Amazon DSP reporting can tell an advertiser what a campaign delivered. Amazon Marketing Cloud can help explain how that delivery worked across audiences, formats and customer journeys. It gives analysts a privacy-safe environment for asking questions that standard campaign dashboards cannot answer in the same depth.
The distinction matters because a campaign can produce a strong headline return while hiding duplication, over-frequency or dependence on shoppers who were already close to purchase. A useful measurement plan needs to separate delivery from diagnosis, attribution from incrementality, and correlation from evidence strong enough to change the next media decision.
There is also a current product change to understand. Amazon says Amazon DSP has evolved into Amazon Ads Agent, with advanced programmatic capabilities available through Display, Video, and Audio campaigns and DVA+ advanced settings. This guide keeps “Amazon DSP” because it remains the familiar market and search term, while explaining measurement in the current Amazon Ads environment.
Quick answer: Amazon Marketing Cloud (AMC) is Amazon Ads' privacy-safe clean room for analyzing pseudonymized Amazon Ads signals together with approved advertiser and third-party inputs. It can support path-to-conversion analysis, audience overlap, reach and frequency, new-to-brand analysis, first-party conversion measurement and custom audience creation. AMC provides aggregated, anonymous outputs. It does not automatically prove incrementality, replace a measurement design or make unlike datasets comparable.
What is Amazon Marketing Cloud?
Amazon Marketing Cloud is a cloud-based clean room built on AWS Clean Rooms. Amazon describes it as an environment where eligible advertisers can analyze pseudonymized signals from Amazon Ads, their own approved inputs and onboarded third-party providers.
AMC is not a replacement for every Amazon Ads report. It is the analysis layer used when the business question requires custom logic, cross-event sequencing, controlled audience definitions or a view beyond one campaign table. Analysts can work with SQL, use curated templates, use APIs or, where available, ask Ads Agent to help generate queries through natural language.
Amazon says AMC is available at no cost to eligible advertisers, while separate Paid Features can expand available signals or analytical depth. Eligibility, market coverage, account structure, signal availability and paid subscriptions should be confirmed for the advertiser before the measurement roadmap is approved.
Verify in the account before analysis
Last verified: October 2026. Source accessed October 16, 2026: Amazon Ads' official Amazon Marketing Cloud product page. Product documentation describes platform capabilities; the advertiser's own AMC instance remains the source of truth for eligible signals, coverage and features.
Check | What the public source says | What to verify in the AMC account |
|---|---|---|
Lookback | Amazon's current product page highlights up to 25 months of ad-traffic lookback. | Confirm the earliest usable event date, conversion-history coverage and whether the same period exists across every required signal. |
Data availability | AMC can analyze eligible Amazon Ads signals and approved advertiser or third-party inputs. | Confirm the actual tables, schemas, events, advertiser uploads, paid features and identity coverage available to this instance. |
Market and account eligibility | AMC is available to eligible advertisers; specific capabilities may depend on product, market and setup. | Confirm the AMC instance, linked advertiser accounts, locale, campaign types, templates, audience activation and study eligibility before promising a deliverable. |
Privacy and output thresholds | AMC returns aggregated, anonymous outputs subject to privacy controls. | Run a small validation query to identify suppression behavior, minimum aggregation requirements and whether the planned segmentation will return usable results. |
Record this verification in the project brief with the person who checked it, the date, the account or instance in scope and any limitations that could change the interpretation.
Standard Amazon Ads reporting vs. AMC
Measurement need | Standard reporting | Amazon Marketing Cloud |
|---|---|---|
Campaign delivery | Impressions, clicks, spend, video delivery and reported conversions by available dimensions | Can combine event-level logic into custom aggregated analyses |
Path analysis | Usually summarized by campaign, line item, creative or audience | Can examine eligible exposure and conversion sequences across touchpoints |
Audience overlap | Limited to the available report structure | Can quantify overlap among eligible campaigns, formats and audience groups |
Reach and frequency | Platform-defined reach and frequency outputs | Can create custom frequency bands and relate them to eligible outcomes |
First-party signals | Used where the ad product supports them | Can analyze approved pseudonymized advertiser inputs with Amazon Ads signals |
Audience activation | Select existing audiences in campaign workflows | Build custom audiences from eligible signals and activate them in Amazon Ads |
Output | Campaign and account reports | Aggregated, anonymous query results subject to privacy controls |
The two layers should work together. Standard reporting supports daily pacing and optimization. AMC supports deeper questions that need a custom analytical design.
