Amazon Marketing Cloud (AMC): A 2026 Guide for Mid-Size Brands

Amazon Marketing Cloud is Amazon’s clean room for advertisers: a place where you can query event-level ad and purchase data about your own campaigns without ever seeing who the shoppers are. Until 2025 it was mostly a DSP and big-brand tool. Since September 2025, any sponsored ads advertiser can open it from the ads console, and it costs nothing to use.

We run AMC on live client instances. Most mid-size brands can get something useful out of it within a month, though not everything the conference decks promise. Here’s what it gives a brand spending £20k–£100k a month, where it falls short, and the queries we’d run first.

What is Amazon Marketing Cloud?

Amazon Marketing Cloud (AMC) is a privacy-safe clean room where your ad events and Amazon’s conversion data sit together at shopper level, and you query them with SQL or no-code templates. You never see individual shoppers. Every result comes back aggregated, and rows describing too few people are removed before you see them.

Amazon describes AMC as “a secure, privacy-safe, and cloud-based clean room solution” that “only accepts pseudonymized information”. In practice, that means one thing a normal ads report can’t do: follow the same (anonymous) shopper across several ads and see what they eventually bought.

Console reports tell you what each campaign did. AMC tells you what campaigns did together.

Question	Standard Amazon Ads reports	AMC
What did campaign X spend and sell?	Yes	Yes
Did shoppers who saw SB and SP convert better than SP alone?	No	Yes
How long from first ad to purchase?	No	Yes
Which campaigns bring in new-to-brand buyers vs repeat buyers, de-duplicated across campaigns?	Partly (per campaign, not de-duplicated)	Yes
Build an audience from people who clicked but didn’t buy	No	Yes (AMC audiences)
Individual shopper records	No	No, never

AMC is Amazon’s own Amazon clean room. It runs on AWS, but you don’t need an AWS account.

AMC Amazon access: who can use it, and is it free?

As of September 2026, AMC is free for eligible advertisers, and any advertiser running sponsored ads can open it directly from the Amazon Ads console under Measurement & Reporting. Amazon made this self-service in September 2025, including for the UK. Before that, sponsored-ads-only brands needed an agency, a partner or an account rep to request an instance, which took days or weeks.

DSP advertisers get access through an ad tech account executive, and agencies and tools on the Amazon Ads Partner Network connect by API.

What does cost money is the paid features layer: extra datasets you subscribe to per instance. Two to know about:

• Amazon Retail Purchases. Five years of Amazon purchase history for your brand’s products. At launch in May 2025 Amazon priced it at $500 a month for measurement, with a 60-day free trial, and free for audience creation.
• Amazon Audience Segment Insights. Lets you see which Amazon audience segments your buyers fall into.

In June 2026, Amazon made its own first-party paid features (including those two) free to query until 31 December 2026, according to PPC Land’s report of the announcement. Third-party data such as Experian stays paid. We couldn’t find the Amazon page for that offer, so check the Paid features tab in your own instance before you plan around it.

How AMC got here: the timeline

Date	Change
Oct 2021	Instructional Query Library launched: ready-made SQL with instructions
Oct 2024	AMC audiences usable in Sponsored Products, Sponsored Brands and Sponsored Display
Jan 2025	AI SQL generator announced at CES: plain English in, SQL out
May 2025	Amazon Retail Purchases dataset (5-year lookback, paid for measurement)
Sep 2025	Direct self-service AMC access for sponsored ads advertisers, UK included
Nov 2025	Ad traffic lookback extended from 13 to 25 months (UK from Q1 2026)
Mar 2026	Ads Agent in the AMC query editor can edit, explain and fix queries
Jun 2026	First-party paid features free to query until 31 Dec 2026 (reported)

What data does AMC hold, and for how long?

AMC holds 25 months of ad traffic data as of September 2026, extended from 13 months in November 2025, with the UK added in Q1 2026. Retail purchase history goes back five years, but only through the paid Amazon Retail Purchases dataset. Everything is stored with timestamps in UTC, not UK time.

For a sponsored-ads-only brand, three tables do most of the work:

Table	What a row is	Use it for
sponsored_ads_traffic	An SP, SB or SD impression or click, with campaign, targeting and search term	Reach, frequency, overlap between ad types, spend
amazon_attributed_events_by_traffic_time	An attributed conversion, dated by when the ad was seen or clicked	“What did my March ads eventually drive?”
amazon_attributed_events_by_conversion_time	The same data, dated by when the purchase happened	“What converted in March, and from which ads?”
conversions	A conversion event (purchase, add to basket, detail page view) for your tracked ASINs	Behaviour beyond what the ad attribution credits

DSP advertisers also get dsp_impressions, dsp_clicks and related tables.

