Retail media has become one of the fastest-growing lines in the CPG marketing budget. Amazon, Walmart Connect, and Kroger have built real advertising businesses, and brands now pour money into them. But the way most brands measure that spending is a problem. When you measure retail media using a retailer’s own reporting, you’re reading a scorecard written by the company you just paid.
That’s not a knock on retailers. Their data is useful. The problem is what it can and can’t see. And what it can’t see is costing CPG brands a clear view of their marketing’s impact.
Why Retailers Can’t Really Measure Retail Media
Retail media platforms report on what happens inside their own walls. An Amazon campaign gets measured by Amazon. A Walmart Connect campaign gets measured by Walmart. Each platform counts the conversions it can see on its own site and reports them back to you. Simply ask yourself: how many times has an advertising platform recommended that you lower your spend with them?
Two things go wrong with this:
- Every platform has a reason to report high numbers. The better a campaign looks, the more you spend the next quarter.
- No platform can see what happens off its own site. If your Amazon ad drives a sale in a grocery store, Amazon never counts it.
That second point is bigger than most brands realize. And it persists even when retailers use third-party measurement providers to create an appearance of transparency and independence. The methodology is still only seeing what happens on their platform. Technically independent, fundamentally limited.
Think about how people actually shop for CPG products. Someone scrolling Amazon sees your ad for a snack brand or a household cleaner, but doesn’t buy it right then. A few days later, they grab it off the shelf during their weekly grocery run. The ad did its job. It drove the sale. But because the purchase happened in a store rather than on Amazon, Amazon’s report shows no record of it. The campaign looks weaker than it actually was, and you could end up cutting the budget from something that’s quietly working.
The Halo Effect Retailer Data Misses
Here’s what the retailer scorecard leaves out. When someone sees your ad on Amazon, they don’t only buy on Amazon. They also buy in stores, on your website, or somewhere else entirely, often days later. Every time your ad shows up, it’s reinforcing your brand and nurturing future sales, even when you can’t tie that impact back to a single click.
The opposite problem is actually more common and more costly. Amazon’s attribution system can give its ads credit for a sale that was partly driven by your brand’s own TV, display, or in-store marketing activity, or for a purchase the customer would have made regardless. When platforms overcount their contribution, brands over-invest in retail media at the expense of the channels doing the upstream work.
Our ROI Genome® shows how big these gaps really are:
- 45% of the sales driven by Amazon display ads happen off Amazon.
- 23% of the sales driven by Amazon sponsored search happen off Amazon.
So if you measure your Amazon display spend using only Amazon’s numbers, you miss one-third of the sales it actually drove. Measure sponsored search the same way, and you miss almost a quarter of its impact. But measurement works in both directions: 47% of brands should reduce or maintain their current retail media budgets. A model that only shows you what you’re missing, without showing when you’re over-invested, isn’t giving you the full picture.
This isn’t unique to Amazon. The same pattern shows up across channels. Affiliates, for example, drive 25% of their impact in physical stores rather than online. And physical stores still matter far more than the retail media conversation suggests. Around 80% of sales worldwide still happen in brick-and-mortar stores. A measurement approach that starts and ends with e-commerce conversions has a blind spot in most CPG categories.
Syndicated Data Has the Same Limit
Plenty of CPG brands already know retailer reporting is incomplete, so they turn to syndicated data from providers like Nielsen and Circana. That data is valuable. It tells you what sold, where, and when, across the whole category.
But “what sold” is not the same as “what your marketing influenced.” Syndicated data describes the result. It doesn’t explain why the result happened. Sales move for a lot of reasons at once:
- Price changes
- Promotions and trade spend
- Shelf placement and distribution
- Competitor activity
- Seasonality
All of these push the same sales number up or down while your retail media campaign is running. Syndicated data can’t pull those forces apart. It can show you the wave. It can’t tell you which part of the wave your marketing made.
Why This Is Really a Mix Problem
The core issue is that CPG brands are trying to isolate one driver, retail media, from a whole system of drivers that all move together. That’s a measurement question, not an attribution one, and it needs a method built for it.
On average, marketing is directly responsible for 10 to 50% of key business outcomes, depending on the category, and CPG is regularly in the bottom third of that range, reinforcing the need to account for additional factors. The remainder of what moves sales comes from things outside the media plan entirely — pricing, distribution, competitive dynamics, and market conditions.
The question that actually matters is the one retailer data and syndicated panels can’t answer: what would have sold without this investment? Answer that, and you know your true incremental impact. Everything else is just counting what was easy to see.
What Commercial Analytics Adds
With Marketing Mix Modeling at its core, Commercial Analytics was built for exactly this kind of problem. Instead of tracking individual conversions on a single platform, it models the entire commercial picture and determines how much each driver contributed. That includes:
- Retail media and other paid media
- Trade promotion and pricing
- Distribution and shelf presence
- Competitive activity, consumer trends, and seasonality
Because it models everything at once, it can see the halo that retailer reporting misses. Off-platform sales driven by your Amazon campaign show up in the same model as on-platform sales. And because the model weighs every driver together, it can estimate what would have sold without your retail media spend. That difference, between what actually happened and what would have happened anyway, is your true incremental impact.
What makes this work at CPG scale is trusted outside insight. Our ROI Genome® gives brands a way to compare their retail media ROI against similar brands in the same category, not just against their own past results, adding both perspective and value.
Unified data is the practical hurdle. For CPG, the challenge isn’t just pulling together retailer data and brand data — it’s accounting for the full range of commercial drivers: pricing moves, distribution changes, competitive promotions, and macroeconomic shifts that affect category volume independently of any advertising. The media buyer’s data alone doesn’t capture that picture, and neither does the retailer’s.
71% of large advertisers face at least one data barrier, with a lack of unified infrastructure topping the list at 38%. A good measurement program is built to work with the data you can realistically pull together, not the data you wish you had.
Retailer data can tell you what sold. Only a model of the full commercial mix can tell you what your retail media actually moved. The goal isn’t to stop using retailer data or syndicated panels. Both are useful inputs. The goal is to stop letting either one be the final word.