Attribution, path analysis and incrementality are different jobs
These terms should not be used interchangeably. AMC can support all three types of work when the required signals and methods are available, but an observed journey or attributed conversion is not automatically a causal result.
Method | Question | What the output means | What it does not prove |
|---|---|---|---|
Attribution | Which eligible touchpoint receives credit under a defined rule or model? | Credit assigned within the selected population, window and attribution logic | That the credited touchpoint caused the outcome |
Path analysis | Which eligible exposure sequences appeared before an outcome? | Observed order, co-occurrence and association among available touchpoints | Causal credit, or that the most common path is the best path |
Incrementality | What happened because of advertising compared with what would have happened without it? | A causal estimate only when supported by a valid counterfactual, such as a properly designed holdout or experiment | Nothing causal from a descriptive AMC query alone |
Decision rule: use attribution to apply a consistent crediting framework, path analysis to find sequencing and overlap hypotheses, and incrementality testing when the business needs a causal answer.
How Amazon DSP measurement works in AMC
A useful AMC project moves through five controlled stages. The worked example below starts with a real budget decision: should the advertiser reduce retargeting and test more Streaming TV prospecting?
Business question. Write the decision and its threshold before opening a query. Example: determine whether retargeting frequency is concentrated above the agreed band and whether Streaming TV reaches eligible users not already exposed to retargeting. The output will decide whether to move a defined test budget, not declare a winner automatically.
Tables and signals. Confirm the eligible impression and conversion events, campaign and ad-product mappings, audience definitions and any approved first-party customer-stage signal. Record the available date range, matching limitations and privacy thresholds.
SQL or template. Use an eligible reach-and-frequency or path template, or reviewed custom SQL, to create frequency bands, exposure groups and outcome summaries. Parameterize the campaign dates, lookback window, qualifying events and conversion definition so the analysis can be reproduced.
QA. Check source and output counts, date coverage, duplicate handling, nulls, join rates, campaign mappings and suppressed rows. Reconcile totals with standard Amazon Ads reporting wherever the definitions should align; investigate rather than hide a mismatch.
Decision. If the descriptive result shows concentrated frequency and complementary Streaming TV reach, launch a controlled budget or audience test with an owner, hypothesis, guardrail and review date. The AMC result prioritizes the test; it does not by itself prove that the budget move will create incremental conversions.
The clean room does not expose a list of identifiable users. Amazon states that AMC accepts pseudonymized information and provides aggregated, anonymous outputs. Advertiser signals cannot be exported or accessed by Amazon according to the current AMC product page.
The most useful AMC measurement use cases
Use case | Question it answers | Decision it can support |
|---|---|---|
Path to conversion | Which eligible media sequences appeared before a conversion? | Adjust sequencing, retargeting and format roles |
Audience overlap | How much duplication exists between campaigns, audiences or formats? | Reduce waste and clarify incremental audience roles |
Reach and frequency | How are users distributed across exposure bands? | Change caps, budget allocation and creative rotation |
New-to-brand analysis | Which tactics are associated with eligible first-time brand purchases? | Separate acquisition from repeat-customer efficiency |
Time to conversion | How long does conversion typically follow the first or last eligible exposure? | Set realistic windows and retargeting intervals |
First-party outcome analysis | How do approved advertiser conversion signals relate to Amazon Ads exposure? | Measure leads, subscriptions, bookings or off-Amazon sales where the setup permits |
Custom audience creation | Which users meet a behavioral rule based on eligible AMC signals? | Create prospecting, suppression, sequencing or re-engagement audiences |
1. Path-to-conversion analysis
Path analysis examines the eligible touchpoints observed before an outcome. An advertiser might compare journeys that included streaming TV, online video and display with journeys that included display only. The goal is not to declare that the most common path caused the conversion. It is to understand how formats appear together and where sequencing may deserve a controlled test.
