Amazon Marketing Cloud

Amazon Marketing Cloud (AMC) is the most misunderstood product in the Amazon advertising stack — pitched as a magic attribution machine, dismissed as an enterprise toy, and rarely explained in plain terms. Here is the plain term: AMC is a clean room where Amazon lets you run SQL over event-level ad data — every impression, click, and conversion your campaigns generated — under privacy rules that only return aggregated answers. That’s it. No dashboards by default, no recommendations, just a query engine over data granular enough to answer questions the ad console structurally cannot. This guide covers what AMC actually is, the questions it uniquely answers, what you need before an instance is worth having, real example queries translated into plain language, and an honest assessment of who needs it — a small subset of the sellers we talk to through our Amazon PPC management work.

What AMC Actually Is: Clean-Room SQL Over Ad Events

Every report you normally see in Campaign Manager is pre-aggregated: Amazon decides the rows (campaigns, keywords, placements) and hands you totals. AMC inverts that. Your instance receives pseudonymized, event-level tables — one row per impression, per click, per attributed conversion — spanning Sponsored Products, Sponsored Brands, Sponsored Display, and, critically, DSP.

The “clean room” part is the privacy architecture. You can join and query these tables however you like, but you can never extract an individual shopper: queries only return results for aggregated groups above a minimum user threshold (generally 100 users; below that, the row is suppressed). You get cohort truth without user-level data leaving Amazon’s walls.

The interface is SQL. You write queries against tables with names like sponsored_ads_traffic, dsp_impressions, and amazon_attributed_events_by_traffic_time, and results come back as tables you export or pipe into your BI stack. Amazon maintains a library of instructional queries covering common questions, so you’re not starting from a blank editor — but someone still has to understand joins, attribution windows, and what the event timestamps mean. Data typically covers a rolling 13-plus months, and standard instances update on a daily-batch cadence rather than real time.

The mental model that keeps expectations honest: the console tells you what each campaign did; AMC tells you what your ads did together. Standard attribution assigns each sale to a single last touch. AMC preserves every touch, which is the whole game.

The Questions Only AMC Can Answer

If a question can be answered from campaign reports, search term reports, or Brand Analytics, AMC is the wrong tool — slower and more expensive in labor. AMC earns its keep on the questions where last-touch reporting is structurally blind:

Path to purchase. What sequence of ad exposures do converters actually travel? A typical finding: shoppers exposed to a DSP display ad and a Sponsored Products ad convert at 2 to 4x the rate of shoppers who saw either alone. The console can’t show this because each ad type reports its own attributed sales in isolation — the overlap group doesn’t exist anywhere in standard reporting.

DSP and search synergy. Related but decision-grade: of the sales your Sponsored Products campaigns claimed, how many purchasers had already been touched by DSP? If 30% of your “search-attributed” conversions saw display first, your DSP is being systematically under-credited by last-touch logic — and cutting its budget on console ROAS numbers would quietly gut your search performance a few weeks later. This is the single most common budget-allocation error AMC corrects.

New-to-brand overlap by channel. Which channel combinations actually recruit new customers versus re-converting your existing base? Sponsored Brands might show a 60% new-to-brand rate alone, but AMC can show whether those NTB purchasers were really cold or had been warmed by weeks of display frequency first — which changes what “prospecting” budget means.

Frequency and time-lag distributions. How many impressions does a converter see before buying, and over how many days? If the median converter needs 5 touches over 12 days, a 3-day DSP retargeting window is quitting early. If conversions cluster within 24 hours of first touch, long-window retargeting spend is decoration.

Custom audience building. Beyond analytics, AMC audiences let you define cohorts by query — “saw my streaming TV ad 3+ times, never purchased, no exposure in 14 days” — and push them to DSP as targetable audiences. This closes the loop from insight to activation and is arguably AMC’s most operational feature.

