Menu
Guide
How to measure digital out-of-home.
Insist on a control group. A held-out set of comparable people who never saw the ad is the difference between a result and a description of the month you ran in. Everything below is how to tell whether the rest of the number is real.
Four questions to ask any vendor
Including us. Three of the four are hard to answer well, which is the point of asking them early.
A bad one: A before-and-after comparison, or a benchmark against the category average. Both measure the calendar.
A bad one: A headline figure with nothing under it. If nobody can tell you what was divided by what, nothing was counted.
A bad one: Unique devices. A rotating identifier turns one passer-by into several, and the error always inflates.
A bad one: At the bid, or in a dashboard you have to log into and believe.
The same four, turned on us
Our screens carry cameras, so attention is read at the glass. That produces a number most of the category has no way to produce, and it is not a flattering one.
Across 10,626,956 exposures we measured, 85.20% of people were in the viewing area with their attention somewhere else. 12.5% were attentive or glancing. A further 2.30% had the display blocked from view, and those were never billed.
That is the real shape of out-of-home attention, and it is roughly the shape of everyone’s. The difference is that we counted it, so you can plan against the 12.5%.
Deduplicated people and frequency come back on every flight, against a control cohort held out for the full term. Every completed play is signed on the screen with Ed25519 and verified by you against a public key, with no account. How we measure.
We publish a flight that survives all four. Read the delivery report, or start with how to use out-of-home.
Questions buyers ask
- How do I make sure I am measuring digital out-of-home the right way?
- Insist on a control group. A held-out set of comparable people who were not exposed is what separates a result from a description of the month you ran in. Without one, a seasonal week, a new store opening and a great creative all look identical, and you cannot tell which one you paid for.
- What is wrong with standard out-of-home impression numbers?
- They are usually modelled. A traffic study estimates how many people pass a location, a share of them is assumed to have looked, and the result arrives as a confident figure with no way back to a person. It is not dishonest, it is just an estimate wearing the clothes of a measurement.
- How many people actually look at an out-of-home ad?
- Fewer than the impression count suggests. Across 10,626,956 exposures we measured at the screen, 85.20% of people were in the viewing area with their attention elsewhere and 12.5% were attentive or glancing. We publish that because a buyer who knows the real attention rate can plan against it, and because a vendor who will not tell you is not counting.
- Is a unique device the same as a unique person?
- No, and the gap runs in the direction that flatters the seller. Phones rotate their Bluetooth identifiers on a timer, so one person walking past can be counted several times over a long window. Ask whether the number is deduplicated to people and what the frequency was, because reach without frequency is not reach.
- What is proof of play, and why should I care?
- It is the difference between an ad you paid for and an ad that ran. Ask when the impression is counted. Counting at the bid means you are billed for a decision; counting at the completed play means you are billed for something that happened. We sign every completed play on the screen itself with Ed25519 and publish the key, so you can check the record without an account.
- What does a properly measured result look like?
- It survives all four questions above. We publish one in full, method included, so you can put the questions to it yourself before you spend anything: a four-week fast food restaurant flight across taxi and rideshare screens in 41 US markets, measured against a DMA-matched control cohort held out for the whole term.
See who you would be reaching.
87 audiences, each with its index, its price and a deal ID already live in your seat.