If you take only one thing from this piece, take this: the ROAS number on your Amazon Ads dashboard is, in most cases, an overstatement of the actual incremental business value your ads are producing. Sometimes a small overstatement. Often a large one.
This isn't a controversial claim inside marketing circles. It's a controversial claim outside them — because for years, ROAS has been the metric the industry uses to communicate ad performance to executives, boards, and investors. The number is comforting. The number is the number people look at. Saying that the number is structurally inflated is an unwelcome message at the QBR.
Welcome to the unwelcome message.
I'm going to walk through the five specific ways your Amazon ROAS overstates incremental value, and what to do about each. None of this is theoretical. This is what shows up when you actually pressure-test your retail media spend with rigorous incrementality testing.
Lie 1: Branded keyword inflation
The single biggest ROAS distortion in Amazon advertising is what happens on your own brand's name as a search term.
The mechanic is simple. A customer searches for your brand by name. Your sponsored result appears at the top of the page. The customer clicks it. The customer buys. The sale gets attributed to the ad. ROAS on branded queries usually looks spectacular — 8x, 12x, sometimes higher.
Almost none of that revenue is incremental.
That customer was already coming. They typed your brand name. They had explicit intent. They would have clicked your organic result, your storefront link, or your category page and bought anyway. The sponsored placement intercepted the existing intent and got attribution credit for a sale that was already going to happen.
How big is this distortion? Across the brands I've helped audit, branded ROAS typically reports 6-15x. True incremental ROAS on branded queries — measured rigorously — typically lands closer to 1-3x. The gap is the lie.
The fix isn't to stop bidding on branded queries entirely. There's some defensive value, especially when competitors are bidding on your terms. The fix is to size the spend to the real incrementality, not the reported ROAS.
Lie 2: Retargeting double-counting
Display retargeting on Amazon — showing ads to people who already viewed your product or added it to cart — is the second-biggest source of ROAS distortion in the typical retail media stack.
Mechanically: a customer views your product, doesn't immediately buy, gets retargeted with an ad over the next 24-72 hours, eventually buys. The retargeted ad gets attribution. ROAS looks great on retargeting campaigns.
The reality is that customers who add products to cart on Amazon convert at very high rates with or without retargeting. The retargeting ads are often hitting customers who would have completed the purchase regardless. The attribution credit is real. The incremental value is much smaller.
The way to test this: pause retargeting campaigns for 14 days on a subset of products. Compare conversion rates and total sales against the unpaused campaigns. The lift, when measured cleanly, is usually meaningfully smaller than the attributed revenue would suggest.
Most teams don't run this test because the dashboard makes the spend look productive. The dashboard isn't lying. The dashboard is just measuring the wrong question.
Lie 3: Halo effect mis-attribution
This one cuts both ways and frequently confuses operators.
When you run sponsored ads on Amazon — particularly Sponsored Brand and DSP — they produce a halo effect. Customers who see ads but don't click them later make purchases that are attributed to organic placements, not the ad. Some of your "organic" revenue is actually paid-driven, just not credited to the campaign.
This means your reported ad ROAS is under-counting halo effect on aggregate. But it also means that brands which interpret ad-pause-tests too literally over-credit ads when they pause and see organic rise. The reality is messy: paid and organic are entangled in ways that make clean per-channel attribution genuinely difficult.
The fix isn't a perfect attribution model. The fix is recognizing that incrementality lives at the campaign level, not the channel level — and running structured tests that measure total business lift, not channel-level attribution.
This sounds abstract. Concretely: don't ask "what's my ROAS on Sponsored Brand?" Ask "if I pause Sponsored Brand for 30 days on this product set, what happens to total sales — paid and organic combined?" The answer is the only number that tells you whether the spend is producing actual business value.
Lie 4: Last-click bias on Sponsored Brand
Sponsored Brand ads — the banner-style ads at the top of search results — get attributed for the last click in a customer's journey. The customer may have engaged with multiple touchpoints (organic results, sponsored products, A+ Content, customer reviews) before clicking the Sponsored Brand and converting.
In a typical journey, the Sponsored Brand click is the closing action. The Sponsored Product result was the discovery. The reviews were the trust-building. The pricing comparison was the decision. The Sponsored Brand was the click that happened to capture the attribution credit.
This produces a reported Sponsored Brand ROAS that overstates the format's actual incremental value, because the format is most valuable as a closer, not as a discovery mechanism. The brands that read the data carefully recognize that Sponsored Brand spending only pays back when there's a real customer journey to close — and dial back the spend when that journey isn't being created elsewhere.
Brands that don't make this distinction over-spend on Sponsored Brand, see strong reported ROAS, and never quite produce the incremental growth the spend should be creating.
Lie 5: Lifetime value is missing from the math entirely
The fifth lie is the one nobody calls a lie — because it's not really a lie of overstatement. It's a lie of omission.
Your Amazon Ads ROAS reports first-purchase revenue. It doesn't report lifetime customer value. For brands with strong repeat economics — Subscribe & Save, replenishment categories, brand-loyalty patterns — the first-purchase ROAS dramatically understates the actual value of acquiring the customer.
This means that the ad spend producing your reported 3x ROAS may actually be producing a 7x or 10x lifetime ROAS, while the ad spend producing your reported 6x ROAS on branded queries (which acquires almost no new customers) is actually producing closer to 1-2x lifetime ROAS.
The reported numbers are pointing operators in exactly the wrong direction. The brands operating from lifetime-value math instead of first-purchase ROAS are deploying their ad budgets very differently from the rest of the industry — and producing meaningfully better long-term results.
What to do about all of this
Three concrete moves. None require a data science team. All require willingness to look at uncomfortable answers.
1. Pause-test your highest-reported-ROAS campaigns. Pick the campaign your team is most reluctant to pause. Pause it for two weeks. Compare total business performance against a matched control. The result will surprise you. The brands that run this test once almost always make it standard practice afterward.
2. Build a lifetime-value-adjusted ROAS report alongside the standard one. Cohort your acquired customers by acquisition campaign. Track 90-day repeat purchase rate by source. Multiply the first-purchase ROAS by the cohort's relative LTV factor. Now you have a comparable view of which campaigns are actually building business versus generating noise.
3. Reallocate spend based on incremental contribution, not reported ROAS. This is where the savings come from. Most brands running this analysis discover they can cut 15-30% of their retail media budget without losing any incremental revenue, and reinvest the savings in the campaigns that are actually building new-customer acquisition. The math change is significant.
The takeaway
Reported ROAS is not a lie in the sense that anyone is being dishonest. It is a lie in the sense that the metric measures the wrong question, and treating the wrong-question answer as the right-question answer leads to systematically misallocated retail media spend.
The brands winning at Amazon advertising in 2026 have moved past reported ROAS as a primary metric. The brands stuck on flat or declining ad efficiency haven't.
Question the dashboard. Run the tests. Measure incrementality, not attribution. The savings — and the reallocated growth — are usually larger than anyone expects.
