Aglet

Prioritize Inconsistent Identifier Cleanup

Prioritizing identifier work requires separating harmless representation drift from identity ambiguity. Rank cases by the failures they cause, the number of dependent workflows, recurrence, and confidence in a reversible normalization. Make unresolved identity conflicts visible rather than burying them inside a broad cleanup project.

Decide where the work belongs

  1. Rank downstream damage

    For each identifier variant, document missed joins, false merges, duplicate entities, broken references, and changed totals. Prefer evidence from actual queries or reports over an assumed severity. A smaller variant that alters entity ownership may outrank a common variant that only adds whitespace.

  2. Assess repair confidence

    Classify a proposed action as deterministic formatting, source mapping, cross-record reconciliation, or unresolved. Estimate review effort from representative conflicts and identify any dependency on a source owner. Keep ambiguous identity cases behind a review gate even when their count is high.

  3. Make the queue decision

    Choose investigate, normalize in a bounded path, request source clarification, or monitor. Include sample size, affected workflows, rollback or comparison method, owner, and a trigger that would raise priority. Avoid prioritizing solely by row count when the records have different business consequences.

What to carry forward

Return a ranked queue that distinguishes formatting cleanup from identity reconciliation. Each item needs a consequence, evidence-backed population, confidence level, dependency, and next action. Treat the ranking as provisional when joins are not representative or when a canonical identifier has not been established.

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