A useful path query should define:
the conversion event and the population eligible for analysis;
the lookback and campaign dates;
how repeated exposures are grouped;
whether order matters or only channel presence;
how direct converters and non-converters are treated;
which minimum volume and privacy thresholds affect the output.
Do not optimize to a path simply because it appears often. High-spend tactics naturally appear in more journeys. Compare conversion rate, audience size, cost and an appropriate baseline before changing the mix.
2. Reach, frequency and overlap
Average frequency hides the distribution of exposures. A campaign can report an average frequency of four while one large group saw one ad and a smaller group saw the campaign many times. AMC can help group eligible users into frequency bands and compare the rate of an outcome across those bands.
This is useful for diagnosing over-frequency, but the result is not automatically causal. People who receive more impressions may differ from light-exposure users because they use the service more often, remain in the audience longer or have different shopping behavior. Frequency analysis should guide the next experiment, not be presented as proof that another impression created the conversion.
Overlap analysis is equally important. If streaming TV and display reach mostly the same eligible audience, the plan may need different audience rules, suppression or sequencing. If the overlap is low, the formats may be extending reach in complementary ways.
3. New-to-brand and customer-stage analysis
New-to-brand metrics help separate customer acquisition from repeat purchase activity where the eligible Amazon Ads data supports them. AMC can add context by examining which campaigns, audiences or sequences are associated with new-to-brand outcomes.
Keep the interpretation precise. New-to-brand is a platform-defined metric based on the applicable lookback and eligible Amazon purchase history. It is not automatically the same as a first-ever customer in the advertiser's CRM, a new household or an incremental buyer.
When first-party data is available, the advertiser can create a clearer customer-stage framework: known customer, lapsed customer, high-value customer, prospect or unclassified user. The definitions should be documented before querying so the final report does not mix platform and CRM meanings.
4. First-party data and off-Amazon outcomes
Ads data manager allows eligible advertisers and partners to onboard first-party data once and reuse it across Amazon Ads products, including advanced programmatic buying and AMC. Amazon describes UI upload, API and integrated partner routes, with user-level controls and encrypted environments.
First-party signals can extend measurement beyond Amazon retail outcomes. Depending on the implementation, an advertiser may analyze approved events such as lead submissions, subscriptions, bookings, store transactions or product registrations alongside eligible Amazon Ads exposure.
The hard part is usually not the query. It is the signal design. Before onboarding data, define:
the legal basis, notices, consent and permitted advertising use;
the event name and business meaning;
timestamp quality and time zone;
deduplication and order or transaction identifiers;
value, currency and refund treatment;
the identifier and match strategy approved for the setup;
data retention, access and deletion responsibilities.
For the wider programmatic context, read First-Party Data Activation in Programmatic Advertising.
5. Turning measurement into AMC audiences
AMC is not only an analytics environment. Advertisers can use eligible signals to create custom audiences and activate them in Amazon Ads. Amazon's current product page describes audience activation across sponsored ads, video ads and display ads. Earlier AMC audience documentation also explains direct activation into Amazon DSP campaigns.
Practical audience rules can include:
people exposed to streaming TV but not display;
people with repeated product-detail-page visits but no purchase;
cart abandoners eligible for re-engagement;
recent purchasers excluded from acquisition media;
lapsed purchasers with no recent engagement;
high-value customer groups derived from approved first-party inputs.
Amazon's AMC Audiences announcement says custom audiences can use configurable SQL or query templates and can be refreshed on a schedule. Audience activation should still follow the same governance as measurement: documented purpose, minimum size, refresh logic, suppression rules and an agreed success metric.
For the planning layer behind those rules, use the Amazon DSP audience strategy guide for shopper signals and first-party data to define prospecting, suppression and customer-stage logic before translating it into an AMC audience.
AMC templates, SQL, APIs and Ads Agent
AMC supports several levels of technical maturity.
