Measuring in-store media: what you can and can't count

In-store retail media can be measured at three levels: what the screens did (plays and uptime), who was likely there to see them (impressions, modeled from store traffic), and what happened at the register (transactions). The first two are standard for digital out-of-home. The third is where a network built on point-of-sale systems stands apart.

Knowing which level you're looking at matters more than any single number. Plenty of measurement arguments are really two people talking about different levels of the same campaign.

What can a screen actually count?

The screen itself counts delivery. Every networked screen can log which creative played, when it played, and whether the display was on and functioning. In the trade this is called proof-of-play, and it's the in-store equivalent of a served impression: evidence the ad ran as bought.

NRS Digital Media operates 39,000+ measured digital screens across 34,000+ independently-owned stores, so delivery reporting can be grounded in real screens in real locations, not estimates of where a loop might be running. Proof-of-play is necessary but not sufficient. A screen playing to an empty aisle delivered nothing, which is why measurement can't stop here.

How are impressions estimated?

Nobody swipes a card to look at a screen, so out-of-home advertising counts audiences by modeling them. The basic recipe combines venue traffic with screen placement: how many people pass through the store, and what share of them plausibly had the ad in view. Geopath, the industry body for out-of-home audience measurement, exists to standardize exactly this kind of modeling.

A retail network has a better traffic source than most venues: its own registers. The NRS network sees 298 million weekly visits, and that visit count comes from transaction activity rather than a survey or a passerby estimate. Modeled impressions are still estimates, but estimates anchored to counted store traffic are sturdier than estimates anchored to guesses.

What does the register add?

This is the level most media channels never reach. Because the screens run on the same point-of-sale systems that ring up sales, the store that shows the ad is also the store that records the outcome. Across the network, that outcome record covers 1.9 billion transactions annually.

Transaction data lets a brand ask outcome questions directly. Did the advertised item sell faster in stores that ran the campaign than in comparable stores that didn't? Did category share shift during the flight? Because every record carries a timestamp, results can even be read by daypart, comparing the hours a campaign ran against the same hours before it started. The scan-data analytics arm of the business, NRS Insights, works with this same transaction layer, which is why sales measurement here isn't an add-on borrowed from a third party.

Where does in-store measurement hit its limits?

Honesty about limits is part of measurement. In-store media can't count exact eyeballs; impressions remain modeled opportunities to see, not verified views. It can't follow a shopper home or connect an exposure to a specific person, and that's by design, since no personal data is involved. And it records outcomes, not reasons. Sales data can show that a product moved. It can't say what the shopper was thinking.

None of this is unusual. Every channel's measurement has assumptions underneath it. The useful question is whether the assumptions are visible and whether the outcome data is real. When the outcome data comes from the register itself, at least one end of the chain is solid ground.