Say a brand spends $40,000 a month on Amazon Sponsored Products and the Amazon Ads dashboard reports a 6.2x ROAS, a realistic illustrative example of the kind of gap that shows up once that number sits next to a blended MER of 2.1x calculated from total revenue and total spend across every channel. Nobody in that room asks why the gap is so wide. It is not a coincidence, and it is not a reason to be impressed with Amazon specifically. It is the same structural gap that shows up every time a platform grades its own ad spend, just with a retail media network's name on it instead of Meta's or Google's.
Retail media is the fastest-growing line item in a lot of ecommerce ad budgets right now, Amazon Ads, Walmart Connect, Instacart, Target Roundel, and a growing list of others, and almost none of the retail media attribution advice written for it comes from the same discipline ecommerce teams already apply, unevenly, to Meta and Google. Most retail media guides are written for the media buyer deciding where to allocate spend between networks. Almost none of them are written for the person who has to explain to a CFO why the dashboard ROAS and the bank statement tell two different stories.
Why retail media ROAS was never going to match your bank account
A platform-reported ROAS number, on any network, is the output of three choices that platform makes on its own: which attribution window to use, which touch gets the credit, and which sales count as caused by the ad at all. Amazon, Walmart, and every other retail media network make those three choices independently, and none of them are required to agree with each other or with your store's own revenue reporting.
This is not a retail-media-specific flaw. It is the same reason Facebook and Google Ads revenue has never matched GA4 cleanly either: each system is reporting what it saw, using the attribution rules it was built with, not auditing itself against an outside source of truth. Retail media adds one more wrinkle on top of that familiar problem. The "sale" a network attributes to your ad is not even a sale in your own revenue system most of the time. It is a sale inside that network's marketplace, visible to your own analytics only as an export you have to go pull separately.
Amazon Ads: two different "attribution" products, and a halo neither reports
Amazon Attribution and Sponsored Products attribution are not the same measurement system, and conflating them is a common way ecommerce teams misread their own Amazon reporting. Amazon's own guide to Amazon Attribution covers the separate tool used to measure non-Amazon traffic, Meta ads or email sends that point to an Amazon listing, and documents a 14-day, last-touch lookback window for that tool specifically.
Sponsored Products, the native on-Amazon ad product most retail media budgets actually run through, uses its own shorter window: a 7-day click attribution window for Seller Central (third-party) accounts and a 14-day window for Vendor Central accounts, per Feedvisor's breakdown of Amazon's attributed-sales methodology, while Sponsored Brands, Sponsored Display, and Amazon DSP all run on a 14-day window regardless of account type. For a third-party seller running Sponsored Products on a longer-consideration product, that 7-day window alone can miss a meaningful share of real purchases that land a few days later. Amazon Marketing Cloud separately lets advertisers extend a custom lookback window up to 28 days for deeper analysis, which is itself a quiet admission that the default windows were missing real purchases that happen slightly later.
The bigger gap most ecommerce teams never account for is the halo effect. Amazon's attribution reporting counts sales of the exact ASIN the ad pointed to within its window. It does not reliably surface the sales an Amazon ad drove to other products in the same catalog, a customer who clicked a Sponsored Products ad for one SKU and bought a different one from the same brand in the same session or shortly after. That spend shows up as cost with no attributed revenue in the Amazon dashboard, even though it worked, which means Amazon ROAS is simultaneously inflated by its short attribution window on some sales and deflated by an invisible halo on others. Both distortions point the same direction: the dashboard number and reality are not the same thing, in either direction, for reasons the dashboard itself cannot show you.
Walmart Connect: closed-loop is not the same as independently verified
Walmart Connect's own measurement page describes its closed-loop attribution as tying ad exposure across Walmart's digital properties and offsite channels directly to sales on Walmart's site and in its stores, using a multi-touch attribution model unified across onsite, offsite, and in-club touchpoints. That is a genuinely stronger measurement foundation than a lot of open-web advertising gets, because it is built on first-party purchase data rather than a modeled estimate.
It is still Walmart measuring Walmart's own ads against Walmart's own sales, using a model Walmart designed and controls. Closed-loop describes the plumbing, not an independent audit. A brand running Walmart Connect campaigns is in the same position as a brand running Meta ads: trusting a platform to grade its own homework, just with a tighter data pipe than most networks offer. The IAB Europe Commerce Media Measurement Standards exist specifically because advertisers kept running into this gap across the industry: the standards define gross versus net sales, a formal definition of incrementality, and new-to-brand reporting windows precisely because no two retail media networks were defining those terms the same way before the standard existed.
Fold retail media into blended MER, don't report it next to it
The fix here is not a retail-media-specific technique. It is the same reconciliation discipline that should already be applied to a blended marketing dashboard built around MER rather than platform-reported numbers: total revenue divided by total marketing spend, across every channel including retail media, is the number that cannot be gamed by any single network's attribution window. Amazon's or Walmart's dashboard ROAS belongs underneath that blended number as a diagnostic input for how to shift budget within retail media, not next to it as if the two numbers measure the same thing.
In practice that means two things most teams skip:
- Spend goes into the denominator even when the network's own report looks great. A 6.2x Amazon-reported ROAS does not change the fact that the $40,000 spent still counts as marketing spend when calculating total MER.
- Revenue only counts once. If a customer bought on Amazon, that revenue belongs to Amazon's own marketplace revenue line, not blended into your DTC store's GA4 and Shopify revenue, the same order-ID discipline that matters when reconciling Shopify against GA4 applies here: pick one system of record per revenue stream and do not let the same sale get counted by two different reports.
What to actually pull before trusting a retail media ROAS number
Before a retail media ROAS figure goes into any budget conversation, pull the network's own order-level export, not just its summary dashboard, and compare it against the marketplace's own sales reporting for the same window. For Amazon specifically, that means checking Amazon Ads reported sales against Seller Central or Vendor Central order data directly, since the two can diverge even within Amazon's own ecosystem. For Walmart Connect, the equivalent check runs against the Walmart Marketplace or supplier portal's own sales report.
Whatever gap survives that comparison is a real, explainable number, caused by a specific attribution window or halo effect you can name. A number that still will not reconcile after that check is a tracking or reporting problem worth treating the same way a marketing analytics engagement already treats a Meta-versus-GA4 mismatch: verify the event-level data each side is actually built from before assuming either platform is simply wrong.
Decide the budget call by the number that survives reconciliation
Use the retail media dashboard to decide which campaign, creative, or keyword to adjust inside Amazon or Walmart this week. Use blended MER, with retail media spend and revenue folded in correctly, to decide whether retail media deserves a bigger share of next quarter's budget at all. Those are two different decisions, and conflating them is how a brand ends up scaling a channel because its own dashboard told it to, while the bank statement quietly disagrees.
A reconciled reporting build that pulls retail media, paid social, paid search, and store revenue into one warehouse is what makes that second decision possible to make with a real number instead of a network's own marketing copy. Retail media is not going away as a line item. The question worth settling now, before next quarter's budget conversation, is which number in the room everyone has actually agreed to trust.
