Retail media networks sold you on a beautiful promise: closed-loop attribution. Every impression, every click, every conversion traced to a SKU. For the first time, you could see exactly which ad dollar drove which sale. That promise is real — but it's also incomplete in a way that quietly costs brands money.
The missing piece is incrementality. Closed-loop attribution tells you what happened after someone saw your ad. It does not tell you what would have happened without the ad. If a shopper was going to buy your product anyway because it's the #1 organic result, the Sponsored Products click didn't generate a sale — it captured one you already had. Your dashboard credits the ad. Your P&L disagrees.
US retail media ad spend surpassed $60 billion in 2025 and is approaching $70 billion in 2026. Amazon Sponsored Products clicks grew 23% year-over-year in Q4 2025. Amazon DSP spend grew 31%. With that kind of money flowing, the difference between attributed revenue and incremental revenue is not an academic debate — it's the line between profit and loss.
1. Attribution vs incrementality: the distinction that matters
Attribution answers "who touched the sale?" Incrementality answers "who caused the sale?" A shopper sees your Sponsored Product, clicks, and buys. Attribution credits the ad. But if that shopper was already searching for your brand name, the click was a toll, not an investment.
The retail media industry is moving toward incrementality measurement because platform-reported ROAS (sometimes called "iROAS" when it tries to sound more sophisticated) systematically overstates ad impact. Albertsons Media Collective, the first major retailer to offer causal incrementality measurement for in-store digital media, found a $2.41 iROAS and 14% sales lift on a Mondelez beta campaign. That's a real number — measured against a control group, not against a dashboard credit.
The gap between attributed ROAS and incremental ROAS can be 40-60%. If your dashboard says 4x ROAS but the incremental lift is 2x, you're overinvesting by a factor of two. That's not a rounding error. That's a margin leak.
2. Why platform-reported ROAS inflates
Platform attribution models have a structural bias: they credit the last click. If a shopper sees your display ad on Monday, searches organically on Wednesday, clicks a Sponsored Product on Friday, and converts — the Sponsored Product gets 100% of the credit. The display ad gets nothing. The organic search gets nothing. But all three played a role.
This creates two distortions simultaneously:
- Over-credit to lower-funnel ads: Sponsored Products and retargeting look more effective than they are because they capture demand created by upper-funnel activity
- Under-credit to upper-funnel ads: display, video, and awareness campaigns look ineffective because they don't get last-click credit, even when they generate the demand that lower-funnel ads convert
The result: brands pour budget into Sponsored Products (which looks great in the dashboard) while defunding the very channels that create the demand those Sponsored Products capture. Over time, Sponsored Products ROAS degrades — not because the campaigns got worse, but because the demand pipeline feeding them dried up.
3. How to measure incrementality
There are three established methods for getting closer to true incremental impact:
| Method | How it works | Best for |
|---|---|---|
| Geo-holdout tests | Run ads in one set of markets, hold back in a matched control set. Compare sales lift. | Brand-level, campaign-level incrementality |
| Audience holdouts | Withhold ads from a randomized subset of users within the same market. Compare conversion rates. | Sponsored Products, DSP, retargeting |
| Media mix modeling (MMM) | Statistical model that decomposes total sales into contributions from each channel + base demand. | Cross-channel budget allocation |
Geo-holdouts are the gold standard for retail media because they measure what actually changed. Amazon's Marketing Cloud (AMC) supports audience-level experimentation. Walmart Connect offers geo-split testing. The challenge is duration: a clean holdout test needs 4-8 weeks to reach statistical significance, and many brands can't stomach the "lost" revenue from the control group during that window.
But the alternative is worse: spending $70 billion globally on retail media and not knowing which dollars are creating demand versus capturing it.
4. The full-funnel retail media framework
The most effective retail media programs in 2026 are full-funnel, not bottom-funnel. Here's how the pieces fit:
| Funnel stage | Format | Role | What to measure |
|---|---|---|---|
| Upper | DSP display, streaming video, Sponsored Brands video | Generate awareness and demand | Incremental reach, brand search lift |
| Mid | Offsite display, audience targeting, retargeting | Nurture consideration | View-through conversions, assisted conversions |
| Lower | Sponsored Products, Sponsored Display | Capture ready-to-buy shoppers | Attributed ROAS (but read alongside TACoS) |
Amazon DSP now accounts for 40% of total Amazon ad budgets for advertisers active in both Ad Console and DSP — a sign that brands are investing more in upper-funnel and audience-driven retail media. Prime Video ad spend grew 127% year-over-year. Streaming video ad spend outside YouTube rose 13% in Q4 2025. The shift is happening because brands are learning that bottom-funnel-only strategies eventually run out of demand to capture.
5. Reading TACoS alongside attribution
TACoS (Total Advertising Cost of Sale) is the single most honest metric in retail media because it measures ad spend against total product revenue, not just ad-attributed revenue. If TACoS is rising while ROAS stays flat, your ads are capturing more of your existing demand without creating new demand. That's the early warning sign of attribution inflation.
A healthy retail media program shows:
- ROAS stable or rising (ads are efficient at capturing demand)
- TACoS stable or falling (ads are growing total revenue, not just capturing it)
- Organic sales stable or growing (ads aren't cannibalizing organic demand)
- Stock cover adequate (ads aren't driving out-of-stock that kills organic ranking)
If ROAS is green but TACoS is red and organic is declining, your ads are eating your own base. The dashboard looks great. The business is slowly bleeding.
6. The contribution margin test
Attribution tells you which campaigns "worked." Incrementality tells you which campaigns created value. Contribution margin tells you whether that value exceeds the cost of capturing it.
For every retail media campaign, run this test monthly:
- Incremental revenue: attributable sales × incremental lift percentage (from holdout or MMM)
- Minus ad spend
- Minus operational costs: commission, fulfillment, returns, COGS on incremental units
- = Incremental contribution margin
If that number is negative, the campaign is destroying value even if platform ROAS is 4x. If it's positive, the campaign is genuinely creating profit — not just shifting it from organic to paid.
7. How FiveX helps
FiveX connects retail media attribution data with your operational costs — commission, fulfillment, returns, COGS, stock — in one dashboard. You can see attributed ROAS alongside TACoS, organic sales trends, stock cover, and contribution margin per SKU per campaign. Instead of trusting platform dashboards at face value, you get the full picture: what the ad platform says happened, what your P&L says happened, and the gap between them.