Establish what is happening
Define both sides of the comparison
Record the current measure and baseline measure with their populations, filters, aggregation, units, and time coverage. Identify which dimensions are intentionally different and which should match. Do not assume a label such as previous period or benchmark establishes comparability when the underlying definitions are not visible.
Reconcile a controlled comparison
Use a small known dataset or bounded report slice and calculate both values independently. Compare the displayed absolute or percentage change with those values. Keep the arithmetic separate from population selection so a mathematically correct difference is not mistaken for evidence that the two sides measure equivalent work.
Locate the first context mismatch
Inspect saved filters, baseline selection, measure revisions, and any fixed benchmark context. Find the first unintended difference. If both definitions match, keep real behavior change and incomplete data as separate possibilities rather than assuming that every unexpected comparison must come from a baseline configuration defect.
What to carry forward
End with current and baseline definitions, independent values, intended differences, and any unintended mismatch. State which context remains unknown. The next owner should be able to assess comparability before investigating the apparent trend, rather than treating a percentage change as self-explanatory evidence of improvement or decline.
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