Commerce Media Technologies

A grocery chain makes somewhere between two and five percent margin selling food. It makes considerably more than that selling advertising to the brands whose food it sells.

That arithmetic built an industry. Forrester forecasts the commerce media market growing from around $184 billion in 2025 to $312 billion by 2030, roughly eleven percent compound annual growth, arriving at approximately twice the level of global television advertising spend.

It also explains why the terminology keeps moving. What started as retail media, meaning ads on a retailer's own website, has outgrown that description, and the technology category has been renamed to follow it.

Retail media, then commerce media

The original proposition was narrow and effective. A retailer has shoppers on its site with purchase intent, and it has data about what those shoppers buy. Brands will pay to reach them at that moment. Sponsored product listings, display placements, and category takeovers on the retailer's own properties.

Commerce media is the superset. It takes the same first-party commerce data and extends it beyond the retailer's owned properties into connected TV, social, streaming, in-store screens, audio, and other commerce-adjacent environments.

The distinction matters because it changes what the technology has to do. Selling ads on your own website is an ad server problem. Extending your audience data into a CTV buy on someone else's inventory is an identity, privacy, and interoperability problem.

Forrester's own framing anticipated this early, describing media networks from commerce companies as growing and transforming shopping. The savvier retailers were never only monetising the website. They were building media options that reach consumers at every touchpoint, including digitising physical stores with advertising and offering brands access to their first-party data for campaigns running on major publisher sites.

The two sides of the market

The technology splits by who is being served, and Forrester evaluates the supply side specifically.

Sell-side platforms serve the retailer or commerce company operating the network. Ad serving, auction and yield management, inventory management, self-service interfaces for advertisers, reporting, and increasingly the sales support required to bring a media business to market. Forrester's evaluation of this side found that over fifty eight percent of client references used their retail media solution to help sell the media offering itself, which tells you something about what retailers were actually buying. Technology, and help finding demand for it.

Buy-side platforms serve the brand or agency spending the money. Campaign management across many networks, bid optimisation, creative, and the measurement layer that tries to compare performance across retailers who each report differently.

The market's structural problem is visible in that split. A brand advertising across twenty retail media networks faces twenty different interfaces, twenty measurement methodologies, and twenty definitions of what counts as attributable. Forrester Consulting research commissioned by one vendor found more than forty percent of marketers using six to ten different platforms for media buying, with consolidation onto a single platform for managing commerce media spend and performance named as a top priority.

Measurement is the unsolved problem

The founding promise of retail media was closed-loop attribution. An ad impression connects to a purchase, in the same data set, with no modelling in between. That is genuinely more than most advertising channels can offer, and it drove a decade of budget reallocation.

The delivery has not matched it.

Forrester's State of Retail Media research found eighty six percent of commerce media decision-makers across North America and Europe naming strengthening measurement and attribution as a high or critical priority. When that many practitioners in an established market call the same thing a priority, it is not a refinement. It is a gap.

The specific failure is that closed-loop attribution in practice usually means last-click within a limited window, which systematically over-credits the advertising that ran closest to a purchase somebody was going to make anyway. Industry research on this consistently finds incrementality, meaning whether the ad caused the sale rather than merely preceded it, to be the measurement problem advertisers most want solved and least often solve.

That gap between promised and delivered measurement is the single largest source of advertiser frustration in this category, and it is where the technology vendors compete hardest, because a network that can credibly demonstrate incrementality can defend its rates.

The Forrester Wave on the sell side

Forrester's scored evaluation of this market covers sell-side retail media solutions, and Criteo took the highest score in the current offering category.

Criteo's position rests on scale and on yield mechanics: click prediction models that determine how much revenue a retailer extracts from sponsored product inventory, and unified auctions where every demand source competes for the same inventory simultaneously rather than in sequence. That second point is less glamorous than it sounds and matters enormously to a retailer's economics, because sequential auctions leave money on the table.

