Retail media has become one of the most attractive growth stories in commerce because the economics are difficult to ignore. Retailers already own the customer traffic, the transaction and, increasingly, a meaningful amount of first-party data. Turning some of that attention into advertising revenue can produce margins that look considerably better than the underlying retail business.
That has made retail media easy to describe as a new advertising channel. Build sponsored placements, sell search visibility, add display inventory, create supplier packages and watch a new revenue line appear.
The description is accurate. It is also incomplete.
The more interesting asset is not the advertising inventory sitting around the transaction. It is the intelligence generated by the transaction itself: what customers searched for, considered, substituted, ignored, bought, returned and bought again. A retailer that understands that difference can build something far more valuable than another place for brands to spend media budgets.
That distinction matters because advertising inventory eventually becomes comparable. Commerce intelligence is much harder to commoditise.
The transaction changes the economics of advertising
Most advertising platforms help a brand understand whether somebody was exposed to a message and then behaved differently. Retailers have something more powerful: they can often connect the media exposure directly to a commercial outcome.
That closed-loop promise is what made retail media compelling in the first place. A beauty brand can know that customers who saw a sponsored placement subsequently bought the product. A grocery supplier can measure whether promotion drove incremental units. A fashion brand can see how paid visibility influenced product discovery, conversion and potentially repeat behaviour.
Having spent years operating businesses where marketing teams, commercial teams and technology teams each carried a slightly different version of attribution, I understand why this is attractive. Connecting media investment to a transaction feels refreshingly concrete.
The danger is assuming that the transaction makes attribution perfect.
It does not.
A customer may have seen the product elsewhere, searched for it directly, encountered it in-store, received an email and then clicked a sponsored placement immediately before buying. Retail media improves the evidence, but it does not abolish the messy reality of human decision-making.
That distinction becomes important when retailers start promising suppliers certainty they do not actually possess. Closed-loop measurement should make commercial conversations more intelligent, not simply make attribution claims more confident.
Used well, the transaction becomes a powerful source of evidence. Used lazily, it gives bad measurement a more impressive dashboard.
The easiest money can create the wrong retail media business
Once retailers discover the margin available in media, the temptation is predictable. More sponsored results appear. More placements are created. Search pages acquire additional inventory. Supplier negotiations begin carrying media targets alongside trading terms.
Revenue rises quickly.
Customer experience can deteriorate just as quickly.
Retailers occupy an unusual position because they control both the commercial environment and the customer journey. A media platform can optimise advertising yield without worrying whether the shelf itself becomes less useful. A retailer cannot afford that luxury.
If the first search result is there because it is commercially sponsored rather than genuinely relevant, the customer may tolerate it. If the first five are there for that reason, search begins to lose credibility. A product page covered in paid placements may monetise attention today while quietly making the underlying experience worse.
This is where retail media needs stronger discipline than conventional advertising.
The inventory is not sitting beside the business. It is sitting inside the shopping experience.
Every sponsored placement therefore creates two economic effects. It produces media revenue, but it can also influence conversion, basket composition, customer trust and the probability that somebody returns.
Those effects should be measured together.
I have always found this one of the more revealing tests of whether an organisation is genuinely omnichannel and customer-led. If the media team is rewarded for advertising yield while the commerce team absorbs any impact on conversion, the operating model has already created the wrong answer before the first campaign is sold.
Retail media cannot be managed as a high-margin annex to retail.
It has to respect the economics of the customer relationship it is monetising.
The proprietary advantage is what retailers know after the click
Advertising inventory can be copied.
A competitor can add sponsored search. A marketplace can create display placements. A technology provider can offer a similar auction engine. Over time, the mechanics become standard.
The more defensible advantage sits underneath them.
Retailers know which products were purchased together. They know when customers trade up or down. They can see whether a promoted customer returns without another incentive, whether a new product creates incremental category spend or merely shifts demand from an existing item, and whether the acquired customer has any long-term value.
Those questions move retail media away from advertising and toward commerce intelligence.
Consider the difference between telling a supplier that a campaign generated one million impressions and telling them that the campaign recruited a new cohort of customers who subsequently repeated at a higher rate, expanded the category and showed lower promotional dependency.
Both are media results.
Only one changes the commercial conversation.
The same applies to assortment. A retailer may observe that customers searching for one product repeatedly substitute into another brand when inventory is unavailable. Search behaviour may reveal unmet demand before sales history catches up. Basket relationships may show that a supplier’s product is far more important to customer missions than its standalone sales ranking suggests.
At that point, the data is no longer useful only to the marketing department.
It becomes useful to merchandising, category management, pricing, suppliers and product development.
That is where the economics start becoming more interesting.
Supplier relationships can move from negotiation to shared intelligence
Traditional retailer-supplier relationships contain a fairly predictable tension.
The supplier wants more distribution, better placement and stronger promotion. The retailer wants margin, funding and terms. Both sides arrive with data, although historically they have not always arrived with the same data or the same interpretation of it.
Retail media has the potential to improve that relationship because the retailer can bring a much richer understanding of actual customer behaviour into the discussion.
A supplier does not only need to know whether a campaign performed. It wants to understand who is buying, what recruited them, what else sits in the basket, whether the customer came from a competing brand, how promotion altered demand and whether the behaviour continued when paid support stopped.
