Build a customer health scoring rubric for {{segment, e.g. enterprise SaaS, mid-market PLG}}.
Inputs:
- What signals we have access to: {{paste — product usage, NPS/CSAT, support tickets, payment behavior, contract status, exec engagement}}
- Past churn data if available: {{paste — what signals preceded churn historically}}
- What we want the score to drive: {{paste — CSM intervention, renewal forecasting, expansion targeting}}
Output a scoring rubric:
## Component weights
| Component | Weight | Source | Why it matters |
|---|---|---|---|
| Product adoption | 30% | usage analytics | leading indicator |
| Engagement breadth (# active users) | 20% | usage | sponsor-risk proxy |
| Support sentiment | 15% | ticket volume + CSAT | trailing indicator |
| Executive engagement | 15% | meeting attendance, exec NPS | political risk |
| Renewal proximity × satisfaction | 10% | contract data + qual | time-bounded |
| Payment behavior | 10% | invoicing | hard signal |
(Adjust weights for the segment. Justify each.)
## Scoring bands
- **Green** (80–100): healthy, expansion candidate
- **Yellow** (60–79): watch list, CSM should act this month
- **Orange** (40–59): at-risk, needs exec intervention
- **Red** (<40): churn likely, special handling
## Specific thresholds per component
Concrete numbers. E.g. "Product adoption < 40% of seats logged in last 30 days = 10/30 points." Not "low usage = low score."
## What the rubric will get wrong
Honest.
- Customers who score green but actually churn (likely cause)
- Customers who score red but renew (likely cause)
- Where the rubric will go stale (which weight will need rebalancing in 6 months)
## Validation plan
- Backtest against the last 12 months of churned + renewed accounts
- Expected predictive accuracy
- The threshold confusion that would trigger a re-weight
Hard rule: a rubric where 80% of accounts score green isn't a health score — it's wallpaper. Calibrate the bands so they discriminate.customer-successchurnrubrics