Templates
Curated analytical and audience templates give teams a faster starting point for common questions. A template should still be reviewed. Its event definitions, dates, filters and grouping logic may not match the campaign brief without adjustment.
SQL
Custom SQL offers the most control over populations, event logic and output. The analyst should use version control, comments and a query dictionary so another person can reproduce the result.
Ads Agent
Amazon announced Ads Agent in AMC for natural-language help with analytics SQL, audience creation and product guidance. Amazon says the feature is built on Amazon Bedrock and uses knowledge of AMC datasets, syntax and privacy rules.
Natural-language generation can shorten query development, but it does not remove analytical responsibility. The user still needs to inspect the SQL, confirm the population and validate the output before making a media decision.
APIs
Amazon provides AMC APIs for reporting, audience management and signal onboarding. APIs are useful when an agency or advertiser needs scheduled workflows, repeated analysis across accounts, software integrations or governed delivery into reporting systems.
What AMC can and cannot prove
Claim | Can AMC support it? | Condition |
|---|---|---|
These eligible touchpoints appeared before conversion | Yes | The events and sequence logic are available and correctly defined |
These campaigns reached overlapping eligible audiences | Yes | The comparison uses consistent dates, populations and identity coverage |
Higher-frequency users converted at a higher rate | Yes | This is an observed association, not automatic proof of causality |
This campaign caused the reported conversions | Not by query alone | A credible incrementality or experimental design is required |
This output identifies individual customers | No | AMC provides aggregated, anonymous outputs subject to privacy controls |
Every customer journey is visible | No | Only eligible, available and matchable signals are represented |
AMC replaces the advertiser's analytics stack | No | It should complement CRM, web analytics, sales and experimental evidence |
A practical Amazon DSP and AMC measurement plan
1. Start with one decision
Write the decision before the query. “Should we move budget from retargeting to streaming TV prospecting?” is stronger than “Find insights in AMC.”
2. Build a measurement dictionary
Define every campaign, audience, conversion, customer stage, product group, date window and KPI. Record which source owns each definition.
3. Audit signal coverage
Confirm which sponsored ads, advanced programmatic, conversion and first-party datasets are available in the instance. Document blind spots and matching limitations before analysis.
4. Separate reporting from experiments
Use AMC for descriptive and diagnostic analysis. When the decision needs a causal answer, design an incrementality test, holdout or another valid counterfactual where the product and scale allow it.
5. Validate the query
Check record counts, date coverage, joins, null values, duplicates, event filters and output totals. Compare the result with standard reports where the definitions should align.
6. Translate the result into an action
Every analysis should end with an owner, action and review date. Examples include changing a frequency cap, creating a suppression audience, testing another sequence or repairing a missing conversion signal.
7. Preserve reproducibility
Save the SQL or template version, parameters, execution date and interpretation. A result that cannot be reproduced should not become a recurring KPI.
Questions to ask before an AMC project starts
Which business decision should the analysis change?
Which Amazon Ads accounts and AMC instances are in scope?
Which signal tables and conversion events are available?
Are advertiser first-party inputs needed, and who owns onboarding?
Which definitions differ between Amazon Ads, CRM and web analytics?
What privacy, legal and data-governance review is required?
Will the output be descriptive, attributed or experimental?
What minimum volume and privacy rules may suppress results?
Who validates the query and signs off on the interpretation?
How will the result change the next Amazon Ads campaign?
Common mistakes
Starting with SQL instead of a decision. A technically correct query can still answer an irrelevant question.
Calling path analysis attribution. Sequence and association do not by themselves establish causal credit.
Using inconsistent windows. Campaign dates, conversion windows and customer-history definitions must be explicit.
Treating frequency correlation as an instruction to spend more. Heavy viewers may be systematically different from light viewers.
Uploading first-party data without a signal specification. Unclear event meaning creates analysis that cannot be trusted.
Ignoring suppressed or missing rows. Privacy thresholds and low volume can change the apparent distribution.
Comparing AMC totals with a dashboard without aligning definitions. Different populations and windows can make both numbers valid but different.
Building audiences without measurement. Every AMC audience needs a hypothesis, exclusion logic and success metric.