That intelligence can change everything from assortment decisions to launch strategy.
Imagine a new product performs moderately in total sales but disproportionately recruits customers who are new to the category. The advertising return may look acceptable rather than extraordinary. The strategic value could be much greater.
Conversely, a campaign can generate an impressive return on ad spend while almost entirely subsidising customers who would have bought the product anyway.
The first example may deserve more investment than conventional media metrics suggest. The second may deserve less.
This is why I believe the strongest retail media businesses will eventually be judged less by the amount of inventory they sell and more by the quality of the decisions their intelligence helps suppliers make.
That requires a different capability from media sales.
It requires analytics, experimentation, customer economics and enough commercial maturity to tell a supplier something more useful than “buy more impressions”.
Measurement has to move beyond ROAS
Return on ad spend became popular because it is wonderfully simple. Spend one dollar, generate several dollars in attributed revenue, declare success.
The simplicity is also the weakness.
A high ROAS can come from advertising to customers already predisposed to buy. A lower ROAS campaign may recruit genuinely new customers or change category behaviour. Margin can differ dramatically between products. Repeat behaviour can make an initially expensive acquisition far more attractive over time.
Retailers are unusually well placed to see those differences because they sit close to the transaction and, where loyalty penetration is strong, the customer.
That should allow retail media measurement to mature beyond the narrow campaign window.
Incrementality matters. New-to-brand behaviour matters. Margin matters. Repeat matters. Customer lifetime economics matter.
The purpose is not to bury suppliers under another fifty metrics. It is to identify which outcomes actually changed because the investment happened.
That requires experimentation rather than attribution alone.
Holdout groups, geographic tests, customer cohorts and other forms of incrementality measurement may be less visually satisfying than a dashboard that claims precise return on every impression. They are often much closer to the commercial truth.
Retail media has an opportunity here because it can connect marketing measurement with actual retail economics.
If it chooses instead to recreate the same attribution theatre the advertising industry has spent years struggling with, it will have wasted one of its greatest advantages.
AI makes the intelligence layer more valuable
AI will accelerate almost every mechanical part of retail media.
Campaign creation will become easier. Bidding and placement optimisation will improve. Creative variants will multiply. Audience construction, reporting and recommendation will become increasingly automated.
Those capabilities will spread quickly.
The scarce asset will increasingly be the quality of the proprietary signal underneath them.
An AI system that knows a customer viewed a product is useful. One that understands the customer usually buys premium in this category, is unusually promotion-sensitive in another, recently shifted brands, has a high probability of repeat and tends to purchase certain products together can make much better decisions.
That context belongs to the retailer.
More importantly, it is generated by operating the commerce business rather than the advertising business.
This creates a strategic choice.
Retailers can use AI to sell advertising inventory more efficiently, which will certainly create value. Or they can use it to understand demand, customers and supplier economics more deeply, then allow media to become one monetisation layer on top of that intelligence.
The second path is harder.
It requires strong identity resolution, governed data, analytical capability and a commercial organisation willing to share intelligence across traditional functional boundaries. It also requires restraint, because the most valuable use of a signal will not always be another ad.
Sometimes the right answer will be changing the assortment.
Sometimes it will be improving search relevance.
Sometimes it will be reducing promotion.
Sometimes it will be telling a supplier that the customer problem is not awareness at all.
That is why the quality of the intelligence layer matters more than the sophistication of the ad server.
Retail media needs a broader operating model
If retail media becomes strategically important, the organisational design around it matters.
A business built purely around media revenue will naturally optimise yield. A merchandising organisation will optimise category economics. Marketing will care about customer acquisition and retention. Digital teams will care about conversion and experience.
All of them can be right individually and still produce a poor answer collectively.
The retailer therefore needs clear governance around what can be monetised, where sponsored inventory is appropriate, how customer experience is protected and how supplier insights are shared across the enterprise.
It also needs a common commercial language.
If media revenue rises while conversion falls, that is not obviously success. If a supplier buys more advertising but customer acquisition becomes less incremental, the relationship may be getting bigger without becoming better. If data monetisation improves but customers begin to feel the experience is increasingly shaped by whoever paid most, the business is consuming trust to generate revenue.
Those trade-offs belong at senior commercial level because they cross P&Ls.
Retail media becomes much more interesting when it stops being treated as a separate high-margin business and starts becoming a capability that improves the economics of the wider enterprise.
Advertising remains part of that.
It just should not be the boundary of the ambition.
The transaction is the beginning, not the end
Retail media is often described as the monetisation of first-party data.
I think that framing still undersells it.
The strategic opportunity is to build a better understanding of commerce from signals that very few other businesses possess at the same depth: intent, discovery, transaction, fulfilment, returns, loyalty and repeat behaviour.
Advertising is one way to monetise that understanding.
It may not ultimately be the most important one.
The retailer that can tell a supplier where demand is emerging, why customers switch, which promotions are genuinely incremental and how product decisions affect long-term customer value becomes more than a place to buy media.
It becomes an intelligence partner.
That is a more defensible position because it is rooted in something advertising platforms have always struggled to obtain cleanly: the economic truth of what happened after the customer decided.
The transaction is therefore not the end of the media funnel.
It is the beginning of the intelligence business.