Why DSP and Sponsored Ads Double-Count Each Other

The mechanism, explained simply

Each Amazon ad product runs its own attribution logic, scoped only to its own impressions and clicks. DSP's reporting console has no visibility into whether the same shopper also clicked a Sponsored Products ad; Sponsored Products' console has no visibility into whether that shopper also saw a DSP display impression. Each platform independently checks: did an ad from me precede this purchase within my window? If yes, the platform claims the sale. Two platforms can both legitimately answer yes to that question about the exact same purchase.

A worked example of the inflation

Say a brand spends $10,000 on DSP in a month and its DSP dashboard reports $60,400 in attributed sales — a 6.04x ROAS, matching our own portfolio figure across 30 advertisers in July 2026. The same brand spends $8,000 on Sponsored Products and its dashboard separately reports $40,000 in attributed sales, a 5.0x ROAS. Add the two dashboards together naively and you'd report $100,400 in combined attributed sales against $18,000 in spend — a blended 5.58x ROAS that looks like a clean, additive number.

If even 15% of the Sponsored Products purchases were shoppers who'd also seen the DSP ad within its attribution window, those purchases are being counted in both totals. The true combined figure, once that overlap is removed, is lower than the naive sum — and the size of the gap is exactly the overlap percentage, which no single-channel dashboard can tell you on its own.

Why this isn't a bug in either platform

It's tempting to call this a flaw Amazon should fix, but each platform is doing exactly what it was built to do: report on its own ads, using its own attribution window, without reference to anything else. Expecting DSP's console to know about Sponsored Products activity — or vice versa — misunderstands what each tool is for. The reconciliation step was never going to live inside either single-channel dashboard; it has to happen somewhere that can see both, which on Amazon means Amazon Marketing Cloud specifically.

How to actually check your own overlap

Inside AMC, a reach-and-frequency or path query can identify, for a given set of purchases, how many shoppers were exposed to both DSP and Sponsored Products ads within their respective windows before buying. That overlap count — not an estimate, an actual pseudonymized measurement — is what lets you compute a deduplicated combined ROAS instead of the naive additive one. It's also the number that tells you whether your DSP and sponsored ads audiences are genuinely reaching different people, or mostly the same ones twice.

The common mistake, including ours

The mistake — one we've made and corrected — is presenting a combined ROAS by literally summing two channel reports in a client deck, under deadline pressure, without running the AMC reconciliation first. It produces a genuinely impressive-looking number, and it's genuinely wrong, in a way that only becomes visible once someone asks why the deduplicated figure from a proper query doesn't match the sum. We now treat any "combined DSP and sponsored ads ROAS" figure as provisional until it's been run through an overlap check, and we say so explicitly in any report where that check hasn't happened yet.

Why this matters even more once view-through is involved

The overlap problem compounds when one of the two platforms is crediting a view rather than a click. A DSP view-through conversion and a Sponsored Products click-based conversion can both legitimately claim the same purchase — one because the shopper saw the ad, the other because they clicked a different ad and bought — and because neither system requires the other kind of interaction to have happened, there's no natural check that catches the duplication before it reaches a report. This is a large part of why Amazon's own January 2026 attribution update moved toward a more relevance-judged view model in the first place — a blind view window made this exact overlap problem worse, crediting views far more liberally than a shopper's actual decision process usually justified.

What to do once you know the overlap

A high overlap isn't automatically bad news — it can mean your DSP and sponsored ads audiences are working together as intended, with DSP building awareness that sponsored ads then closes. What it does mean is that you should stop reporting the naive sum, report the deduplicated combined figure instead, and use the overlap size itself as a planning input: a very high overlap suggests DSP prospecting audiences may be too narrow, reaching people your sponsored ads were already going to convert rather than genuinely new shoppers.

A low overlap is informative in the opposite direction — it suggests DSP and sponsored ads are reaching largely distinct audiences, which is usually the healthier pattern for a prospecting-plus-search strategy, but it's also worth double-checking rather than assuming: a genuinely near-zero overlap on an account running meaningful spend in both channels can indicate the query itself missed matches, rather than that the audiences truly never intersect.