Amazon Marketing Cloud explained: what UK brands can learn from it, and what it can't tell you

The ads console tells you what Amazon credited to your ads. Amazon Marketing Cloud lets you ask your own questions of the same data: which ads new customers saw before they bought, how many shoppers your campaigns reached twice, which combinations of ad types came before a purchase. It sits alongside the ad types and costs covered in our Amazon advertising UK guide, and since September 2025 any brand running sponsored ads can open it free of charge.

That makes it worth knowing. It also makes it easy to mistake more detail for more truth. This guide covers what AMC is, how UK brands get in, what it holds, and the one discipline that keeps it honest: judge every finding on incremental profit, not attributed sales.

What Amazon Marketing Cloud is

Amazon describes AMC as “a secure, privacy-safe, and cloud-based clean room solution” in which advertisers analyse and build audiences from pseudonymised signals, both Amazon Ads signals and their own¹. It is built on AWS Clean Rooms¹.

In practice your AMC instance is a private database containing event-level records of your Amazon Ads activity: impressions and clicks from sponsored ads and Amazon DSP, and the purchases and other conversions attributed to them⁴. Each record is keyed to a pseudonymised shopper identifier that can never leave the clean room⁵. You write queries against those records; AMC runs them and returns aggregated tables.

Three things make that different from the reports in the ads console:

• You choose the question. Standard reports give fixed metrics. AMC lets you query event-level and user-level signals with your own metrics and groupings¹.
• You can follow a shopper across ad types. Because records share a shopper identifier inside the clean room, you can see, in aggregate, how many people saw a Sponsored Brands ad and later clicked a Sponsored Products ad before buying.
• You can bring your own data. Advertisers can upload pseudonymised first-party records, such as hashed emails from a website’s customer list, and join them to Amazon Ads events. Amazon says uploaded data stays in your instance and cannot be accessed or exported by Amazon⁴.

AMC is not an ad platform and does not run campaigns. Audiences you build there are activated in Amazon’s own ad products¹⁵.

Access in the UK: free, and self-service for sponsored ads

On 18 September 2025 Amazon made AMC available to every advertiser running sponsored ads: Sponsored Products, Sponsored Brands, Sponsored Display or Sponsored TV. It appears in the ads console under Measurement & Reporting². The announcement covered the UK among its marketplaces², and Amazon’s product page lists the UK, says AMC “is available at no cost to eligible advertisers” and that those using sponsored ads “can access AMC via self-service"¹.

On 29 September 2026 Amazon renamed its unified ad platform Amazon Ads Agent¹³, so that console is where you will now find AMC. Don’t confuse that with Ads Agent in AMC, the AI assistant inside AMC described below.

What you get through self-service, and what you don’t:

Two practical notes. The AMC interface is in English only⁴. And if you sell in several European stores through separate advertising accounts, each instant-access instance covers one account, so you cannot join a UK and a German account in a single query that way.

What data AMC holds

When an instance is created it is back-filled with Amazon-owned advertising data, and Amazon’s developer documentation describes a rolling 13 months of event-level data⁴. In November 2025 Amazon announced that the ad traffic lookback would extend to 25 months, with the UK included from the first quarter of 2026⁹. The documentation still says 13 months, so check the date range your own instance actually returns before you plan a year-on-year analysis.

The tables that matter most for a sponsored ads advertiser:

• Traffic: impressions and clicks, by campaign, ad type, placement, device and time.
• Attributed conversions: one row per conversion credited to an ad. The pre-attributed tables use a last-touch model that prioritises product relevance, then favours clicks over views, with a lookback of up to 14 days¹⁰.
• Conversions for custom attribution: tables that let you choose your own model, such as first touch, last touch or equal weight, with a lookback of up to 28 days¹⁰.
• New-to-brand flags: a purchase is new-to-brand if the shopper has not bought from your brand in the previous 12 months, and AMC carries that flag and new-to-brand sales and units on attributed purchases¹¹.

One sentence in Amazon’s help changes how you should read all of this: the conversion tables “contain signals related to ad-attributed Amazon conversions and pixel conversions only. Non-ad attributed Amazon conversions aren’t published"¹⁰. AMC sees what happened after your ads. It does not see the shoppers who bought without one.

Privacy and aggregation thresholds: why rows go blank

AMC never lets user-level or event-level information out. Amazon Ads tables can’t be exported, and every report is aggregated so that no row represents fewer than a minimum number of shoppers⁵. The minimum depends on the columns you use. Each column has a sensitivity level⁶:

When a row falls short, AMC returns it with the sensitive values blanked rather than revealing a small group⁶. Audiences built in AMC must exceed 2,000 records and cannot be exported⁵.

For a UK brand spending a few thousand pounds a month, this is the most common frustration. Cut a path-to-purchase analysis by campaign, week and device and half the rows come back empty. Amazon’s own advice is to lengthen the date range, broaden filters and group by less granular dimensions, for example campaign rather than ad⁶. Our rule of thumb, which is our practice and not Amazon’s, is to start at the coarsest grouping that answers the question and only split further if the totals hold up.

