Skip to main content

Run Measurement Forecasts

Run Measurement Forecasts​

Use SDK helpers to create the project and import observations, then run forecasts through the measurement analytics API.

Use an authenticated client authorized for measurement, an existing project and recorded observations. Select bounded request fields from the published API contract; a route alone is not a complete request body.

POST /api/v1/measurement/projects/{projectSlug}/forecast-runs
GET  /api/v1/measurement/projects/{projectSlug}/forecast-runs/{forecastRunUuid}

A usable forecast response includes status, method, generatedAt, inputCount, sampleCount, seed, P10/P50/P90 percentiles, warnings, and provenance. Sparse data should render as a warning state in Measure, not as an overconfident forecast.

Decision value-of-information runs use the same analytics surface:

POST /api/v1/measurement/projects/{projectSlug}/decisions/{decisionSlug}/simulation-runs
GET  /api/v1/measurement/projects/{projectSlug}/decisions/{decisionSlug}/simulation-runs/{simulationRunUuid}

Example: forecast checkout-runway after importing its completed-work series, retain the returned run UUID, and read that exact run. In Measure, compare its percentiles with the project target. Decision simulation also requires a configured decision and inputs; creating a project alone does not create those.

Expected result: a persisted run with provenance and values or explicit warnings. Sparse results call for more valid evidence, not repeated runs until a preferred number appears. Correct validation errors before another request; refresh the recorded run when an earlier request's result is uncertain.