Review product metrics with AI
Compare a change with its baseline, check which group accounts for it and distinguish a useful hypothesis from something the data actually establishes.
Review product metrics with AI when a headline figure has changed but the next decision is unclear. More sign-ups and fewer first purchases could point to an awkward checkout or simply to a different mix of new visitors.
The metrics-review workflow gives an agent a structure for examining those alternatives. A comparison is only useful if the periods, definitions and eligible users match. Someone who joined yesterday cannot yet count towards a seven-day retention outcome.
What it does
- Compare two periods under the same metric definition.
- Check which cohort accounts for an activation change.
- Put the product’s main measure beside its target.
- Separate a change in acquisition from a change in engagement.
- Identify the next piece of evidence needed for a decision.
Install and use
- Supply aggregate figures, definitions, dates and relevant product events. A table is sufficient; direct reads need the maker’s Product Management plugin and authorised tools.
- Try: “Sign-ups rose but first purchases fell. Compare the two complete weeks and tell me which explanations these figures cannot distinguish.”
- Check an important result against the source. Include counts as well as rates before suggesting an investigation.
FAQ
Source
Made by Anthropic. Original skill: metrics-review, published on GitHub under the Apache-2.0 licence.
View on GitHub