Amazon Marketing Cloud for Mid-Market Brands: The First Three Queries

Amazon Marketing Cloud (AMC) is the ads clean room where a sponsored-ads brand can ask questions the campaign manager cannot answer: what share of ad-attributed purchases are actually new to the brand, whether Sponsored Products and Sponsored Brands are paying for the same shopper, and how many days pass between the ad and the order. Since September 18, 2025, Amazon has made AMC available at no cost to advertisers running Sponsored Products, Sponsored Brands, Sponsored Display, or Sponsored TV. You do not need an Amazon DSP account to open it. For a $100K–$10M brand, the useful first month is three queries on the ASINs that pay the rent — not a custom SQL program and not a DSP proposal.

This is a platform deep dive for Seller Central operators who already watch ACoS in the Ads console and TACOS across the account. AMC sits beside those numbers. When the three queries disagree with the weekly ACoS screenshot, use the longer window and fix campaign structure — SPAG structure (single-product ad groups) is where that work usually lands. When the detail page cannot answer a shopper, the work is the listing, on the Amazon operations path.

Who can open Amazon Marketing Cloud

AMC is a privacy-safe workspace. You query aggregated, pseudonymized ad and shopping events. You do not get a list of shopper names, emails, or addresses. Small result cells are suppressed, so a query can be valid and still return a blank row when the ASIN or the week is too thin. That blank is a volume limit, not proof that the feature is broken.

Amazon’s own product page lists AMC at no cost for eligible advertisers. The September 18, 2025 announcement is the access change that matters for mid-market brands: anyone running those four sponsored-ads products sees Amazon Marketing Cloud under Measurement & Reporting in the Amazon Ads console. The homepage includes no-code templates (campaign performance, audience insights, and similar starting points) plus a query editor for custom SQL. Before that date, DSP was the usual gate, and sponsored-ads-only brands mostly reached AMC through an Amazon Ads partner. Direct self-serve is the current path for sponsored ads.

1. 01

   Open Measurement & Reporting

   In the Amazon Ads console for the marketplace you actually sell in, open Measurement & Reporting and choose Amazon Marketing Cloud.

2. 02

   Accept the terms on that ads account

   The sponsored-ads flow creates an instance tied to the account. You are not filing a DSP access request first.

3. 03

   Confirm rows before you interpret anything

   Run a no-code template, or a short query that lists campaign name, the console campaign ID, and ad product type. If that returns nothing, fix the date range or the account before you debate new-to-brand share.

Three setup traps show up in the first session.

The menu item is missing. The user role cannot open Measurement & Reporting, the account has no sponsored-ads delivery, or you are in a marketplace where that ads account is not active. Switch user and marketplace before you treat AMC as enterprise-only.

The date range ends yesterday. Amazon’s instructional guidance treats event data as lagging by up to 48 hours. End the range at least two days back or the run can fail as unavailable.

The campaign ID column is the wrong one. campaign is the console name. campaign_id_string is the ID you copy from the console. campaign_id is the numeric ID the Ads API uses. For sponsored ads those two IDs are not interchangeable. Filtering campaign_id with a console ID often returns zero rows, which looks like an empty account. DSP orders can show the same value in both ID fields, so a DSP spot-check hides the mistake until it hits Sponsored Products.

ad_product_type splits families, not formats: sponsored_products, sponsored_brands, sponsored_display, and sponsored_television. There is no sponsored_brands_video value; video sits inside sponsored_brands. Impressions and clicks live on sponsored_ads_traffic. Attributed conversions live on amazon_attributed_events_by_traffic_time (click inside the range; the order can land later) or amazon_attributed_events_by_conversion_time (the order is inside the range). dsp_impressions is the DSP table — skip it if you are not on DSP.

Query one: what share of ad purchases are new to the brand

The Ads console already shows a new-to-brand flag on some Sponsored Brands and Sponsored Products reports. AMC is worth opening because you can define the lookback and cut it by ASIN and ad product in one place. The question to ask, in plain language: of purchases attributed to ads in this range, what share had no purchase from this brand in the prior 365 days?

Run it on the hero parent and its top variations, split by ad_product_type. An account-level average hides Sponsored Products harvest inside a Sponsored Brands prospecting campaign. Read the share next to margin: the same rate means different things for a repeat consumable and a one-time durable. The signal you want is a flattering ACoS on a weak acquire rate.

A hypothetical, not a client result. Say a supplement brand spends about $18,000 a month, sees Sponsored Products ACoS at 24%, and calls the account healthy. Then query one, on a 365-day new-to-brand definition, shows about 3 in 10 ad-attributed purchases on the hero ASIN are new to the brand. The other seven in ten are existing buyers the ads intercepted. A 24% ACoS can be real and still be harvest. The console is not lying about cost per attributed sale. It is silent on whether the sale needed the ad.

Pair this with TACOS. Low new-to-brand share with flat or rising TACOS means you are paying for demand the listing already owned: put branded and retargeting queries in their own bucket so a prospecting campaign cannot take credit for them. High new-to-brand share with falling TACOS and steady revenue means ads are acquiring and organic plus Alexa for Shopping (formerly Rufus) are carrying more of the repeat. Keep that spend.

Use at least four settled weeks, ending 14 days before the budget meeting. Launches, coupons, and stockouts move the rate. If the hero row is suppressed, widen to the parent or the month. Leave a blank cell blank.

