Amazon published a detailed technical walkthrough on May 1, 2026, explaining how Ads Data Manager (ADM) works from an integration standpoint, covering account architecture, the four-step API workflow, and the three activation paths available once data is ingested. The video, presented by Mayank Arora, partner solutions architect at Amazon Ads, provides one of the clearest public explanations to date of how the system is structured and what technical teams need to do to connect it programmatically.

What ADM is

According to Arora, ADM is "a stand-alone solution for managing advertiser first party data across Amazon ads." The emphasis on "stand-alone" matters. ADM is not embedded inside Amazon DSP or the Ads Console as a secondary feature - it is positioned as a dedicated infrastructure layer that sits at the center of first-party data workflows and connects outward to multiple Amazon advertising products.

The data flowing into ADM can originate from CRM systems, analytics platforms, or customer interaction records. Once inside the system, those data assets can be activated across Amazon DSP, Events Manager, and Amazon Marketing Cloud (AMC). Data can enter ADM through three channels: the console UI, the API, or partner integrations.

This architecture matters for advertising teams that have historically needed to upload similar audience data separately into different Amazon products. ADM is designed to remove that duplication. A single data set, once created inside ADM, can be sent to multiple destinations without re-uploading.

Three structural components

ADM is organized around three core building blocks. The first is the manager account, which is the configuration layer where ADM is set up. The second is the data room - a secure, isolated environment where all data assets reside. According to Arora, there is one data room per manager account, meaning the data room is the single container for everything that flows through ADM for a given organization.

The third component is the data set. A data set is a collection of customer records stored inside the data room. Each record contains identity signals, which may include hashed emails, phone numbers, mobile ad IDs, or universal IDs. Optional attributes such as behavioral or demographic data can accompany the identity signals, though they are not required.

One structural recommendation Arora flags is configuring data sharing links early. According to the walkthrough, sharing rules and destination connections should be set up before any large-scale upload begins. Doing so ensures that when data sets are ingested, they are already mapped to the correct advertiser accounts or destinations, avoiding delays in activation. It is a detail that is easy to overlook when moving quickly through an integration but has downstream consequences for how promptly data reaches its targets.

Identity signals and hashing requirements

All identity signals must be hashed before being sent to the API. This is not optional. The system does not accept plain-text personally identifiable information. According to the walkthrough, the signals that can be submitted include hashed email addresses, hashed phone numbers in E.164 format, first names, and zip codes. The more identity signals a record includes, the better the match rate when Amazon attempts to resolve that record against its own user graph.

This design is consistent with how other large advertising platforms handle first-party data ingestion. Google's Data Manager API, launched in December 2025, similarly requires normalization and hashing before transmission, with email addresses requiring lowercase conversion and removal of dots in Gmail addresses before the hashing step is applied.

The hashing requirement also connects to user consent. Each uploaded member record in ADM requires a user consent field. In the demo, Arora sets the geographic consent to "US" and marks both consent flags as true. This structure means ADM is designed to carry consent signals alongside identity signals, which is relevant for advertisers operating under privacy regulations that require explicit consent records attached to data activations.

The four-step API workflow

The integration follows four sequential steps. Step one is creating a data room, which is a one-time operation for each manager account. Step two is creating an audience data set. Step three is uploading hashed audience members into that data set. Step four is verifying the upload in the console and connecting the data to its intended destinations.

In the Postman demonstration, Arora walks through each step using live API calls. To create a data room, a POST request is sent with three required headers: the API client ID from a Login with Amazon application, a bearer token for authorization, and the manager account ID. The response returns an assignment and creation timestamp confirming the data room has been provisioned.

Creating an audience data set also uses a POST request, this time to the ADM audiences endpoint. The required fields include a name, a country code, an optional description, and an ID retention flag. In the demo, ID retention is set to true. According to Arora, this setting retains hashed identifiers for 90 days, which improves match rates by giving Amazon more time to resolve identifiers against its user graph.

The API response returns a status 201 and a data set ID. That ID is required for the next step. It must be saved and passed as a path parameter when uploading members.

Uploading audience members is another POST request, this time to the ADM audiences endpoint followed by the data set ID and the path segment "members." Each member object contains three parts: an external user ID (the advertiser's own customer identifier, such as a CRM record ID), a user identity object containing the hashed PII signals, and a user consent object. A successful upload returns a status 200 and an ingress ID, which functions as a tracking identifier for that particular upload batch. According to the response shown in the walkthrough, all records in the batch are confirmed as accepted.

After uploading, the record count in the console UI updates to reflect the ingested data. In the demo, one test record is uploaded and the console shows one record populated following a refresh.

Three activation paths

Once a data set is populated, three activation options are available through the console. The first is creating an advertiser audience, which makes the data set available for targeting in Amazon DSP campaigns. The second is sharing the data set to Amazon Marketing Cloud, which enables advanced measurement and custom attribution analysis. The third is attributing off-Amazon conversions, which connects the data set to Events Manager for conversion tracking.

These three paths serve different functions in the advertising funnel. DSP targeting uses the audience for programmatic ad delivery. AMC sharing uses the audience for analytics, including audience overlap analysis, path-to-purchase modeling, and custom attribution. Events Manager connection uses the audience for closing the measurement loop by attributing conversions that happen outside Amazon's owned properties back to advertising activity.

