Skip to content

Media · Mezzi · Deep dive

Retail media: advertising near the point of purchase

Retail media in 2026: onsite, offsite and in-store formats, how it is bought at auction and measured with closed loop and incrementality, close to the purchase.

Software house — su misura, costruito da noi

RETAIL MEDIA · CLOSE TO PURCHASE

In · Shopper in the buying phase

Out · Measurable purchase

01

Onsite · site and app

The spaces on the retailer's site and app: sponsored products in internal search, display, video on the product page. The most mature surface and the closest to the buy button.

02

Offsite · data beyond the site

The retailer's purchase data activated off its own property — social, programmatic, connected TV, email. It extends reach beyond onsite inventory while keeping first-party precision.

03

In-store · point of sale

The physical store as a medium: screens, audio, smart shelves, QR at checkout. In 2026 it becomes measurable inventory, no longer signage.

04

First-party purchase data

The engine that makes all this possible: first-party purchase data, consented and tied to real behaviour, driving targeting and measurement without third-party cookies.

05

Closed loop · incrementality

The closed loop and incrementality: linking exposure to a verified purchase and proving, with control groups, how much sales were genuinely created.

What it is, in plain words

Advertising inside the store, next to the purchase

Retail media means buying ad space inside a retailer's selling environments — its website, its app, its physical stores — and doing it with the first-party purchase data that retailer owns. It is the supermarket shelf turned into a measurable advertising medium.

Where it lives

Not on a neutral channel, but inside the seller's own property: the product page, the internal search result, the digital shelf, the in-aisle screen. That is where the person is already deciding what to buy.

What makes it different

The retailer knows real purchases, not just clicks. It can tell whether someone who saw the ad then added the product to the basket: that is the famous "closed loop" between exposure and sale.

Why it is exploding in 2026

Third-party cookies are going dark and purchase data becomes the cleanest asset there is. Whoever owns the checkout owns the data: that is why every major chain is also becoming a small publisher.

The formats

Three surfaces, one logic

Retail media isn't a single format: it's a family of spaces ranging from internal search to the digital in-aisle sign. They are bought separately, but they all answer to the same data.

01

Onsite — on the retailer's site and app

Sponsored products in internal search results, display banners, video on product pages. It is the most mature format and the closest of all to the "add to basket" button.

02

Offsite — the retailer's data, off its own site

The same purchase audiences activated on social, programmatic, connected TV and email. It extends reach beyond onsite inventory, now saturating, while keeping the precision of the first-party data.

03

In-store — inside the physical store

Digital screens, audio, smart shelves, QR activations at checkout. In 2026 it stops being static signage and becomes inventory measured to the same standards as digital.

Retail media doesn't sell generic attention: it sells proximity to the moment someone reaches for their wallet.

How it is bought

From the auction to the audience, in the clear

Buying almost always happens inside the retailer's own platform, self-serve or managed. The mechanics are simple; the discipline lies entirely in not scattering across networks.

1
Self-serve or managed service

Networks offer an interface where you set campaigns and bids yourself, or a managed service for premium formats and large budgets. Often both run on the same network.

2
Auction, usually second-price

Sponsored products work like paid search: you bid on keywords or categories and win the space, paying just above the second-best bid. Display more often goes by CPM.

3
Audiences built on purchase data

Targeting doesn't start from inferences on a cookie, but from real behaviour: who bought the category, who abandoned it, who is new to the brand. This is what makes retail media so precise.

4
Many networks, one plan

A serious brand in 2026 runs six or more different networks, each with its own logins, rules and naming. Buying well means orchestrating them into one plan, not chasing them in silos.

Onsite campaigns launch in days; the real difference is cross-network orchestration, which is continuous work, not a one-off setup.

How it is measured

Closed loop, but eyes open

Retail media promises the cleanest measurement on the market — the ad tied to a verified purchase. The promise is real; the risk is mistaking an attribution figure for proof of incrementality.

What you look atWhat it really tells youThe trap to avoid
Closed-loop attributionLinks exposure to the real purchase via the purchase ID or loyalty card.Often last-click in short windows: it captures demand, it doesn't always create it.
Network ROASThe return reported by the platform you buy on.It's the network grading its own homework: without independent verification it means little.
IncrementalityThe extra sales that wouldn't have happened without the ad, measured with control groups.It's the metric that counts — and the one fewest networks genuinely offer.
New-to-brand customersHow much of the sale comes from people who had never bought the product.It separates growth from rewarding those who would have bought anyway.
Halo and cannibalisationWhether spend shifts sales between networks or channels instead of adding to them.Visible only by comparing networks against each other, not inside one dashboard.
When it works and when it doesn't

The signs you're using it well — or badly

Retail media is powerful close to conversion and fragile when treated as a pure attribution machine. You recognise it like this.