Query two: where two campaigns pay for the same order

The second query asks whether two ad products touched the same shopper inside one conversion window. The practical cut for most mid-market accounts is Sponsored Products against Sponsored Brands, then Sponsored Products against Sponsored Display if you run both. You are counting purchasers (or converting clicks) who had a traffic event from more than one ad_product_type in the 14 days before the purchase.

Each campaign reports its own ACoS, so a shopper who clicks a Sponsored Brands headline and later a Sponsored Products ad can look like two efficient campaigns. The business paid twice for one order. AMC puts both events in one clean room, which is where that double pay shows up.

Same hypothetical. Say that on the same $18,000 month, a large share of Sponsored Brands purchasers also had a Sponsored Products click inside those 14 days, often on a branded query. The Brands ACoS still looks fine in its own report. The fix is an owner for the query: branded terms in one campaign, category terms in another. Pair that split with SPAG structure — single-product ad groups, one advertised product per ad group — so each ASIN's terms have one owner, with AMC as the evidence.

Two limits keep the number honest. Overlap on the same intent — both ads closing a branded query — is double pay. Overlap where video introduces the brand and the product ad closes a non-branded query can be the plan. Label each campaign’s job before you call the cell waste. And a missing “saw all three products” row is often privacy suppression, not a zero. Decide on the pairs that returned: Products × Brands, and Products × Display.

High overlap on branded queries means move budget toward non-branded queries the detail page can win, then watch TACOS the next month. Low overlap plus low new-to-brand share is an acquisition problem. Act on query one.

Query three: how many days the purchase actually takes

Default ACoS reads are impatient. A campaign report graded on a short window will punish any product the shopper researches. AMC lets you measure the gap between the traffic event and the conversion event yourself — purchases at 1 day, 7 days, 14 days, and beyond — instead of trusting whatever window the campaign type assumed.

Amazon’s playbooks tell you to wait 14 days after the period ends before you trust attributed totals. The traffic-time table can keep receiving conversions after your range; those totals move. Pulling the report the morning after a Prime Day push, or the Monday after you paused a campaign, under-counts the slow orders and makes the pause look smarter than it was.

Same hypothetical. Say the median ad-attributed purchase on the hero ASIN lands on day 9. A weekly ACoS meeting that judges last week’s test on a 7-day view is grading the test before the window closes. Campaigns that look expensive on day 7 can land inside the margin target once day-9 and day-14 orders arrive. The operating rule is simple: do not kill or scale a hero campaign on a window shorter than the day most of its orders show up.

Use the curve to set the meeting, not to build a 12-touch model. Orders that land inside 2 days can live with a weekly ACoS readout, plus the 14-day settle and the 48-hour lag. Categories where a large share lands after day 7 — furniture, researched supplements, higher-priced hardlines — need a 14-day or 28-day decision. A click that converts on a sibling variation is a failed ASIN in the console and a successful family in AMC if you include brand halo. Judge the hero alone or as a parent, and keep that choice fixed across meetings.

How to read the three results in one meeting

Bring one page, not a SQL log. Four lines are enough for the operators in the room.

After the first clean run, freeze structural changes until the next 14-day window closes, or the re-read will not compare. Give each live campaign one job — acquire non-branded, close branded, or re-engage — because overlap only indicts two campaigns doing the same job. Move budget in one direction that month, then re-pull the same three queries. You want new-to-brand share up, same-intent overlap down, and a day-to-purchase curve stable enough to trust the ACoS you manage.

Audience builds (new-to-brand expanders, overlap exclusions) wait until those three numbers repeat. Activating a lookalike off a noisy first pull is how a brand targets its own buyers and calls it prospecting.

AMC also will not tell you whether the detail page deserved the click. A high new-to-brand share that dies on the product page is a listing problem. Keep listing optimization for Alexa for Shopping open beside the queries: AMC says who arrived; the attribute set decides whether the shopper, and Alexa for Shopping, can finish.

When another query is the wrong next step

Open AMC once the account has something to count. Two weeks of spend, a suppressed hero row, and empty attributes produce a clean room full of blanks — and a temptation to buy DSP so the blanks fill in. DSP adds dsp_impressions. It does not invent purchase history, and it does not complete backend fields.

Settle these before a fourth query:

• No settled spend. Wait until you have a month you can end 14 days in the past. A launch week is a story, not a baseline.
• The listing cannot answer. Material, size, compatibility, and count live in attributes, not in a headline. If those fields are empty, Alexa for Shopping and the filter rail have nothing to quote. Spend the week on the catalog. Catalog enrichment is the SKU for that gap; AMC will still be there when the attributes are.
• Browse node or variation family is wrong. Overlap and new-to-brand both get muddy when three products share a parent that should have been split, or when the ASIN sits in a node shoppers do not use. Correct the family first so the query is about demand, not about a bad parent.
• You wanted Search Query Performance. SQP is share of marketplace demand on a query, organic and paid. AMC is the path of your ad events. They answer different meetings. Do not export one and relabel it as the other.

The September 2025 access change made the clean room ordinary. It is still not the first dollar. A measured harvest on a thin listing is an efficient way to rent demand the catalog could have earned. Use the three queries to see whether ads are acquiring, duplicating, or being judged too early — then move budget, or attributes, to match that read.