Account Health Scoring
Build a composite health score from usage, engagement, support, and adoption signals to predict churn risk and expansion opportunity across all accounts.
npx gtm-skills add product/retain/health-score-dashboardOutcome
Score 20+ accounts with >=70% classification accuracy on known outcomes
Leading Indicators
- Score accuracy vs known outcomes
- Dimension signal quality
- Back-test churn correlation
Instructions
-
Verify PostHog has 30+ days of account-level usage data with Group Analytics enabled.
-
Select 20 accounts from Attio: mix of known-healthy, known-at-risk, and unknown.
-
Label each account before scoring to prevent bias.
-
Manually compute 4-dimension health scores: usage (35%), engagement (25%), support (20%), adoption (20%).
-
Use percentile ranking against the 20-account cohort for normalization.
-
Compute composite score (0-100) and classify risk tier.
-
Validate: do >=70% of known at-risk accounts fall into At Risk or Critical tiers?
-
If accuracy <70%, adjust dimension weights and re-score.
-
Document final model weights and scoring functions.
-
If passed, proceed to Baseline with validated model.
Recommendations
Time
8 hours over 1 week
Play-specific cost
Free