Run a durable study

A structured study separates research design from execution. This is useful when the work has multiple questions, assets, segments, or a method such as MaxDiff.

1. Define the population

Start with a specific audience brief:

Build an audience of procurement leaders at European manufacturers with responsibility for industrial software purchases.

Use create_group_from_brief when you need a new grounded audience. Use list_groups when the audience may already exist.

2. Create the panel

Use create_panel to create the study container around the decision or research goal. Keep the objective concrete: what decision will the results inform?

3. Choose the execution path

For one straightforward question, use ask_panel.

For a multi-question study:

  1. call plan_panel_study;
  2. inspect the proposed plan;
  3. confirm or revise it;
  4. call run_panel_study;
  5. poll get_panel_study until the durable run completes.

The server-side run continues independently of the chat session.

4. Inspect disagreement

Do not reduce the panel to one average answer. Look for:

  • segment differences;
  • recurring objections;
  • minority views with strong reasoning;
  • uncertainty or evidence gaps;
  • reactions that change by market or buyer role.

Use get_panel_summary for the whole-study view and semantic artifacts.

5. Export the result

Use export_panel to produce PDF, CSV, XLSX, or JSON output. Choose a format based on the next consumer: PDF for a decision brief, spreadsheets for analysis, and JSON for another system.

Open Minds MCP documentation