Run a recurring metrics review that catches drift before it's a crisis
A metric that's slowly drifting wrong looks fine in any single week and alarming a quarter later.
- Outcome
- A lightweight, repeatable review cadence that catches slow drift in key metrics early, distinguishes it from normal noise, and flags it while it's still small enough to fix easily.
- Time
- 30-45 minutes per review, run on a recurring cadence
This workflow is part of PromptKit Pro
A handful of prompts and workflows stay free for everyone — Pro unlocks the full library, including this one, plus the guided runner.
A single deep-dive analysis catches whatever's happening at that moment, but slow drift — a metric quietly getting worse a little at a time — doesn't look alarming in any individual week and is easy to miss without a repeatable process specifically designed to catch it. This playbook builds a lightweight recurring review: track the metrics that matter against their normal range, distinguish real drift from noise, and flag it while it's still small and easy to fix.
When not to use this
For a one-off analysis of something that already happened, use a direct analysis approach instead — this is specifically for an ongoing, repeatable review cadence.
Did this work?
You catch a genuine slow drift while it's still small, with a clear reason to distinguish it from normal week-to-week noise.
Tested on claude-opus-5. Evidence status is draft; it moves to battle-tested only on recorded runs, never by hand.