  • It works when you defend and grow at the shelf. You are present where the decision actually happens, in the right category, with purchase data driving the bid.
  • It works when first-party data is your advantage. Without third-party cookies, an audience built on real purchases is the cleanest base for targeting and measurement.
  • It breaks when you chase reported ROAS alone. A network grading itself inflates its own credit: without incrementality you reward sales you'd have had anyway.
  • It breaks when networks stay siloed. Six logins, six naming conventions, six reports that don't talk: you spend more and can't tell what is truly incremental.
  • It breaks when you use it as the only medium. Retail media harvests demand close to purchase; with no medium creating demand upstream, you harvest less and less.

True whether you sell at the shelf or for the brand: the balance between harvesting and creating demand changes, the principle doesn't.

The TMM approach

Media-neutral, transparent, measurable

We don't sell a network: we plan retail media as one part of the medium, not the medium. We treat it with the same rigour as any channel — clear rules, a legible chain, independent measurement.

Media-neutrality first

We decide how much retail media you need by its role in the purchase journey, not by who pays the highest commission. If demand has to be created elsewhere, we say so.

One reading across networks

We bring six or more networks back to comparable metrics and naming, so you know what is incremental and what is merely reported — not six dashboards that don't talk.

Incrementality, not self-scoring

We test credit with control groups and independent verification, to separate sales created from sales simply harvested.

Chain and take rate in the clear

We make margins, commissions and paths legible: you know how much of the budget buys space and how much stays along the chain.

Straight questions

Before you ask

Does retail media replace the rest of the media plan?

No. It is formidable close to conversion, where demand already exists. But it harvests demand, it doesn't create it: with no channels building awareness and desire upstream, over time it harvests less and less. Plan it as one lever of the medium, not the whole medium.

Is it only useful if I sell on large marketplaces?

No. The surfaces are the same at every scale: a chain's internal search, the in-aisle screens, a local banner's data. What changes is the breadth of inventory, not the principle of buying close to purchase with first-party data.

Is the closed loop proof that the advertising worked?

Not on its own. Closed-loop attribution links exposure to purchase, but it is often last-click in short windows: it tells you who bought after seeing, not who bought because they saw. The proof is incrementality, measured with control groups.

Why is everyone talking about first-party data in 2026?

Because third-party cookies are no longer a reliable base for targeting and measurement. A retailer's purchase data is first-party, consented and tied to real behaviour: it is the cleanest asset left, and that is why retail media grew so fast.

How do you avoid scattering across too many networks?

With single orchestration. Each network is a walled garden with its own metrics and naming; running them in silos means paying more and not knowing what is incremental. You need shared rules, a comparable reading and independent verification above the platforms.

Is in-store really measurable or just signage?

In 2026 it is becoming genuinely measurable. With industry standards on impressions and viewability, and a link to the same loyalty ID as onsite and offsite, the in-aisle screen competes for budget on the same terms as the other formats — no longer on gut feel.

Cases

From problem to result — anonymised.

FMCG · anonymised

Many networks, no common reading

Problem Campaigns scattered across several networks, each with its own reported ROAS and naming: no way to tell what was genuinely incremental.

Method Reconciliation to comparable cross-network metrics + incrementality tests with control groups, above the platforms.

Result Budget shifted from merely "reported" activity to what creates new sales, with one reading for every network.

Retail · anonymised

Onsite saturated, growth stalled

Problem Internal-search inventory now saturated and expensive, with spend rewarding those who would have bought anyway.

Method Offsite extension of the purchase data + new-to-brand measurement, to separate harvesting from growth.

Result Reach extended beyond the digital shelf and credit rebalanced toward acquisition, not just defence.

Beauty · anonymised

Closed loop mistaken for proof

Problem Decisions made on the network's last-click attribution, which certified existing demand as a campaign effect.

Method Control groups and independent verification placed alongside the platform's attribution.

Result A clear distinction between sales created and sales harvested, with budget reallocated to what truly moves the business.

Go deeper

Bring this to your stack.