Blog posts
Many retailers are running the same promotion in four places at once: the weekly email, the front screen of the loyalty app, the entrance screen and main aisle and the in-store audio ads. Four surfaces, one offer, one shopper — and four different systems deciding independently who sees what.
Nobody designed it this way. It accumulated — the email platform with CRM, the app from digital, the screens from store operations or a signage supplier, each procured against its own business case. Each one works. What does not work is the seam between them, and the seam is where the money is.
Which is why omnichannel has stopped being an ambition and become a requirement. Retail media tenders now often ask for one platform across the retailer's own site, its app, the screens in store and in-store audio, with measurement that closes the loop to actual transactions. The question is no longer whether a retailer wants omnichannel retail media, but what has to be true underneath for it to work.
The instinct is to reach for an orchestration layer above everything, or a second platform replacing what is already in place. That instinct is right about the problem and wrong about the solution, because it treats every surface as an equivalent node on a network.
They are not equivalent. For most European grocery, convenience and specialty retailers, the physical store is the single largest concentration of customer attention and purchase intent the company owns — the dominant channel by a wide margin, and, as a media audience, usually the least defined. The site knows its visitors, the app knows its users, the CRM knows its segments. The store knows footfall and till receipts, and most of the people in it are anonymous at the moment they see a screen.
So the inherited state is not a hierarchy with the wrong thing at the top. It is four audiences with nothing at the centre. An omnichannel programme needs one, and the store is the natural candidate: site, app and email build reach and frequency around the visit; the store is where the visit converts. When the store sets the plan, every other channel inherits a target already validated against real purchasing behavior.
Once the store is the anchor, the omnichannel question narrows usefully: what is the most important thing that has to be common for the whole to add up? The answer is the audience layer —and the definition of who the audience is, how large it is, where it can be reached, and what it costs.
The media product is where the data from the audience layer turns into something a retailer can sell and a brand can buy. It takes an audience and adds the reach that comes with it. It sits at the centre because it is defined once, on the shared core, and does not depend on the channel that carries it. The same product can be delivered to in-store screens and audio or to the retailer's app and website, with one report back. Without that common unit, each channel has its own inventory and price list, and omnichannel remains a promise.
Packaging he audience as a media product is not enough alone. Something has to turn a sold audience into delivery — and this is where omnichannel programmes most often come apart, because the last step is not the same step in every channel.
Most of what a retail media platform does is channel-neutral. Campaigns, creative assets, user rights across head office and stores, the network itself, the audience definitions and the measurement — none of it cares whether the output is a screen in an aisle or a placement in an app. Everything built for the screens is just as valid for the web. That shared core is large.
The last step is different. A screen in a store is driven by a player: playlists and loops, dayparts, opening hours, hardware that fails, and operational content that has to be correct rather than merely delivered — a price, an availability flag, a menu that changes at eleven. A placement in an app or on a site is filled by an ad server: a decision at request time, impression by impression, with creative variants, frequency capping and competitive separation.
Neither can do the other's job. A digital signage player has no concept of a request or an impression; an ad server has no concept of an estate, a loop, a store that closed early, or content whose job is accuracy rather than reach. Stretching either across both is how a retailer ends up with in-store as a poor imitation of digital, or app placements scheduled like a playlist — and two sets of numbers that never reconcile.
So the architecture that holds is one platform with two delivery endpoints at its edge: players for the store estate, an ad server for app and site, both fed by the same campaigns and audience definitions and both reporting in the same units. That is how In-Store IMPACT is built, and it is why the audience layer — in the shared core, not in either endpoint — is what makes the numbers add up.
One distinction is worth making plainly, because tenders often blur it. Site, app, store screens and in-store audio are owned inventory: the retailer controls placement, data and measurement.
Offsite — the retailer's audiences activated on media it does not own — is bought inventory, measured by what the partner returns, on different economics. Both belong in an omnichannel strategy, but listing them as one undifferentiated set is how programmes promise a closed loop they cannot deliver.
Make the owned channels add up first; that is what makes an offsite extension worth anything later. An audience proven against transactions is portable. One that has not is just a segment name.
There is a harder reason the store belongs at the centre, and it has nothing to do with reach. In-store is the only channel where exposure and purchase happen in the same visit. A shopper sees the message at the shelf edge and either puts the product in the trolley or does not, and the point-of-sale data says which. Every other channel has to model its way to that answer; the store observes it.
An audience layer built from those transaction and loyalty signals is not the store's audience extended elsewhere; it is the one definition every channel can be checked against.
That gives an omnichannel programme its natural order: define audiences from data tied to purchase, prove them where the purchase happens, then extend the same definitions outward. That order compounds evidence.
For any retailer assembling an omnichannel retail media offer, the practical test is short: can you sell one audience across your store, your app and your site, deliver it in all three, and report it as one number?
If not, the missing piece is not usually to add another ad server but a single unified system that provides a common audience layer — and where that layer lives determines whether the next five years of channel investment compounds or fragments.
Omnichannel retail media starts in the store.
