Self-Service AMC vs a Dedicated Instance: Which Amazon Marketing Cloud Setup an Agency Should Pick

An agency that runs Amazon Marketing Cloud for its clients makes the same decision once per brand: run AMC on the brand's own self-service instance, or create a fresh, dedicated instance inside the agency's AMC account. The choice decides which paid features the queries can use, who keeps the workflows when the engagement ends, and what happens to audiences built from that instance.

This post compares the two setups from the agency side, then shows how the choice looks in DataDoe, where enabling AMC for a client starts with exactly this pick, and the queries then run from Claude or ChatGPT next to the brand's sales, ads and profit.

Two ways into AMC

Amazon offers more than one route to a clean room instance:

• Self-service AMC. Every Sponsored Ads advertiser can open one standard instance per Sponsored Ads account and marketplace. It is the same instance the brand sees in the AMC console in Amazon Ads.
• An AMC account that can create instances. Amazon provisions these, usually after a request through an Amazon Ads account executive or partner, and new instances are created in them through the Amazon Ads API. They are common for agencies. Each new instance starts empty.

Both are Amazon's own AMC. The queries, the 14-day attribution window and the privacy thresholds that drop rows with too few customers are the same in both. What differs is ownership, contents and lifetime.

Self-service vs dedicated, side by side

• Where it lives. Self-service sits in the brand's Sponsored Ads account. A dedicated instance sits in the agency's AMC account, created for that brand.
• What is in it on day one. Self-service already holds whatever the brand ran in the AMC console: its workflows, schedules and results. A dedicated instance is new and empty.
• Paid features. Subscriptions such as Amazon Insights belong to one instance. If the brand pays for them in its self-service instance, a dedicated instance does not inherit them.
• Shared with the console. Work in a self-service instance is visible in the AMC console and the other way round. Deleting a workflow from one side deletes it on the other.
• Terms. For self-service, the Amazon login behind the Ads connection has to accept the AMC terms once per region (North America, Europe, Far East) in the AMC console.
• Query limits. Amazon allows 30 on-demand runs a day and 10 running at once, per instance. Scheduled runs do not count toward the daily 30. On self-service, the 10 parallel slots are shared with runs the brand starts in the console.
• When the client leaves. Amazon does not let anyone delete a self-service instance through the API, so it stays with the brand, with its workflows and schedules. A dedicated instance is deleted when it is disconnected, together with its workflows and schedules.

When the brand's own instance is the right pick

• The brand pays for Amazon Insights. Three of DataDoe's ready-made templates need it: organic vs ad-touched new-to-brand customers by ASIN, repeat purchase and customer value cohorts, and time to second purchase. They only run on the instance that holds the subscription. Amazon has made its 1P paid features free to query through 31 December 2026, but the subscription is still required.
• The brand's team also works in the AMC console. Self-service lets both sides see the same workflows and results.
• The brand should keep the work. If the relationship ends, the workflows and schedules stay in the brand's account. That is easier to promise in a contract than a handover.
• The brand already has history in AMC. Its existing workflows can be listed, validated and run by their Amazon ids.

When a dedicated instance makes more sense

• The agency has an AMC account and wants a clean start. Its own SQL and schedules live in an instance it created, separate from whatever the brand ran before.
• Console traffic should not compete. A dedicated instance does not share its parallel slots with the brand's own console runs.
• The analysis is the agency's product. When the engagement ends and the instance is disconnected, the agency's workflows go with it rather than staying in the client's account.
• No paid features are involved. Most templates, such as new-to-brand by ASIN, keyword and search term halo, path to conversion and cross-sell, use the free data every instance has.

The trade is simple. Self-service keeps the brand's paid features and history and survives a disconnect. Dedicated gives the agency a clean instance it controls, which disappears with the connection.

