Retail media is the fastest-growing ad category in the world. It is also the most under-audited. The combination is a problem, and it's about to get more expensive.
I work in marketing leadership. I see what brands actually spend on Amazon Ads, Walmart Connect, Roundel, and Instacart. I see the ROAS dashboards, the QBR slides, and the agency decks. And I'm telling you — gently, but I'm telling you — a meaningful portion of the dollars flowing into retail media right now are not buying growth. They are buying the appearance of growth. Sales that would have happened anyway, dressed up in the language of attribution.
This is not a critique of retail media as a channel. The channels are real. The opportunity is real. But the way most brands measure their spend is broken in ways that compound, and the brands that fix it inside the next twelve months are about to outperform their competitors on the same total budget.
Time to talk about it.
ROAS is not incrementality. They are not even cousins.
Return on ad spend tells you the ratio of attributed sales to ad dollars. It says nothing — nothing — about whether those sales were caused by the ad.
Take the cleanest example. You bid on your own brand name. A customer Googles your brand, intends to buy your product, sees your sponsored result, clicks, buys. The sale gets attributed to the ad. Your branded ROAS looks fantastic. Your incrementality is approximately zero. That customer was already coming. The ad was a tax on a sale you'd already won.
Now scale that example up. Sponsored Product ads on your own listing. Sponsored Brand ads where your organic ranking is already #1. Display retargeting on customers who've already added the item to cart. All of these produce great-looking ROAS reports. Most produce zero incremental revenue.
The dashboard isn't lying. The dashboard is measuring the wrong question.
The four-bucket audit nobody runs
Every dollar in your retail media budget falls into one of four buckets. Most teams cannot tell you the percentage in each. That's the audit.
1. Defensive — protecting sales you already have. Branded keywords, conquest defense, listing protection. Useful in moderation. Dangerous in excess. A reasonable allocation is the cost of doing business in a competitive marketplace. An unreasonable allocation is paying a tax to your retailer for shoppers who were already going to find you.
2. Offensive — winning sales you would not have won. Non-branded category-level keywords, top-of-funnel placements, new-to-brand customer acquisition. This is where actual growth lives. It is also the smallest, hardest, and most-underfunded bucket in nearly every brand's budget.
3. Tactical — moving inventory or hitting short-term goals. Promo amplification, deal-day pushes, end-of-quarter velocity. Legitimate when used surgically. Toxic when it becomes a permanent line item in the run-rate.
4. Wasted — spend producing no measurable outcome. Display retargeting on people who already converted. Sponsored Brand bids on hyper-branded queries. Audience expansions chasing irrelevant intent signals. Every brand has some of this. Most have more than they think.
If your team has never produced this view of the budget, that's the finding. The four-bucket cut is the conversation that should be happening every quarter and almost never is.
How to actually test for incrementality
You don't need a data science team. You need the willingness to run small structured tests and look at the answers honestly.
Pause and watch. Pick a campaign. Pause it for two weeks. Compare organic sales for the affected ASINs against a matched control group. If organic sales rise by something close to the paused campaign's attributed sales, the ad was largely non-incremental. The number of brands that refuse to run this test because "we can't lose the revenue" tells you exactly how dependent they've become on attribution they have never validated.
Geo splits. Where the platform allows, run a campaign in some regions and not others. Compare total sales lift, not attributed sales. The delta is your incremental contribution. The gap between attributed and incremental is your waste.
New-to-brand isolation. If your platform reports new-to-brand sales separately, you can build a credible incrementality picture without geo or pause tests. Your true growth rate is the new-to-brand line, not the total attributed line.
None of these are exotic. All of them get routinely skipped in favor of a dashboard with big numbers that feel reassuring.
The branded keyword tax
The biggest single line of waste in most retail media budgets is over-spending on branded queries.
The argument for it is "we have to defend." Sure — to a point. But that spend should be calibrated to actual competitive pressure on those queries, not to whatever the platform's recommendation engine suggests. Auto-suggested branded budgets exist to grow the platform's revenue. Not yours.
A useful exercise for any marketer reading this: take your top branded queries. For each one, ask honestly what would happen if you paused the bid for thirty days. If your organic listing is strong and the competitive set is light, the answer is "almost nothing." That is your tax bill. You are now in a position to negotiate it down.
Three actions for this week
If you do nothing else with this article:
Build the four-bucket view of your retail media budget. Defensive, offensive, tactical, wasted. Honest percentages, not aspirational ones.
Pause the campaign your team is most defensive about for two weeks. Watch organic. The campaign your team most resists testing is almost always the one most worth testing.
Ask your retail media partner for an incrementality report. If the report they hand you is just a re-skinned ROAS deck, you've just learned something important about how your budget is being managed.
The takeaway
Retail media is not bad. Wasted retail media is bad. The brands winning the next decade will treat retail media spend the way a serious finance team treats every large operating expense: skeptical by default, audited continuously, and held to a standard higher than "the dashboard looked good in QBR."
Stop optimizing for ROAS. Start optimizing for the only metric that pays for the business: incremental revenue per dollar of spend. The number is almost always lower than the dashboard shows. That gap is where your waste lives. It is also where your next phase of growth comes from, the moment you stop funding it.
