Course-aligned lesson expansion from original D15. Access requirement: paid Part 1 or Bundle.
Part 1 | Week 3 production chain | deliverable_production | 15-/30-minute paths
Turn a prompt case library into a usable metric tracking plan with reader, use, evidence, owner, cadence, and done standard.
A manager says: “We approved a reusable prompt case. How will we know whether it actually improves the work?”
It sounds like you should list impressive metrics. The trap is that a metric is useless unless it has evidence, owner, cadence, and a decision rule.
“Track engagement, quality, time saved, and user feedback.”
The weak version contains engagement, no evidence location, no owner, no decision rule.
What will show the work changed?
Where will the proof come from?
Who updates and reviews it?
What result changes the next action?
Use this rule before filling the production table. It turns the day from abstract AI advice into a work-ready habit.
What will show the work changed?
Where will the proof come from?
Who updates and reviews it?
Your manager does not need a polished-looking document. They need a usable deliverable: a small metric tracking plan that says what to track, where evidence comes from, who updates it, when it is reviewed, and what result changes the next action.
A team is testing whether reusable prompt cases from D14 improve customer response drafts and internal status briefs. D15 turns that library into a tracking plan the manager can actually review.
D14 manager-readable prompt case library
D16 evidence_pack.md
Learning checkpoint: Use the previous artifact as working input, produce today's artifact, and hand it forward with evidence, limitation, owner, and repair needs visible.
A manager asks, 'How will we know this deliverable worked?' A weak answer lists tasks completed. A useful answer names reader, decision, metric signal, evidence location, owner, cadence, and a done standard.
A manager asks, 'How will we know this deliverable worked?' A weak answer lists tasks completed. A useful answer names reader, decision, metric signal, evidence location, owner, cadence, and a done standard.
Weak: 'Track engagement.' Repaired: 'Support lead checks reply rate and escalations every Friday; pass if fewer than three unclear-owner replies.'
Weak: 'Track engagement.' Repaired: 'Support lead checks reply rate and escalations every Friday; pass if fewer than three unclear-owner replies.'
A deliverable is usable when five anchors are visible: reader, decision/use, form, evidence, and done standard.
| Anchor | Status | Evidence or repair note |
|---|---|---|
| Reader | ||
| Decision/use | ||
| Form | ||
| Evidence | ||
| Done standard |
“Here are possible metrics: views, clicks, signups, feedback, engagement.” It looks useful, but nobody knows evidence, owner, cadence, or what action changes.
“For the 7-day prompt-case test, track customer draft revision count, manager correction count, and time to first usable brief. Review after five samples.”
Core path: define three metrics. Deep path: define five and include a decision rule for each.
| No. | Metric / evidence signal | Reader | Evidence used | Owner | Evidence readiness | D16 evidence question | Cadence | Decision rule / done standard |
|---|---|---|---|---|---|---|---|---|
| 1 | ||||||||
| 2 | ||||||||
| 3 |
Do not call a deliverable complete if it has no reader, decision/use, evidence used, owner, cadence, or done standard.
Each metric has a evidence, owner, cadence, and decision use.
The plan is a pile of impressive metrics without collectable evidence or a decision rule.
Use the metric rows above as a per-metric evidence-needs list for D16. D16 should collect or verify evidence access, owner, status, and the open question for each metric.
Use this pass to find the row that could confuse someone, create unsafe AI use, or fail as workplace evidence before you export.
Weak row: Metric: engagement.
Repair move: Add evidence source, owner, cadence, and done standard.
Better row: Support lead checks customer draft revision count every Friday; continue if five samples need fewer than two manager corrections.
Choose the row you would be comfortable showing to a manager. It should have a clear task, boundary, owner, and next action.
Choose the row most likely to fail, confuse someone, expose sensitive input, or need review.
Write the exact change. Sentence starter: The weak row is weak because ___. I will repair it by adding ___ before using AI.
This file is today’s work receipt. It should show what changed in your work habit, not just that you filled a table.
What this artifact proves: This file proves you can turn a prompt case into a trackable metric plan.
Weak export: broad metrics like engagement with no source, owner, cadence, or decision rule.
Good export: one metric tied to evidence source, owner, review cadence, and a done standard that changes the next action.
Click Generate Markdown to create metric_tracking_plan_pack.md.
D15 starts the deliverable production chain and hands evidence needs to D16.