Establish what is happening
Map the expected response
Write which filter, fields, values, and measure should affect the card. Record the card's population, aggregation, default state, and any documented exclusions. Keep a fixture with one row inside and one outside the selection so the expected delta is unambiguous.
Trace the card population
Follow the control through selection state, predicate construction, query or model boundary, aggregation, formatting, and card rendering. Compare selected values with included identifiers and the number shown. Mark the first stage where the card ignores, over-applies, or reinterprets the filter.
Bound visual relationships
List cards and charts on the page, noting which should share the filter and which are intentionally independent. Separate a mapping defect from a measure-grain or date-boundary issue. Ask the report owner to confirm the intended relationship when visible proximity suggests an unsupported link.
What to carry forward
Triage is complete when the filter-to-card expectation, first propagation divergence, affected surfaces, and report owner are recorded. Keep the finding provisional when the card's model or exclusion rules are not visible enough to distinguish intentional independence.
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