Amazon Marketing Cloud for Attribution: What You Can Actually Answer

Questions AMC can answer directly

Within Amazon's own ad ecosystem, AMC can genuinely answer: what sequence of touches preceded a purchase across DSP, sponsored ads and streaming; how much audience overlap exists between two campaigns or ad types; what share of purchases involved more than one Amazon ad product; and, with a properly designed query, what the incremental lift looks like between an exposed group and a matched or randomised control. These are all built from signals Amazon already holds — impressions, clicks, purchases, browse events — pseudonymized and joined inside the clean room.

The boundary: off-Amazon conversions

Here's the limit worth being precise about, because it's the one most explainers gloss over. AMC's native signals cover activity on Amazon — Amazon's own ad exposure and Amazon's own purchase events. It has no native visibility into a conversion that happens entirely off Amazon: a sale completed on your own Shopify checkout, for instance, generates no signal inside AMC unless you deliberately upload it. If your DSP strategy includes link-out campaigns sending traffic to your own site, AMC alone will not tell you whether those visits converted there — that data has to come from your own site's analytics and be brought into AMC as first-party data if you want it joined against Amazon-side exposure.

What uploading your own data does and doesn't fix

You can bring first-party data into AMC — a CRM list, hashed emails, or in principle a conversion event feed from your own site — and query it alongside Amazon's signals. That closes part of the gap: it lets you check whether a shopper exposed to a DSP ad later shows up in your own converted-customer list, for example. What it doesn't do is give AMC native, automatic visibility into every off-Amazon conversion the way it has for on-Amazon purchases — the upload has to be built, maintained, and matched correctly, and a bad match on the join key will silently shrink the result without an error message telling you why.

This also isn't a real-time process. Uploading and matching first-party data is a periodic exercise, not a live feed most brands run continuously, which means a query answering an off-Amazon question is only ever as current as the last upload — worth checking before treating a result as reflecting last week's activity rather than last month's.

A worked illustration of the boundary

Say a brand runs a DSP link-out campaign sending 20,000 clicks a month to its own Shopify site. AMC can tell you, precisely, how many shoppers saw or clicked that campaign, and whether any of them subsequently purchased on Amazon within the standard attribution window. It cannot tell you, on its own, how many of those 20,000 clicks converted on the Shopify site itself — that number lives entirely in Shopify's own analytics or a separate conversion pixel, and reconciling the two into one full-funnel view requires deliberately joining both data sources, not querying AMC harder.

Questions AMC cannot answer, full stop

AMC cannot tell you whether a competitor's advertising influenced your sales — it has no visibility outside your own account's signals. It cannot measure brand perception or awareness directly; that's what Amazon's separate Brand Lift product is for, using survey-based comparison between exposed and control audiences rather than transaction data. And it cannot, on its own, prove incrementality just by running a standard attributed-sales query — that requires a specifically designed holdout or matched-control query, not the default reporting templates.

The common mistake, including ours

The mistake is assuming AMC is a universal measurement layer that sees everything relevant to a purchase decision, when it's actually scoped tightly to Amazon's own signals plus whatever first-party data you deliberately bring in. We've fielded a client question — reasonably — asking why their AMC query couldn't explain a conversion spike that coincided with a Shopify promotion running at the same time; the honest answer was that AMC had no visibility into that promotion at all, because it never touched Amazon's ad signals. The fix wasn't a different query — it was recognising the question needed off-Amazon data AMC was never going to have.

Where the boundary sits relative to the rest of this measurement territory

This limitation is exactly why the link-out and off-Amazon destination pages elsewhere on this site emphasise pairing Amazon-side measurement with your own site's analytics rather than expecting one tool to cover both. AMC is genuinely the strongest tool available for understanding what happens across Amazon's own ad products — DSP, sponsored ads, streaming — together. It was never built to be, and isn't, a substitute for your own web analytics on traffic that leaves Amazon and converts somewhere else entirely.

How to work within the boundary properly

Before building an AMC query, be explicit about whether the question is answerable from Amazon-side signals alone, or needs off-Amazon data brought in. For the second case, build the first-party upload pipeline deliberately — with a clear, tested join key — rather than assuming AMC will somehow reconcile a Shopify conversion it was never fed. And for anything genuinely off Amazon, pair AMC's on-Amazon view with your own site analytics rather than expecting one tool to answer a question that spans two separate systems.