The date range is set in the UI date picker, not in your SQL; a WHERE clause can only narrow it. Attribution follows Amazon’s standard 14-day click window for sponsored ads. AMC doesn’t change the attribution model. It lets you see across campaigns.

How do AMC’s privacy rules change what you see?

AMC removes any output row that describes too few distinct shoppers, and it does this silently. You don’t get an error. You get fewer rows, or a dimension like the search term blanked out. Amazon doesn’t publish one fixed number. The commonly cited figure is around 100 distinct users per row, and in our experience the threshold varies by column.

Intentwise, an AMC partner, describes the thresholds as tiered: some columns load with as few as two unique users, others need at least 100. The schema browser in the AMC query editor shows each column’s aggregation threshold (Low, Medium or High) when you click its info icon. Check it before building a query around a column.

It’s the most common reason AMC output looks wrong:

• A missing row isn’t a zero. It might mean nobody did it, or too few people did it to show.
• Fine-grained questions fail first. Search term × campaign × week will lose most of its rows on a mid-size account. Campaign type × month will survive.
• Blanked text isn’t NULL. Suppressed search terms and targets often come back as an empty string, so WHERE customer_search_term IS NOT NULL doesn’t filter them. Use TRIM(customer_search_term) <> ''.

The working rule we use: if a question needs more than two dimensions, widen the date range before you add the third. A 90-day window survives where a 7-day one won’t.

Do you need Amazon Marketing Cloud SQL skills?

Not to get started, but you’ll hit the limits of no-code quickly. As of September 2026, AMC offers no-code analysis templates with charts, a library of low-code instructional queries, an AI SQL generator, and Ads Agent inside the query editor. SQL is still how you ask a question nobody has written a template for.

What each option is good for:

Option	Skill needed	Good for	Watch out for
No-code templates	None	Overlap, path to conversion, NTB summaries	Fixed questions, fixed cuts
Instructional Query Library	Light SQL editing	Standard analyses with instructions on reading the output	Must adapt to your campaign names and ASINs
AI SQL generator / Ads Agent	Enough to check the output	First drafts, fixing syntax, explaining a query	Can use columns your instance doesn’t have; review before running
Hand-written SQL	Working SQL	Anything specific to your account	AMC-specific rules (below)

AMC SQL is a standard SQL dialect with privacy rules on top. The ones that catch people out, from our own query logs:

• No SELECT *. Name every column.
• user_id can be used in joins, GROUP BY and COUNT(DISTINCT ...), never in the output.
• ORDER BY and LIMIT are rejected in the instances we use. Sort in the export.
• spend in sponsored_ads_traffic is stored in micro-units. Divide by 100,000,000 and check one campaign’s total against the console before trusting it.

Below are three templates we use as starting points. They are illustrative: written against the table and column names we see in live instances as of September 2026, but schemas vary between instances and Amazon changes them. Check each column in your own schema browser first.

Query 1: Which combinations of ad types do buyers see?

A path-to-conversion question in its simplest form: of the shoppers reached by SP, SB and SD, which combinations bought?

-- Illustrative template. Set the date range in the AMC UI (e.g. last 90 days).
WITH exposure AS (
  SELECT
    user_id,
    MAX(CASE WHEN ad_product_type = 'sponsored_products' THEN 1 ELSE 0 END) AS saw_sp,
    MAX(CASE WHEN ad_product_type = 'sponsored_brands'   THEN 1 ELSE 0 END) AS saw_sb,
    MAX(CASE WHEN ad_product_type = 'sponsored_display'  THEN 1 ELSE 0 END) AS saw_sd
  FROM sponsored_ads_traffic
  WHERE impressions > 0
  GROUP BY user_id
),
buyers AS (
  SELECT
    user_id,
    SUM(total_purchases)     AS purchases,
    SUM(total_product_sales) AS sales
  FROM amazon_attributed_events_by_traffic_time
  GROUP BY user_id
)
SELECT
  e.saw_sp,
  e.saw_sb,
  e.saw_sd,
  COUNT(DISTINCT e.user_id)  AS reached_shoppers,
  COUNT(DISTINCT b.user_id)  AS buying_shoppers,
  SUM(COALESCE(b.sales, 0))  AS attributed_sales
FROM exposure e
LEFT JOIN buyers b
  ON e.user_id = b.user_id
 AND b.purchases > 0
GROUP BY e.saw_sp, e.saw_sb, e.saw_sd

Divide buying_shoppers by reached_shoppers for a conversion rate per combination. If SB + SP converts at twice the SP-only rate, that’s a case for keeping SB on terms where it looks weak alone. Correlation, not proof: shoppers who see both are often further along already.

