Product usage anomaly
Daily anomaly scan across product events — flag spikes and drops with the most likely deploy or cohort cause.
Eluu — product-usage-anomaly
Shown in preview 
Runtime ~2 min per run
Tokens ~8K per run
Owner a product analyst
Works with
Pick whichever tool your team already uses.
- Messaging Slack
Teams - Spreadsheet Google Sheets
Excel - Analytics Mixpanel
Amplitude 
GA4
What it does
Product usage drops rarely get noticed for days — by the time someone runs a query, the regression is buried in a release that already shipped. This recipe runs a daily anomaly scan across every tracked event, flags ±2.5 z-score moves, cross-references the deploy log, and posts the suspects to your product-anomaly channel.
How it works
- Pull events. Daily counts for every tracked event over the last 28 days are pulled along with a trailing baseline.
- Detect. Event-level z-scores are computed; anomalies above 2.5 absolute are flagged; the last 7-day deploy log is cross-referenced; most likely cause per anomaly is suggested.
- Post. Per-anomaly mini-cards land in #product-anomaly with the deploy candidate (PR or release tag) tagged.
Setup
- Connect a product analytics tool (Mixpanel, Amplitude, or GA4).
- Connect a spreadsheet destination (Google Sheets or Excel) for the detect step.
- Connect a messaging channel (Slack or Teams).
Variations
- Run hourly for critical events instead of daily.
- Add a per-cohort breakdown (free vs paid, region).
- Auto-create a Linear ticket if the anomaly persists past 24 hours.