Instance Requirements: What You Need Before It’s Worth Having

The gating items are practical rather than financial:

• An Amazon Ads account and advertiser relationship. Instances are provisioned through Amazon Ads (historically via a DSP seat or an ads API partner); agencies and tool vendors commonly provision and manage instances for brands.
• The instance itself is free. Amazon doesn’t charge for a standard AMC instance or its queries. Optional paid features (extended lookbacks, some enrichments) exist, but the base capability costs nothing from Amazon.
• DSP activity — practically, if not technically. Sponsored ads data flows in regardless, but without DSP events the cross-channel questions above mostly evaporate, and what remains is largely answerable from free reports. AMC without DSP is a SQL engine pointed at questions you could already answer.
• Someone who writes SQL — or a tool that does. This is the real cost. Options are an in-house analyst, an agency, or the growing layer of platforms (Pacvue among them) that ship packaged AMC dashboards and query templates so you consume answers instead of writing joins.
• Enough volume for the math to survive privacy thresholds. Aggregation minimums mean thin campaigns produce suppressed rows and unstable percentages. This is one reason AMC skews toward larger advertisers independent of cost.

Example Queries in Plain Language

You don’t need to read SQL to understand what an AMC query does. Four we run regularly, translated:

1. The overlap query: “Split everyone who purchased in the last 90 days into three groups — touched only by search ads, only by DSP, or by both — and show me conversion rate and sales per group.” This one query has reallocated more DSP budget than any other analysis we run; the both-groups conversion multiple is the headline number.

2. The path query: “For converters touched by multiple ad types, what order did the touches happen in?” Typical output: display → search → purchase dominates, confirming display’s role as the opener that search closes — and justifying upper-funnel spend the console calls unprofitable.

3. The time-lag query: “Show the distribution of days between first ad impression and purchase.” If 70% of conversions land within 7 days, your campaigns can be judged on weekly windows; if the tail runs 3 weeks, your recent-performance panic is measuring incomplete data.

4. The audience query: “Build an audience of shoppers who clicked a Sponsored Products ad for my hero ASIN in the last 30 days but didn’t buy, and send it to DSP for retargeting.” Insight becomes activation in one step — cart abandoners and consideration-stage shoppers targeted with precision the standard DSP audiences can’t match.

5. The frequency-cap query: “Bucket DSP-exposed shoppers by how many impressions they saw — 1 to 3, 4 to 10, 11 plus — and show conversion rate per bucket.” The curve almost always flattens somewhere, and everything past the flattening point is wasted spend. This is how you set a DSP frequency cap from evidence instead of the default, and on awareness-heavy campaigns it routinely recovers 10 to 20% of display budget.

None of these require genius SQL — most are adapted from Amazon’s instructional library — but all of them require someone to sanity-check attribution windows, dedupe logic, and suppressed rows before the numbers reach a budget meeting.

AMC vs. Brand Analytics vs. Standard Reports

Sellers often conflate AMC with the other analytics surfaces Amazon offers, so a quick boundary map:

The three don’t compete — they stack. Brand Analytics tells you which queries matter, campaign reports tell you what each campaign did about them, and AMC tells you how the channels combined on the way to each sale. If you haven’t exhausted the first two, you have no business paying analyst hours for the third.

Who Actually Needs AMC

Honest triage, because AMC is the definition of a tool that’s transformative for some accounts and pure overhead for others:

You need AMC if you’re running DSP (or streaming TV) alongside sponsored ads at meaningful combined spend — call it $30K to $50K+ per month — and you’re making real budget-allocation decisions between channels. At that point, last-touch console numbers are actively lying to you about DSP’s contribution, and AMC is the only instrument that corrects the lie. It’s also how you move the conversation from ACoS to incrementality — the ACoS versus TACoS framing taken to its logical end.

You don’t need AMC yet if you’re running Sponsored Products with some Sponsored Brands and no DSP. Every decision you face — keywords, bids, negatives, placements, budgets — is fully served by search term reports, placement reports, and Brand Analytics, and the highest-return use of analyst hours is mining those, not writing clean-room SQL. Adopting AMC before DSP is buying a telescope to read a book.

The in-between case — DSP-curious brands at $20K to $40K monthly search spend — should treat AMC as part of the DSP decision itself: if you test DSP, instrument it with AMC from day one so the incrementality question is answerable at week 12 instead of being argued from console ROAS forever.

If you’re in that first or third group, the practical path is rarely hiring a clean-room analyst — it’s working with a team that already runs AMC instances across accounts and knows which twenty queries matter. That’s part of how we run DSP-inclusive engagements in our Amazon PPC advertising service, and a free audit is the fastest way to find out whether your account is leaving cross-channel money on the table.