Query 2: New-to-brand buyers by campaign type

The console shows NTB per campaign. This shows it by ad type, side by side, in one pass.

-- Illustrative template. DSP rows (if any) have a NULL ad_product_type.
SELECT
  COALESCE(ad_product_type, 'dsp')          AS ad_type,
  COUNT(DISTINCT user_id)                   AS shoppers,
  SUM(total_purchases)                      AS purchases,
  SUM(new_to_brand_total_purchases)         AS ntb_purchases,
  SUM(total_product_sales)                  AS sales,
  SUM(new_to_brand_total_product_sales)     AS ntb_sales
FROM amazon_attributed_events_by_traffic_time
GROUP BY COALESCE(ad_product_type, 'dsp')

NTB share = ntb_purchases / purchases. Add campaign to the GROUP BY once you’ve seen the totals survive the privacy filter. This is the query that most often changes how we read ACoS against TACoS: a campaign with poor ACoS and a high NTB share is buying customers, not just orders.

Query 3: How long from ad to purchase?

-- Illustrative template. Buckets the gap between the attributed ad event and purchase.
WITH lag AS (
  SELECT
    COALESCE(ad_product_type, 'dsp') AS ad_type,
    user_id,
    SECONDS_BETWEEN(
      CAST(traffic_event_dt_utc    AS TIMESTAMP),
      CAST(conversion_event_dt_utc AS TIMESTAMP)
    ) / 86400.0 AS days_to_purchase
  FROM amazon_attributed_events_by_conversion_time
  WHERE total_purchases > 0
)
SELECT
  ad_type,
  CASE
    WHEN days_to_purchase < 1 THEN 'a: same day'
    WHEN days_to_purchase < 3 THEN 'b: 1-2 days'
    WHEN days_to_purchase < 7 THEN 'c: 3-6 days'
    ELSE 'd: 7-14 days'
  END AS lag_bucket,
  COUNT(DISTINCT user_id) AS buying_shoppers
FROM lag
GROUP BY
  ad_type,
  CASE
    WHEN days_to_purchase < 1 THEN 'a: same day'
    WHEN days_to_purchase < 3 THEN 'b: 1-2 days'
    WHEN days_to_purchase < 7 THEN 'c: 3-6 days'
    ELSE 'd: 7-14 days'
  END

If a big share of purchases lands three or more days after the ad, yesterday’s numbers only show part of a campaign’s sales. That matters most for higher-priced products and daily bid rules.

What are AMC audiences, and can sponsored-ads-only brands use them?

Yes. Since October 2024, AMC audiences can be used for targeting in Sponsored Display and as bid boosts in Sponsored Products and Sponsored Brands. So you can build an audience (say, shoppers who clicked an SP ad in the last 30 days and didn’t buy) and bid up when they search again, with no DSP.

AMC offers two kinds:

• Rule-based audiences. You define the rule in SQL or a template: clicked but didn’t buy, bought product A but not product B, bought 60–90 days ago on a replenishable item.
• Lookalike audiences. Amazon models shoppers similar to a seed you choose.

Two practical limits:

1. Size. Intentwise reports that AMC only lets you create a custom audience with 2,000+ shoppers. A £20k-a-month account can hit that on broad rules and miss it on narrow ones.
2. Where they run. Audiences built from Amazon’s signals run inside Amazon’s own ad products: sponsored ads and Amazon DSP. You can’t export them to Meta or Google.

For DSP advertisers, AMC audiences are more flexible, which is why the PPC vs DSP question often comes up at the same time. If you’re weighing an Amazon DSP agency partly for AMC access, you no longer need one for that. You need one if you want DSP itself run well.

Is Amazon Marketing Cloud worth it for a £20k–£100k/month brand?

Access is free, so the real cost is time: typically a few hours a month once set up, more to start. At £20k–£100k a month in ad spend, AMC tends to pay off when you run more than one ad type and you’re making decisions the console can’t answer. If you run Sponsored Products only, it usually isn’t worth the time yet.

