Measurement plan
This plan defines proposed learning measures for the fictional Northstar prototype. It reports no customer, adoption, or business outcomes.
| Decision question | Measure and denominator | Proposed signal | Guardrail |
|---|---|---|---|
| Can a PM make a traceable decision? | Participants with a justified decision and both F01/F02 sources identified / five participants | 4/5 unassisted | A status change alone does not count. |
| Does the score support rather than replace judgment? | Participants explaining impact, confidence, effort, and score direction / five participants | 5/5 identify it as directional | Do not treat agreement with the score as success. |
| Is the simulated tag boundary clear? | Participants identifying the tag as editable simulated context / five participants | 5/5 | No inference that a model or accuracy evaluation exists. |
| Does grouping preserve reach correctly? | Participants naming the four-account F01/F02 union / five participants | 4/5 unassisted | Never add duplicate record counts as reach. |
| Is the workflow worth its cost? | Participant comparison after current-method and prototype tasks / five participants | Qualitative pattern only | Do not claim time savings from five sessions. |
Record completion time only among successful attempts. Keep all five analyzed participants with retained consent in the primary denominator, including incomplete, abandoned, and moderator-assisted attempts; record those outcomes separately rather than dropping them from the denominator. A participant may stop or withdraw at any point. For a withdrawal, remove partial task data from the study analysis, retain no identifying notes from it, and report only the aggregate withdrawal count as an operational fact. Recruit a replacement with retained consent before setting the fixed five-participant denominator; never recast a withdrawal as a failure or incomplete completion. The Discovery Plan supplies the consent and comparison design, while Validation supplies scoring anchors. If two participants repeat a critical misunderstanding, pause expansion and revise the evidence or decision flow before measuring broader adoption.