Aglet

Learn From Repeated Metric-Grain Confusion

A grain lesson should retain the small example that made two plausible totals disagree. Preserve source rows, grouping keys, the metric definition, and the report wording. Then improve the shared definition or review practice so a future measure exposes its row unit before readers compare it with another number.

Keep the lesson for the next incident

  1. Keep the revealing example

    Store a fixture with repeated events, shared membership, and the intended count at each grain. Record the source revision, filters, aggregation, visible label, and export header. Preserve the original misleading interpretation beside the corrected definition for future review.

  2. Classify recurrence patterns

    Review earlier metric questions for hidden joins, repeated memberships, changed grouping, ambiguous labels, and visuals that used different grains. Group by the missing contract or review signal. Use named examples and observed consequences instead of claiming a universal source of error.

  3. Assign a definition owner

    Make the metric owner maintain the row-unit statement and the report reviewer check it when measures or joins change. Add a trigger for new aggregations, drill paths, or exports. State which grain cases the fixture detects and which business definitions still need human judgment.

What to carry forward

The learning record is complete when it connects confusion to a missing grain definition or review cue, preserves an interpretable fixture, and names an owner and trigger. Keep intentional measures at different grains explicit so future readers do not mistake a valid distinction for a defect.

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