Use this when partners, departments or reporting audiences use the same word differently. Bring one existing measure and the request it must answer.
Start with the decision behind the label.
In our fictional training workflow, the course coordinator tracks skill use after 30 days. A sponsor requests “successful completion.” Those measures may both matter, but they answer different questions. Completion records cannot establish later skill use.
Compare the population, qualifying event, observation period, evidence source, denominator and treatment of missing answers. Similar wording is not enough.
Same meaning: reuse the approved measure.
Different view, supported by records: calculate and label the alternative.
Additional justified evidence: plan collection.
Conflict or missing definition: resolve it before reporting.
Work through one difference.
The training team has 40 completers, 25 follow-up respondents and 15 reporting skill use. A respondent rate is 15/25. A confirmed-user share of all completers is 15/40. Both are calculable, but neither is a completion rate, and nonrespondents must not be described as non-users.
If a partner supplies only a percentage, request the counts and definition before aggregation. Averaging percentages from different-sized groups can give the wrong combined result even when the labels match.
Approve the rule before extending the list.
- Write the question the measure must answer.
- Put both definitions and their sources side by side.
- Compare population, event, timing, source and calculation.
- Record whether to reuse, translate, add evidence or resolve a conflict.
- Test the agreed interpretation with ordinary, duplicate and missing records.
Retain different measures when they describe different outcomes or stages. Remove redundant measures when they serve no distinct decision or justified requirement. A required measure with missing evidence needs an explicit collection or clarification decision, not quiet removal.
External standards can be mapped after the local measure is defined. Keep the exact standard version and differences in scope. AI can propose comparisons with citations; the responsible person approves the relationship.
Your output
One agreed meaning, one unresolved difference if necessary, and an owner. Next, write its counting rule so another person can apply it.
Test edge cases before collecting the next period
Use a small test set containing ordinary and difficult records. Ask two people to apply the specification independently. Differences reveal an unclear rule before it becomes a reporting discrepancy.
| Test record | Expected review |
|---|---|
| One person submits twice | Apply the agreed identity and latest-valid-response rule; count the person once |
| A response arrives outside the follow-up window | Retain its timestamp; apply the approved window rule rather than shifting the date |
| No follow-up response | Record unknown and reflect it in coverage |
| Two sources report different statuses | Keep the conflict and resolve using the agreed source/review policy |
| A participant changes jobs | Do not assume continuous retention; use the explicit continuity rule |
| One partner sends only a percentage | Request counts and denominator; do not average percentages without valid weights |
A missing owner does not automatically mean a measure is unnecessary: assign responsibility before retiring a required measure. Likewise, an unavailable value may signal a collection gap. Record the decision to retain, revise or stop a measure and any reporting commitments affected.
Add this to your plan
Record what you decided in this chapter in the same working evidence plan you started in Foundations, so definitions, sources and checks stay in one place.