How the choice works in DataDoe

AMC is in beta in DataDoe and available to every organization without a request. The steps, for each client account:

1. Connect Amazon Ads for the seller or vendor on the same marketplace.
2. On the Accounts page, the organization owner clicks AMC in the account's row. DataDoe always offers the brand's own self-service instance. If the Ads connection also has AMC accounts on that marketplace, such as the agency's, the owner can pick a new dedicated instance in one of them instead.
3. For self-service, DataDoe asks for confirmation first, because enabling creates the instance in the brand's Ads account if it does not exist yet. Setup runs in the background, and owners get an email when AMC is ready.
4. To validate, start or cancel queries and to save workflows, the account needs "Read and write" access. With "Read only", Claude can still read the schema, the workflows and the results.

To switch a client from one setup to the other, disconnect AMC, wait until disconnecting finishes, and enable it again with the other option. Past results are no longer available through DataDoe after any disconnect, so download what you need first. AMC also has to be disconnected before the client's account can be deleted from DataDoe or its Ads connection changed.

Running every client from one conversation

All client brands can sit in one DataDoe organization, and Claude reaches each one by its account id. A few prompts that work the same on either setup:

• "For each client with AMC enabled, check which templates are available on its instance and which are blocked by a missing paid feature."
• "Run the new-to-brand gateway ASINs template for last month on Brand A and put the top ASINs next to their profit after ads."
• "Schedule the keyword and search term halo template weekly for every client and tell me which ones are already scheduled."
• "List the results that came back from schedules and console runs since Monday."

Checking a template against a client's instance tells Claude whether every data source it needs is there, before anything runs at Amazon. Queries usually take 15 minutes or more. Results of schedules and console runs appear in DataDoe within an hour. Amazon charges nothing for AMC itself, apart from paid features such as Amazon Insights; in DataDoe, AMC is part of the Base plan at $97 a month with a 14-day free trial, and results are priced in AI tokens from the plan.

Audiences depend on the instance too

AMC audiences for Sponsored Ads can be created from Claude through DataDoe Actions: rule-based audiences selected with SQL, such as shoppers who added to cart and did not buy, and lookalike audiences built from a seed. They can raise bids in Sponsored Products and Sponsored Brands or serve as targets in Sponsored Display. DSP audiences are not supported. Amazon lists a finished audience in Sponsored Ads about 24 to 48 hours after it is ready, and Claude then attaches it with the campaign or target Actions.

Audiences are built in the instance AMC runs on. If a dedicated instance is disconnected, audiences created there may stop being available in Sponsored Ads, which is one more reason to put long-lived audiences on the brand's own instance. Actions stay off until you enable them, most AI clients ask before a change, and every change is logged.

What to keep in mind

• AMC in DataDoe is a beta.
• Amazon Insights is a paid Amazon feature subscribed per instance. DataDoe does not subscribe on anyone's behalf.
• Small brands or short periods can return fewer rows or an empty result because of Amazon's privacy thresholds.
• Run reports at least 14 days after the period ends, so late conversions are counted.

For what AMC answers and how queries run from your AI client, see Amazon Marketing Cloud in Claude and ChatGPT. For managing many client accounts in one place, see DataDoe for agencies. For the rest of the ad account, see Amazon Ads in Claude and ChatGPT, and for DSP reporting, the DSP reporting API post.

Questions people ask

How does an agency get access to Amazon Marketing Cloud?

Either through each brand's self-service AMC, which every Sponsored Ads advertiser can open, or through an AMC account provisioned by Amazon, usually requested through an Amazon Ads account executive or partner.

What is the difference between self-service AMC and a dedicated instance?

Self-service is the brand's own standard instance, shared with the AMC console and kept by the brand. A dedicated instance is a new, empty instance in another AMC account, such as an agency's, and it does not carry the brand's paid features.

Which instance should we use if the brand pays for Amazon Insights?

The brand's self-service instance. Paid features belong to one instance and do not carry over to a dedicated one.

What happens to the AMC instance when a client leaves?

A self-service instance stays in the brand's Ads account with its workflows and schedules. A dedicated instance created through DataDoe is deleted when AMC is disconnected.

Do we need to write SQL?

No. Ready-made templates cover the common questions, and Claude can write and validate SQL against the instance's live schema when you need something else.