A billing exception taxonomy gives each problem a name that points toward the right evidence and owner. It should not be a long list that hides judgment. Define terms in plain language and state which role may assign or change each category.

Separate kinds of work

Source exceptions include missing records, conflicting identifiers, inaccessible systems, malformed files, and duplicate candidates. Decision exceptions include coding, coverage, contract interpretation, write-off, refund, release, and policy choices. Mixing them causes staff to treat a missing document as a business answer.

Each category needs entry condition, required evidence, prohibited shortcut, owner, and exit state. “Access blocked” should identify system, search path, date, and approved route. It should not be closed as “missing document” because the assigned role cannot view the source. “Coding review required” is safer than “coding error” before qualified review.

Review category drift and unexplained “other” use. Preserve prior category after reopening. Report opening population, category counts, exclusions, unknowns, and definition changes before comparing periods. The goal is an honest map from evidence condition to accountable decision.

### Control points

  • Define entry and exit conditions.
  • Separate source gaps from decisions.
  • Require evidence and a bounded question.
  • Keep authority visible in labels.
  • Minimize unexplained “other.”
  • Preserve category history.
  • Explain taxonomy revisions.
  • ## Operating notes

+1. Write the operating purpose in one sentence before naming a metric, because a number without a decision context can invite the wrong action.

+2. Use the smallest stable identifier available and record where it came from, while keeping unnecessary personal or account details inside approved systems.

+3. Preserve the source wording beside any summary so a later reviewer can distinguish an observed fact from a preparer's interpretation.

+4. Record the date an event occurred, the date it was received, and the date it was reviewed when those moments have different operational meanings.

+5. State the timezone and business-day convention before comparing records created by people or systems working in different locations.

+6. Keep a missing source separate from a source that was not searched, and keep both separate from a source that was found but access-limited.

+7. Describe a calculation with its inputs and formula so another specialist can reproduce it without trusting an unexplained remainder.

+8. Do not borrow a field from a similar record merely because the target record is incomplete; similarity is a search clue, not evidence.

+9. Give every exception a named owner and one bounded question, rather than sending a broad request that hides several unrelated decisions.

+10. Set a next review date that follows the evidence or deadline, and record why that date is appropriate to the source.

+11. Use a carry-forward state when work is incomplete at a cutoff, and preserve the reason instead of counting it as silently finished.

+12. Retain rejected candidate matches because they explain why a plausible alternative was not accepted by the reviewer.

+13. Keep preparation, approval, posting, correction, communication, and release as distinct events in the record.

+14. Make clear which role may inspect, calculate, draft, approve, change, or communicate each part of the workflow.

+15. Treat a later event as an addition to history; do not erase an earlier observation simply because the current state looks cleaner.

+16. Compare periods only after checking that the source population, filter, definitions, and exclusion rules are materially the same.

+17. Report excluded, duplicate-risk, access-blocked, and unresolved items beside completed items so the denominator remains understandable.

+18. Explain a changed category or filter before describing a lower queue count as improvement or a higher count as deterioration.

+19. Use a specific exception status that tells the next person what evidence is missing and what action is permitted next.

+20. Do not use a generic pending label when the queue is waiting for different people, sources, or decisions.

+21. Preserve the source path and capture time for portal checks, exports, correspondence, and generated reports.

+22. Limit copied information to what the next reviewer needs, particularly when a working note could otherwise expose protected data.

+23. Route permission questions through the approved owner path and never use another person's account to complete a check.

+24. Keep a version link when a report, invoice, claim, statement, or source document changes after the first review.

+25. Record the exact event that caused a row to reopen, rather than describing the reopening as a general correction.

+26. Close a row only when its approved result, documented hold, or accountable carry-forward is visible.

+27. Use a short final reconciliation to prove that the opening population did not lose records between intake and disposition.

+28. Review a sample of completed rows for source traceability, boundary compliance, and clear next action before trusting an aggregate measure.

+29. Escalate contradictions early, because a confident summary built on conflicting sources is harder to repair than an explicit unresolved state.

+30. Make the handoff useful to a new shift: source, current state, evidence, owner, question, deadline, and next review should be findable together. ## Additional review note

The taxonomy should be tested with mixed examples before it becomes a reporting standard. Include a missing source, a conflicting source, an access block, an owner-only decision, a duplicate candidate, and a reopened item. Ask whether two reviewers would assign the same category from the written rule. Revise ambiguous definitions before using category counts as a performance signal. Classification quality is part of queue quality.

A practical review sequence

Start a taxonomy review with a small, known set of recently handled billing exceptions rather than with an empty spreadsheet. Choose examples that include a clean resolution, an unresolved item, a reopened item, and at least one case where the first label changed. For each example, ask a reviewer who was not involved in the original classification to assign a category from the written definitions. Record the assigned label, the evidence they used, the question they believed they were answering, and the point at which they hesitated. This exercise exposes categories that sound precise but still depend on private knowledge.

When two reviewers disagree, do not settle the difference by averaging labels or selecting the more familiar term. Identify whether the disagreement concerns the evidence condition, the decision authority, the stage of work, or the desired outcome. A source gap may remain a source gap even when a specialist later decides how to proceed. Update the definition only when the intended boundary is clear, then retest the original examples and a fresh sample. Keep the previous wording and effective date with the change record so trend reports can explain why category counts moved. A taxonomy is useful when it makes the next permitted action easier to see, not when it creates the longest possible list.