This is our working heuristic, not an Amazon rule:

Your situation	AMC worth it now?	What to use it for
SP only, under ~£20k/month	Usually not	Fix structure and search terms first
SP + SB, £20k–£50k/month	Yes, lightly	NTB by ad type, SB+SP overlap (Queries 1–2)
SP + SB + SD, £50k–£100k/month	Yes	Overlap, time to conversion, remarketing audiences for bid boosts
Any spend, running or testing DSP	Yes	Cross-channel paths, DSP audiences, incrementality-style tests
Replenishable product, strong repeat rate	Yes, even at lower spend	Lapsed-buyer audiences (Retail Purchases reportedly free to query until end-2026)
Fewer than a few hundred orders a month	Probably not	Most rows will be suppressed

The last row matters most. AMC’s privacy filter punishes small data. With orders in the low hundreds a month, most cuts come back empty. Your search term report is a better use of that time.

AMC also shows patterns, not causes. To prove SB caused extra SP sales you need a test (pause a campaign type for a period and compare), and AMC is where you measure it.

What should you run in your first 30 days of AMC?

Aim for one budget decision in four weeks, not a dashboard. Pick a question the console can’t answer and whose answer would change your budget split. This is the sequence we use.

1. Week 1: Open AMC in the console, run the no-code overlap and NTB templates, and check that campaign totals roughly match the console for the same dates.
2. Week 2: Run Queries 1 and 2 over 90 days. Note which rows are missing, because that tells you where your data is too thin.
3. Week 3: Run Query 3. If a large share of purchases arrives after day three, review how you judge new campaigns, and whether daily bid rules are reacting too early.
4. Week 4: Build one rule-based audience (clicked, didn’t buy, last 30 days) and apply it as a bid boost in your best SP campaign. Measure over the following four weeks.

If you’re running a SKAG structure, AMC’s targeting column lines up with your keyword-level campaigns, which makes the output far easier to act on.

---

FAQ

What is Amazon Marketing Cloud?

Amazon Marketing Cloud is Amazon’s privacy-safe clean room for advertisers. It combines your ad events with Amazon’s conversion data at an anonymous shopper level, so you can query cross-campaign questions like overlap, new-to-brand share and time to purchase. You only ever see aggregated results, and small groups of shoppers are removed from the output.

Is AMC free?

Yes, as of September 2026 AMC is free to eligible advertisers. Paid features such as Amazon Retail Purchases normally carry a subscription; Retail Purchases launched at $500 a month for measurement. Amazon’s own first-party paid features were reported as free to query until 31 December 2026. Third-party datasets remain paid.

Can I use AMC without Amazon DSP?

Yes. Since September 2025, sponsored ads advertisers can open AMC directly from the Amazon Ads console under Measurement & Reporting, with no DSP spend or partner needed. You can analyse Sponsored Products, Sponsored Brands and Sponsored Display data, and use AMC audiences for SD targeting and SP and SB bid boosts.

How far back does AMC data go?

Ad traffic data goes back 25 months as of September 2026. Amazon extended it from 13 months in November 2025, with the UK added in Q1 2026. Five years of retail purchase history is available through the paid Amazon Retail Purchases dataset.

Do I need to know SQL to use AMC?

Not to start. AMC has no-code templates, an Instructional Query Library and an AI SQL generator through Ads Agent. But anything specific to your account usually needs some SQL, even if only to adapt a template or check what the AI wrote. Review every generated query before you trust its output.

What are AMC audiences?

Custom audiences built from AMC data, either rule-based (for example, clicked but didn’t buy) or lookalike. They can be used for Sponsored Display targeting, as bid boosts in Sponsored Products and Sponsored Brands, and in Amazon DSP. Partners report a minimum audience size of around 2,000 shoppers.

---

Sami Sultan runs Ecom Enable, an Amazon growth partner for DTC brands. AMC access, pricing, lookback and audience rules checked 22 September 2026. If you’d like us to run these queries on your own account, AMC analysis is part of our advertising service, or start with a free Amazon audit.

Sources: Amazon Ads: Amazon Marketing Cloud product page; Amazon Ads: Sponsored ads advertisers can now access AMC (Sept 2025); Amazon Ads: Direct access to AMC for sponsored ads advertisers; Amazon Ads: AMC expanded ad traffic lookback window; Amazon Ads: AMC Audiences for sponsored ads; Amazon Ads: Instructional Queries in AMC; Amazon Ads: Generative AI in AMC; Amazon Ads: Enhanced Ads Agent for AMC; Amazon Ads: Amazon Retail Purchases dataset; PPC Land: AMC 1P paid features free until end of 2026; PPC Land: AMC lookback from 13 to 25 months; PPC Land: Explaining AMC; Intentwise: Amazon Marketing Cloud at scale; AdLabs: AMC reference guide.