The competitive field around it has moved considerably. Two significant competitors, CitrusAd and Microsoft-owned PromoteIQ, wound down operations, while a set of newer entrants including Koddi, Topsort, and Osmos compete for retailer accounts, alongside Mirakl Ads and dunnhumby. On the buy side, Skai, Pacvue, and Perpetua serve brands managing spend across networks.

Retailers have been putting this technology back out to tender as the market matures, which is a normal sign of a category moving from land-grab to renewal cycle.

Build or buy

This is the decision that defines the category and it does not have a general answer.

Building in-house makes sense at the very largest scale, where media revenue justifies a dedicated engineering organisation and where the retailer's data advantage is significant enough to be worth protecting. The largest networks run their own stacks for exactly this reason.

Buying makes sense for almost everyone else, and the practical questions are integration and demand.

Integration, because a retail media platform has to connect to the product catalogue, order management, point of sale, and customer data platform. Deployment timelines in this market are quoted in weeks by vendors with strong API tooling and considerably longer where systems are fragmented, and the difference is your systems rather than their software.

Demand, because inventory without advertisers earns nothing. A retailer launching a network needs national brand budgets to flow to it, and that is a business development problem more than a technology one. It is why so many references reported using their platform to help sell the offering, and why a vendor's existing advertiser relationships can matter more than its feature set.

The threat Forrester has named

Here is the prediction that should shape how anyone reads this category right now.

Forrester's 2026 predictions include agentic commerce reducing retail media ad sales by twenty percent, on the reasoning that generative AI search engines become transactional platforms.

Follow the logic. Retail media works because a human shopper browses a page, sees a sponsored placement, and is influenced by it. Remove the browsing human and the mechanism has nothing to act on. An agent instructed to reorder household supplies within a budget does not notice a sponsored listing, is not swayed by a hero banner, and does not have a moment of consideration to interrupt.

If a material share of routine purchasing moves to agents, sponsored placement inventory loses its audience without losing its inventory, which is the worst possible position for a media business.

Forrester also predicts a surge in adtech mergers and acquisitions, with more than a hundred transactions in 2026 as vendors and private equity firms respond to a restructuring Google adtech landscape.

Neither prediction is certain. Both are worth putting in front of anyone about to sign a multi-year commitment in this category, because they describe scenarios where the demand assumptions underlying the business case change materially inside the contract term.

Where the money actually comes from

One structural point that gets lost in the technology discussion.

Retail media budgets have historically been funded from trade promotion and shopper marketing rather than from brand advertising. That money was already flowing from brands to retailers as slotting fees, co-op advertising, and in-store promotional support. Retail media largely repackaged it into an auction-based digital product with better reporting.

Two consequences follow.

The growth figures partly describe reallocation rather than pure new spending, which matters when assessing how much further the category can grow before it is drawing from genuinely new budgets.

And the buyer relationship is not a normal advertising relationship. A brand negotiating retail media spend with a retailer is negotiating with a partner that also controls its shelf space, its promotional calendar, and its distribution. That asymmetry does not appear in any technology evaluation and it shapes every commercial conversation in this market.

What to test

If you are a retailer selecting a platform, ask about yield mechanics rather than features. Unified versus sequential auctions, click prediction accuracy, and fill rates determine your revenue far more than the interface does.

Establish the integration path against your actual stack, including the point of sale system nobody wants to touch. Vendors quote deployment in weeks on the assumption of clean APIs.

Ask what the vendor does about demand generation, concretely. Which advertiser relationships come with the platform, and what sales support is included.

If you are a brand, insist on incrementality testing rather than accepting attributed return on ad spend. Holdout groups, geo experiments, or matched market tests. A network confident in its performance will support this, and the ones that resist are telling you something.

Compare methodologies before comparing numbers across networks. Attribution windows, view-through treatment, and what counts as a conversion vary enough that reported performance figures from two retailers are frequently not measuring the same thing.

And ask both sides what happens when the shopper is an agent. Nobody has a complete answer yet, but the quality of the thinking separates vendors who have considered it from vendors who have not.

Forrester does not endorse any vendor named here, and tier placement should not be read as a recommendation to buy.