Build a useful investigation brief
Build matched comparison fixtures
Define current and baseline populations with known records, units, and aggregations. Calculate expected values and their intended difference independently. Add a controlled mismatch in one filter or definition, keeping other inputs fixed, so the comparison can reveal whether the report preserves the dimensions that are supposed to remain equivalent.
Trace reference selection and revisions
Inspect how the baseline is selected and whether it captures a definition or reads a mutable current setting. Compare filters, coverage, and measure versions on both sides. Record the first unintended difference, including a saved report whose baseline context changes later without the displayed comparison label or history reflecting that change.
Compare arithmetic and explanation
Verify the absolute or percentage change separately from the selected values, including zero or unavailable baselines where relevant. Inspect the explanation shown to readers. Evaluate one implicated selection or definition change against the matched control so a correction improves comparability without silently changing the report’s intended benchmark question.
What to carry forward
Deliver both context definitions, independent calculations, and the first selection or interpretation boundary that diverges. State which comparisons remain intentionally different and why. A supported correction makes the reference and arithmetic understandable rather than presenting a precise-looking change whose baseline population or measure meaning cannot be reconstructed.
Technical background: Metabase documentation.
Keep the decision with the work.
Use a Work Item in Aglet to record the problem, the evidence you have, and the next decision. Add an owner and priority, then keep updates in the discussion so the next person can follow the reasoning.
Create an account See the product workflow