Retail is a huge space and when done right, it can even beat a well set B2B channel.
While retail has won the growth argument, it is yet to work on its credibility.
US retail media spending is expected to reach $107.6 billion in 2026, nearly three times its level five years ago.
At the same time, 67% of CMOs say they plan to increase their retail media spending this year. The problem is that only 53% believe retail media networks currently provide adequate measurement and attribution capabilities.
That gap matters.
No one is asking whether they generated sales.
Brands are more concerned about the sales they would have as it is generated.
And this is a question of incrementality, which is becoming the retail media’s important measurement challenge.
ROAS Tells You What Happened. Incrementality Asks Why
Retail media has an obvious advantage over many traditional forms of advertising: it sits close to the transaction.
For instance, a shopper sees an ad, searches for a product, clicks on a sponsored listing and buys it. The retailer can often connect those actions with remarkable precision. That makes retail media highly measurable.
But measurable does not automatically mean incremental.
Suppose a shopper was already planning to buy a particular brand of shampoo. They search for it on a retailer’s website, see a sponsored product listing and purchase it.
The campaign may receive credit for the sale. But did the advertising actually create the sale? Possibly not.
The consumer may have bought the product regardless.
This is where traditional return on ad spends, or ROAS, starts to become an incomplete answer. ROAS can tell marketers how much revenue was associated with advertising spend. It does not necessarily tell them how much additional revenue the advertising created.
The Retailer Knows More Than the Brand — But That Doesn’t Solve the Problem
Retail media networks possess something advertisers desperately want: first-party commerce data.
They can see searches, product views, purchases, basket behavior and, in many cases, loyalty activity.
That makes them incredibly valuable advertising environments.
It also creates a measurement paradox.
A retailer may have an excellent view of what happens inside its ecosystem while having a less complete view of what happens outside it.
A consumer may discover a product on social media, hear about it from a creator, search for it on Google, visit a physical store and eventually purchase it through a retailer’s app.
Which channel created the demand?
The final transaction may happen inside the retail environment, but the decision may have been influenced much earlier.
This is why retail media cannot be evaluated entirely as a collection of isolated campaigns.
It is increasingly part of a connected consumer journey that spans paid media, organic discovery, promotions, pricing, brand activity, search, physical stores and multiple retailers.
Measuring one piece of that journey in isolation can produce an impressive number and still lead to a poor investment decision.
The Attribution Trap
Attribution is not useless. It is simply being asked to answer a question it was never designed to answer.
- Last-touch attribution is useful for understanding the final identifiable interaction before conversion.
- Multi-touch attribution can distribute credit across multiple interactions.
But neither automatically establishes causality.
Now imagine two groups of shoppers.
One group sees an advertisement for a product.
The second group does not.
If the first group buys more, the difference might indicate advertising impact.
But there may be other explanations. Perhaps the people exposed to the advertisement were already more likely to buy. Perhaps they were searching for the product before seeing the ad. Perhaps a promotion was running simultaneously. Perhaps distribution improved in the same period.
To establish incrementality, marketers need some form of credible counterfactual:
What would have happened if these consumers had not received the advertising?
That is a much harder question. And it is the reason incrementality is increasingly moving from an analytics exercise to a strategic capability.
Experiments Are Powerful. But They Are Not Always Practical.
The cleanest way to establish incremental impact is often experimentation. A brand can create a treatment group that receives advertising and a control group that does not, then compare outcomes.
In theory, the difference provides a much stronger indication of causal impact.
In practice, retail media experimentation can be difficult.
Campaigns operate across multiple retailers. Audiences overlap. Inventory changes. Promotional calendars interfere with clean tests. Brands may not have enough scale to run statistically meaningful experiments. Some retailers may offer different measurement capabilities or methodologies.
That is why the industry is moving toward a broader measurement toolkit. The implication for marketers is important:
There is no single magic incrementality model.
The right methodology depends on the business question, available data, campaign scale and level of causal certainty required.
MMM Is Becoming Relevant Again
Marketing mix modeling, or MMM, is also gaining renewed attention because it can look beyond an individual campaign or platform.
Instead of asking whether one advertisement received credit for a purchase, MMM attempts to understand how different marketing investments contribute to business outcomes over time.
That broader view becomes particularly valuable as retail media increasingly interacts with other channels.
A retailer promotion may increase the effectiveness of an advertising campaign. Brand advertising may increase the number of people searching for a product. Retail media may then help convert that existing demand.
Looking at only the final touchpoint can make the retail media investment appear more powerful or less powerful than it actually is.
Modern MMM is also becoming more accessible.
Google has continued expanding Meridian, its open-source marketing mix modeling framework, including integration with Google Analytics 360. Its stated objective is to help marketers unify first-party and cross-channel data, understand causal performance and model future investment scenarios.
That matters because measurement is shifting from retrospective reporting toward investment decision-making.
The goal is to answer:
“If I move $10 million from one channel to another, what is likely to happen?”
AI Will Make Measurement Faster. It Won’t Automatically Make It True.
Artificial intelligence is entering measurement at exactly the right moment and with exactly the wrong potential expectation. AI can help marketers process fragmented datasets, identify patterns, run scenarios and accelerate analysis.
But AI cannot magically manufacture causality from poor data.
AI can improve the speed of measurement. It can improve the accessibility of measurement. It can help identify relationships humans might overlook.
But if the underlying methodology cannot distinguish correlation from causation, AI can simply produce a more sophisticated-looking version of the same mistake.
For senior marketers, that means AI-powered measurement should not be judged by how impressive the dashboard looks.
It should be judged by whether it improves the quality of budget decisions.
The retail media market is becoming more competitive.
As more retailers build media networks and more brands allocate significant portions of their budgets to them, simply having access to valuable audiences will not be enough.
Measurement quality will increasingly become part of the value proposition. That is already visible in the market.
Retailers that can demonstrate credible, comparable and transparent business outcomes will have an advantage over networks that simply report high ROAS. Trust among consumers is going to be a discerning factor to boost retail sales for any brands.
Read More: Agent-to-Agent Salestech: What Happens When Buyer Bots and Seller Bots Negotiate Deals?