How Ad Geeks can help
Ad Geeks helps brands and agencies connect advanced Amazon Ads buying with a practical measurement plan. We can structure the AMC question set, audit available signals, define conversions and customer stages, build or review queries, create governed audiences and translate results into campaign actions.
We also help teams decide what AMC should not be asked to prove. That includes separating descriptive analysis from incrementality, reconciling Amazon Ads outputs with first-party analytics and documenting limitations before findings reach decision-makers.
Explore Amazon DSP access and managed activation through AdGeeks, compare Amazon DSP and DV360, review the Amazon DSP pricing and managed-service guide, or align measurement with an Amazon DSP audience strategy.
Frequently asked questions
Is Amazon Marketing Cloud the same as Amazon DSP reporting?
No. Standard reporting supports campaign monitoring and optimization. AMC is a privacy-safe clean-room environment for custom analytics, signal collaboration and audience creation across eligible datasets.
Does Amazon DSP still exist?
Amazon says Amazon DSP has evolved into Amazon Ads Agent. Existing programmatic capabilities continue through Display, Video, and Audio campaigns with DVA+ advanced settings. “Amazon DSP” remains the familiar market term.
Do advertisers need SQL to use AMC?
Not for every task. Amazon provides templates and Ads Agent assistance that can reduce the need to write SQL from scratch. Complex or high-stakes analysis still benefits from SQL review and analytical expertise.
Is AMC free?
Amazon says AMC is available at no cost to eligible advertisers. Paid Features and third-party signal subscriptions can add separate costs. Access and eligibility vary by advertiser, market and account setup.
Can AMC measure off-Amazon conversions?
Potentially, when approved advertiser conversion signals are onboarded and the implementation supports the use case. Data quality, permitted use, matching and event definitions determine what can be analyzed.
Can AMC prove incrementality?
Not through an ordinary query alone. AMC can support analysis and some experimental workflows, but a causal claim needs a valid counterfactual such as a holdout or another credible incrementality design.
Can AMC audiences be activated in Amazon Ads?
Yes, subject to eligibility, policies and minimum audience requirements. Amazon says custom audiences can be built from eligible AMC signals and activated across sponsored ads, video and display campaigns.
Does AMC expose individual customer data?
No. Amazon states that AMC accepts pseudonymized inputs and returns aggregated, anonymous outputs under privacy controls.
How far back can AMC analyze ad traffic?
Amazon's current AMC product page highlights an expanded 25-month ad-traffic lookback. The exact signal coverage in an advertiser's instance should still be checked before a longitudinal analysis is designed.
What should the first AMC analysis be?
Start with a question tied to a real decision and a dataset you can validate. Audience overlap, frequency distribution or a clearly defined path analysis is usually more useful than a broad request to “find insights.”
Final takeaway
Amazon Marketing Cloud makes Amazon DSP measurement more flexible, but flexibility is not the same as certainty. The strongest AMC work starts with a decision, defines the eligible population, validates the signals and states what the method cannot prove.
Use standard reports to run the campaign. Use AMC to explain patterns, build audiences and design the next test. Use experiments when the business needs a causal answer.
Sources
Amazon Ads, “Amazon Marketing Cloud.” Accessed and verified October 16, 2026. https://advertising.amazon.com/solutions/products/amazon-marketing-cloud
Amazon Ads, “Amazon DSP is now part of Amazon Ads Agent.” Verified October 15, 2026. https://advertising.amazon.com/solutions/products/amazon-dsp
Amazon Ads, “Accelerate Amazon Marketing Cloud workflows with Ads Agent (beta).” November 11, 2025. https://advertising.amazon.com/resources/whats-new/unboxed-2025-ads-agent-in-amazon-marketing-cloud
Amazon Ads, “Build a holistic first-party data strategy with Ads data manager.” October 15, 2024. https://advertising.amazon.com/resources/whats-new/ads-data-manager-beta
Amazon Ads, “Amazon Marketing Cloud announces custom audience creation and activation feature.” May 1, 2023. https://advertising.amazon.com/resources/whats-new/amc-custom-audience-creation-and-activation
