Record a reproducible model analysis

Use this checklist when handing off a Dunsky model analysis. It is shared documentation guidance, not a claim that every model automatically records these fields.

Keep an analysis record

Record	Why it matters
Model and package version or commit	Identifies the calculation implementation
Input files or dataset versions	Identifies the data actually used
Scenario and configuration	Records the assumptions selected for the run
Run identifier and tag, when available	Connects the analysis to platform results
Execution environment	Distinguishes local, development, and production runs
Output files and interpretation notes	Preserves the evidence used in the deliverable
Validation checks and reviewer	Records what was checked before use

Review before comparing

Check the model version, units, geography, time horizon, scenario, and input versions. If any differ, describe the difference before interpreting a change in outputs. For a platform workflow, see Results and downloads. For local execution, follow the package guide linked from the model catalogue. Keep client-specific inputs and outputs in their approved project locations. The documentation should explain how to interpret a field without embedding unrelated project data.

​Keep an analysis record

​Review before comparing

Keep an analysis record

Review before comparing