Critically, the same data set can be activated across all three destinations simultaneously. An advertiser does not need to choose one path and exclude the others. The data room architecture is designed to support this multi-destination activation from a single upload.

How ADM fits into Amazon's broader data infrastructure

ADM did not emerge in isolation. Amazon introduced Ads Data Manager at the unBoxed 2024 event in October 2024, where it was described as an interface allowing advertisers to securely upload their signals once and use them across Amazon DSP and Amazon Marketing Cloud. At that point, it was integrated with partner platforms including Treasure Data, Salesforce, and Tealium, providing import paths for first-party data stored in external systems.

Since that launch, Amazon has continued building out the surrounding infrastructure. In September 2025, Amazon made AMC directly accessible to sponsored ads advertisers within the Amazon Ads console, removing the previous requirement for partner or representative intervention before getting access. That change lowered the barrier to using the AMC activation path that ADM connects to.

In November 2025, Amazon extended AMC's ad traffic lookback window from 13 to 25 months, enabling advertisers to run analyses spanning more than two full calendar years. This expansion directly affects the value of sharing data sets from ADM to AMC, since a longer historical window means more complete year-over-year analysis for any audience that has been ingested and retained over time.

Also in November 2025, Amazon launched a unified Campaign Manager platform that collapsed the DSP and Ads Console into a single buying environment. ADM's DSP activation path now feeds into this unified interface, meaning audiences created through ADM and shared to DSP can be applied within the same workspace that manages sponsored ads campaigns.

The AI layer has grown significantly around the same infrastructure. Ads Agent, announced November 11, 2025, automates campaign management tasks across AMC and DSP using natural language processing, translating business questions into SQL queries and recommending targeting segments. Data sets activated from ADM into AMC become queryable inputs for the Ads Agent's analytics capabilities.

Why this matters for marketing and technical teams

The ADM API walkthrough is relevant to several distinct groups. For developers and marketing technology teams responsible for data onboarding, the video provides a concrete demonstration of the authentication setup, request structure, and response handling needed to build a production integration. The Postman demo shows exactly what headers are required and what a successful response looks like, reducing ambiguity in the early stages of implementation.

For advertising operations teams managing audience strategy, the multi-destination activation model changes how data reuse is approached. A single CRM export, hashed and uploaded once into ADM, can simultaneously feed DSP targeting, AMC analytics, and conversion measurement through Events Manager. Previously, coordinating that coverage required separate uploads to separate systems, each with its own format requirements and authentication flows.

The 90-day ID retention setting is a specific configuration choice with practical consequences. Setting ID retention to true keeps hashed identifiers active for 90 days, which Amazon says improves match rates. For advertisers running always-on campaigns where audience freshness matters, understanding whether ID retention is enabled and what the retention window is affects how audience segments age over time.

The consent handling built into the member upload structure is also significant for compliance workflows. Each record in ADM carries consent flags, meaning the system is designed to carry the consent provenance of each data point through to activation. This is directly relevant for advertisers operating under GDPR in Europe or CCPA in California, where demonstrating that consent was collected and transmitted alongside data activations is a compliance requirement.

The emphasis Arora places on setting up sharing rules and destination connections before uploading large data sets reflects a sequencing principle that applies broadly to API integrations: configuration and routing should precede data volume. Building a large audience in ADM without pre-configured destinations means data sits in the data room without flowing anywhere, requiring retroactive configuration and potential re-processing.

Amazon's advertising revenue has grown substantially in recent periods. Amazon's ad revenue reached $21.3 billion in Q4 2025, representing 23% year-over-year growth. That scale means the platforms ADM connects to - DSP, AMC, and Events Manager - are operating at significant volume, and the data infrastructure that feeds them has direct bearing on campaign performance outcomes for advertisers of all sizes.

Full API documentation, code samples, and integration guides are available at the Amazon Ads developer site, according to the walkthrough. The documentation covers authentication, endpoint specifications, and implementation details for teams building programmatic data onboarding at scale.

Timeline

Summary

Who: Amazon Ads published the walkthrough, presented by Mayank Arora, partner solutions architect at Amazon Ads. The intended audience is technical teams at advertising agencies, brands, and marketing technology vendors building programmatic first-party data integrations.

What: A technical video walkthrough of Amazon Ads Data Manager (ADM), covering the system's three-component architecture (manager account, data room, data sets), the four-step API integration workflow, the identity signal and hashing requirements, and the three activation destinations - Amazon DSP, Amazon Marketing Cloud, and Events Manager.

When: Amazon published the walkthrough on May 1, 2026. ADM was originally introduced at unBoxed 2024 in October 2024.

Where: The walkthrough covers the Amazon Ads API, accessible through the Amazon Ads developer site. ADM operates globally as part of Amazon Ads infrastructure, with activation destinations spanning Amazon DSP, AMC, and Events Manager.

Why: ADM addresses a longstanding operational problem for advertisers: the need to upload the same first-party data separately into multiple Amazon advertising products. By centralizing ingestion into a single data room and enabling multi-destination activation from a single data set, ADM reduces duplication, ensures consistent data across products, and - when sharing rules are configured correctly - allows data to flow to its destinations immediately upon upload.

Share this article
The link has been copied!