No-code templates, instructional queries and SQL

AMC’s query language is SQL, with Amazon’s own extensions⁴. You no longer need to write it to get started:

• No-code templates. The AMC homepage has analysis templates that return visual results “with just a few clicks"². In the use case library, filter by No code to see templates that need only campaign or ASIN inputs⁷.
• Instructional queries. The use case library holds Amazon-written SQL queries, each with business context, instructions and guidance on reading the results, from path-to-conversion analyses to overlap analyses⁷. You can open any of them in the query editor, run it as it is, or change the campaigns, groups or ASINs it uses⁷.
• Ads Agent in AMC. Announced in November 2025 as a beta, available in the UK, it generates analytics SQL from plain-English requests, with knowledge of AMC’s datasets and privacy rules. You open it from “Ask AI for AMC help” in the AMC navigation⁸.
• Your own SQL. The query editor for anything bespoke³.

Our view is that AI-written SQL is a useful start, not a finished analysis. Read every query before you trust its output: check the date window, which conversion table it uses, and whether it counts purchases or shoppers. A wrong join produces a confident, wrong number.

What brands can learn from AMC

These are the analyses we find most useful for founder-led brands on sponsored ads. The use case library includes path-to-conversion, overlap and frequency analyses among its instructional queries and templates⁷; the rest can be built from the same tables.

Path to purchase. Which sequences of ad types came before a conversion: Sponsored Brands then Sponsored Products, Sponsored Products alone, and so on. The question it answers is whether your upper-funnel spend is part of the journeys that end in a sale, or running alongside them.

New-to-brand. What share of ad-attributed purchases came from shoppers who hadn’t bought from you in 12 months¹¹, by campaign and ad type. A campaign whose attributed sales are mostly repeat customers is probably harvesting demand you already had.

Overlap. How many shoppers were reached by two or more campaigns or ad types, and whether those exposed to both converted at a different rate. It is the best way to find campaigns bidding against each other for the same people.

Frequency. How many impressions it takes before conversion rates stop rising, which tells you where extra spend on the same audience stops paying.

Time to conversion. How long after the last ad interaction purchases happen. If many land after day 7, the seven-day, last-click window in sellers’ standard reports¹² is under-crediting that campaign.

Audiences. Rule-based segments, such as shoppers who viewed a product but didn’t buy in 30 days, for activation in Amazon’s ad products¹.

Every one of these describes the behaviour of people who saw your ads. None of them, on its own, tells you whether the ads changed that behaviour. That needs a comparison.

Measure incremental profit, not attributed sales

Attributed sales are sales that happened after an ad, within a window, under a set of rules. Some of them would have happened anyway: the shopper who searched your brand name, clicked your Sponsored Brands ad at the top of the page and bought, would most likely have bought from the organic result underneath. Amazon’s rules decide who gets the credit. They don’t decide what the ad caused.

The number that matters to the business is this:

Here is why the difference matters, in an illustration. A brand spends £10,000 a month on sponsored ads and Amazon attributes £40,000 of sales to them, without VAT: a ROAS of 4.0. Its average margin before advertising is 30%, so the dashboard view is £40,000 × 30% − £10,000 = £2,000 a month of profit from ads. A four-week holdout then suggests only 55% of those attributed sales were incremental: £22,000. On that basis the ads produced £22,000 × 30% − £10,000 = −£3,400 a month.

Split by campaign type, the picture usually isn’t uniform:

The decision this points to is not “stop advertising”. It is to find which spend creates sales and which buys sales you already had, and move money accordingly. That is the job AMC is good for: it tells you where to test. The test itself gives you the answer.

A holdout test you can run without a data team

The same method we describe in advertising at portfolio scale, applied to a single question.

Business Reports define Ordered Product Sales as the price of items ordered multiplied by units sold for the period¹⁴, so it captures organic and advertised sales alike, which is exactly what the test needs. Keep both sides on the same price basis: Amazon Ads reports attributed sales without VAT¹², while Business Reports sales are what the shopper paid.

Who should use AMC now

In the accounts we manage, AMC earns its time when three things are true: the brand spends enough that its analyses don’t come back blank, it runs more than one ad type, and someone will act on the findings. Below that, most of the value comes from simpler work first: a break-even ACoS for each product, a clean campaign structure and a weekly negative-keyword routine. Our guides to break-even ACoS and Amazon PPC strategy cover those.

If you meet those conditions, a sensible first month looks like this: open AMC from Measurement & Reporting, run the no-code templates for campaign performance and audiences, run the new-to-brand and overlap instructional queries for your biggest campaigns, and pick one slice of spend to test. If you also send traffic from social media or Google to Amazon, measure that separately with Amazon Attribution.

The tool is free. The time is not. Spend it on the questions whose answers would change a budget.