-time decision. Based on the detected audience, the system decides which creative plays on the screen (dynamic content) or what price that play sells for (dynamic pricing). Here, measurement stops being a backward-looking report and becomes a live decision engine.

These three layers form a chain: if measurement is weak, the currency is untrustworthy; without currency, activation is blind. The rest of this article focuses mainly on the first layer — measurement itself — because the soundness of the other two flows from it.

Modeled audience or measured audience?

The most important distinction in DOOH measurement is whether the audience is modeled or directly measured. The two answer different questions and carry different levels of confidence.

A modeled audience infers “who passed this location” from indirect data. The classic method combines mobile-device location data with field panels to statistically estimate the flow of people within a given screen’s catchment. Most established providers — Geopath in the US, Nielsen, and Route in the UK — rely on this modeling approach and generally do not place a camera at the screen. Modeling scales broadly and is low-risk for privacy, but it cannot close the gap between “who passed” and “who looked.”

A measured audience collects data from the screen’s own vantage point through real observation. A camera placed at the screen, paired with computer vision software, tries to count how many people are looking toward the screen at that moment and for how long. IAB’s DOOH measurement guidance explicitly anticipates AI camera- and sensor-based audience detection. We go deeper into the difference between modeling and measuring, and the comparison of mobile-location and camera data, in a separate article.

Dimension	Modeled audience	Measured audience
Data source	Mobile location + panel	Screen-side camera + computer vision
Question answered	Who passed by?	Who looked, and for how long?
Typical provider	Geopath, Nielsen, Route	On-screen CV measurement providers
Privacy risk	Low	Depends on design (low if anonymous)
Scale	Broad	Per-screen, requires hardware

What exactly gets measured in DOOH?

The output of the measurement layer is not a single number but several complementary metrics.

Impressions are the base currency: they describe how many people a play reached. In DOOH, a single play can correspond to many impressions; we cover that multiplier logic in the article on impressions.

Opportunity and likelihood to see define the quality of an impression. Here the industry uses OTS (Opportunity-to-See) and LTS (Likelihood-to-See) metrics; IAB’s 2025 guidance reflects a shift from a rough “opportunity to see” toward LTS, which better captures the actual likelihood of viewing.

Dwell time measures how long a person stays within the screen’s catchment and is the key input for estimating attention.

Attention is the qualitative edge of measurement, and honesty matters here: true eye-gaze cannot be reliably measured at distances of 2-10 metres. Instead, the industry measures whether the head is oriented toward the screen (head-pose) and combines that with duration. In other words, it is a “facing the screen, for this long” approach — not a precise claim that “the pupil is on this pixel.” We unpack that nuance in the attention measurement article.

Demographics are the most contested and most limited metric. Camera-based age and gender estimation is possible, but the error margins are serious. Gender estimation can reach high accuracy in good conditions but degrades with angle, occlusion, and children, with documented bias issues. Age is the weakest leg: real-world mean error can run to roughly 7-11 years, so only a coarse age band is considered reliable. We honestly bound what demographic measurement can and cannot promise in a dedicated article.

Detection is not recognition

In camera-based measurement, the critical distinction is not to confuse face detection with face recognition. Detection only asks “is there a face in frame, is it oriented toward the screen.” Recognition tries to match an identity. Privacy-first measurement providers (such as Mecrai) can be designed to perform detection only, generate no biometric template, and output only anonymous, aggregate counts. That architectural choice directly determines the legal and ethical acceptability of the measurement.

How does measurement connect to currency and programmatic?

A raw metric does not unlock budget on its own; it has to be standardized. The main bodies shaping DOOH measurement standards are the IAB (technical framework and OpenRTB), the MRC (audit and accreditation), and currency providers such as Geopath; we detail their roles in the measurement standards article. In MRC’s OOH accreditation, Phase 1 covered non-audience metrics, while the Phase 2 rules covering audience metrics have only more recently begun to take shape.

The technical link is made through the dooh object and the imp.qty.multiplier field in OpenRTB 2.6: the measured audience is converted into the number of billable impressions per play. That multiplier can take a fractional value (for example 14.2 or 0.32). A DOOH play is thereby translated into the “impression” language the rest of the digital ecosystem understands and enters the real-time bidding (RTB) flow. The more reliable the measurement, the more defensible that multiplier becomes — which is precisely why currency is what unlocks budget.

Summary

DOOH audience measurement is a three-layer chain that converts raw anonymous counts first into a commercial currency and then into live activation. The strength of that chain comes from being able to measure the audience directly rather than only model it — and from honestly defining the limits of every metric.