Aglet

Learn From Recurring Mismatched Counts

Learning from count mismatches means making population and arithmetic assumptions visible. Preserve the smallest cohort that exposed the difference, the intermediate set comparison, and the definition decision. Turn those lessons into fixtures and review practices that catch drift before reports disagree again.

Keep the lesson for the next incident

  1. Document count contracts

    Write whether each consumer counts rows, entities, events, or groups, including filters, status interpretation, time zone, timestamp, and distinctness. Name intentional exceptions and their owner. Keep the contract beside the query or report so future comparisons begin with aligned definitions.

  2. Retain set fixtures

    Keep compact fixtures for boundary dates, status changes, duplicate rows, one-to-many joins, missing keys, and merged groups where relevant. Assert intermediate sets as well as totals. Label expected differences when consumers intentionally count different concepts.

  3. Review recurrence signals

    Assign owners for shared definitions and each consuming count. Review discrepancies by cohort and refresh window on a defined cadence, with a trigger for a new unexplained set difference or widening gap. Close follow-up only after repeated comparable observations honor the documented contracts.

What to carry forward

Close the learning record with explicit count definitions, set-level fixtures, ownership, and a recurrence trigger for unexplained membership or total changes. Preserve intentional differences as documented behavior. The durable improvement is faster reconciliation grounded in shared populations, not a permanent promise of equal numbers.

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