Core Concepts

What Simulate does

Adobe Brand Intelligence Simulate predicts how creative assets will land with an audience. You supply one or more creatives - a headline, an email, an image, a video - and Brand Intelligence runs them past a synthetic audience drawn from your organization's populations and segments. When the run finishes, it returns an executive summary and detailed findings for each creative.

A simulation replaces a guess about which creative will resonate with an answer from the audience. When you integrate Simulate into an assistant, the simulation result is the answer: the assistant relays it and does not add its own opinion of the creatives.

Resource model

Workspace
└── Simulation (one study: objective, template, audience, assets)
    └── Run (one execution of the simulation)
        └── Executive summary + findings

Workspace - a container for simulations. A user can belong to several workspaces, each either personal or shared. One is flagged is_active: the workspace the user currently has open in the Brand Intelligence web app. Every simulation belongs to exactly one workspace.

Simulation - one study testing a set of creatives against an audience. It records the objective, the template, the target population and segments, the sample size, and the assets.

Run - one execution of a simulation. A simulation's reported status follows its latest run when one exists.

The four decisions

Creating a simulation means settling four decisions, in this order:

1. Objective - what the simulation should find out, for example "which of these two subject lines drives more opens".
2. Assets - the creatives to test.
3. Template - the study template to run.
4. Audience and participants - which segment or segments to target, and how many simulated respondents.

The objective and the assets come from the user. The template and audience options come from get_simulation_options, which lists what your organization has available.

Templates

A template defines what a simulation measures. Each template returned by get_simulation_options carries:

• a title and a description of what it measures
• its key_questions and scored dimensions
• the design types it allows
• form defaults: default_name, default_objective, default_design_type, and default_sample_size

When the request omits a field, the template's default applies.

Design type follows from the number of assets. A comparative design (sequential_monadic) needs at least two assets; with only one asset, the server runs a monadic design instead.

Audiences

Audiences come from your organization's populations. Each population contains segments, such as "Existing customers" or "Gen Z". A simulation targets one population and, optionally, one or more of its segments. An empty segment list targets the whole population.

Sample size is the number of simulated respondents. When omitted, the template's default_sample_size applies.

Asset types

Each asset is a flat object tagged by type. The type follows what you are testing, not the format of the file you have.

When you are testing emails and have an image of the email body, that image is the email's body_image, not a standalone image asset.

label is optional on every asset. Unlabeled assets are numbered "Option 1", "Option 2", and so on, by their position in the list.

Asset sources

Only media assets and email body images carry bytes. Each declares a source telling Brand Intelligence where to read them from:

• url - the file is already reachable at a public HTTPS URL. Pass the URL directly; no upload step is needed.
• upload - the file is local. Stage it first with open_asset_upload (an inline upload panel) or prepare_asset_upload (a ready-to-run upload command), then reference the returned upload_id.

Supplying both a URL and an upload_id, or neither, is rejected. See Using MCP for when to use each upload path.

Simulation lifecycle

create_simulation returns one of two outcomes:

Simulations are asynchronous. After a launch, poll get_simulation until both of these are true:

• status is resultsReady or archived, and
• summary_status is ready.

summary_status tracks the executive summary separately from the run:

A run can reach resultsReady while its summary is still pending, so keep polling until both conditions hold.

Results

Once ready, get_simulation returns:

• executive_summary - a summary of how the creatives performed, plus recommendations.
• analysis - the detailed findings.
• inputs - what the simulation was configured with: objective, design type, participants, population, segments, dimensions, and assets.
• webview_url - a link to open the simulation in the Brand Intelligence web app.

Report results with their confidence. If the creatives perform within the margin of error, the honest answer is that there is no clear winner.

Authentication

Every MCP request requires a Bearer token. MCP clients obtain one through an interactive OAuth sign-in; see Using MCP. The tenant is resolved server-side from the caller's Adobe organization, so no tool takes a tenant id.