Example prompts
What this Skill does
Measures each account's usage four ways (event count, active days, active users, and hours where session data exists) and takes the weakest 5, 10, 25, and 40% as cohorts. The cut-offs are percentiles of your own data, so "low usage" means low for your product. The stricter the band, the higher the churn usually goes.
When it fires
- An account sits in the bottom 5 / 10 / 25 / 40% of usage
- Usage = events, active days, active users, or hours
- Measured over the first 30 / 60 / 90 days or all time
Inputs
- Brazebee MCP: connected to Claude or ChatGPT with your Brazebee API key
- Product usage: Amplitude, PostHog, or Mixpanel connected in Brazebee
- Billing and CRM: Stripe and HubSpot for ARR, plan, renewal date, and account owner
How it works
- 01 / Computes usage per account for each metric and window
- 02 / Sets the threshold at the 5th, 10th, 25th, and 40th percentile
- 03 / Builds a cohort for each metric and band
- 04 / Compares each cohort's churn rate against the rest
Output
Under-Usage: fewer than 6 active days in first 30 days (bottom 10%)
Cohort: 280 accounts · $850k ARR
Churn: 11% in cohort vs 1.6% in the rest (6.9x)
Act on: 34 accounts below the line this month
Route: CSM · Webhook · "Re-engagement outreach" Example numbers. Your thresholds and rates come from your own data.
Tips
- Active days is usually the most honest metric: one heavy day can inflate event counts
- Active users shows single-user accounts, a separate churn risk
- All-time usage is biased by tenure. Prefer a fixed window (first 30/60/90 days)
