---
name: customer-health-scoring
description: Calculate and monitor customer health scores using usage data, support signals, billing status, and engagement metrics to identify at-risk accounts and expansion opportunities. Use when building health score models, monitoring account health, triggering health-based workflows, designing CSM playbooks, or improving customer retention through proactive health management. Triggers on phrases like "health score", "account health", "customer health", "health monitoring", "health dashboard", "risk scoring", "engagement scoring", "customer success scoring", "health tier".
---

# Customer Health Score Monitoring

Calculate and monitor customer health scores to identify at-risk accounts and expansion opportunities.

## Workflow

### Health Score Model Design

Trigger: Model creation; quarterly model review; business metric change:

1. **Signal identification**: Map all customer health signals — usage (login frequency, feature adoption, session duration, key actions), support (ticket volume, CSAT, escalation rate, sentiment), billing (payment status, subscription changes, arrears), engagement (NPS, survey responses, community activity, training completion), growth (user count change, seat utilization, API calls, data volume).
2. **Weight assignment**: Assign weights to each signal category based on predictive power (correlation with churn/expansion) — usage 30%, support 20%, billing 20%, engagement 15%, growth 15%; validate weights against historical churn data.
3. **Scoring algorithm**: Define score calculation — normalize each signal to 0–100; apply weights; calculate weighted average; classify into tiers (Green 70–100, Yellow 40–69, Red 0–39); configure real-time vs. daily calculation.
4. **Threshold calibration**: Set tier thresholds based on historical data — % of churned customers in each tier; optimize to maximize true positive rate (churned customers in Red tier) while minimizing false positives.
5. **Dashboard and alerting**: Build health score dashboard — account list with color-coded tiers, trend chart, drill-down to signal details; configure alerts (Green → Yellow: CSM review; Yellow → Red: intervention required; Red → any: executive escalation).
6. **Playbook integration**: Link health tiers to CSM playbooks — Green: expansion opportunities; Yellow: proactive engagement; Red: retention intervention; configure automated triggers.
7. **Validation and launch**: Back-test model against last 12 months of churn data; calculate precision and recall; adjust weights; launch with CSM training.
8. **Continuous improvement**: Monthly model accuracy review; quarterly weight adjustment; annual algorithm review; add new signals as data becomes available.

### Health Score Calculation

```
CUSTOMER HEALTH SCORE — CALCULATION MODEL
============================================

Signal Categories and Weights:

1. Usage Score (30%)
   Metrics:
     - Login frequency (30-day avg vs. 90-day baseline): +10 if stable or up, -20 if down > 50%
     - Active features (count of features used in last 30 days): +15 if ≥ 70% of available, -15 if < 40%
     - Key action completion (critical workflows completed): +10 if ≥ 80% complete, -10 if < 50%
     - Session duration trend: +5 if stable or up, -10 if down > 30%
   Score range: 0–100

2. Support Score (20%)
   Metrics:
     - Ticket volume (30-day vs. baseline): +15 if low/stable, -20 if spike > 2×
     - CSAT average (last 5 tickets): +20 if ≥ 4.0, -15 if < 3.0
     - Escalation rate: +10 if 0 escalations, -15 if ≥ 2 escalations
     - Sentiment trend (ticket analysis): +15 if positive/stable, -15 if negative trend
   Score range: 0–100

3. Billing Score (20%)
   Metrics:
     - Payment status: +30 if current, -30 if failed payment
     - Subscription changes: +10 if upgrade/stable, -20 if downgrade request
     - Arrears: +20 if $0 outstanding, -20 if > 30 days overdue
     - Contract timeline: +10 if > 6 months to renewal, -10 if < 30 days to renewal
   Score range: 0–100

4. Engagement Score (15%)
   Metrics:
     - NPS score: +20 if ≥ 9 (promoter), -15 if ≤ 6 (detractor)
     - Survey response rate: +10 if responsive, -10 if non-responsive (3+ surveys)
     - Community participation: +10 if active, -5 if inactive
     - Training completion: +10 if ≥ 80% complete, -10 if < 40% complete
   Score range: 0–100

5. Growth Score (15%)
   Metrics:
     - User count change: +20 if growing, -15 if declining
     - Seat utilization: +10 if ≥ 70%, -10 if < 40%
     - API call volume: +10 if growing, -10 if declining
     - Data volume: +10 if growing, -10 if declining
   Score range: 0–100

Overall Health Score = (Usage × 0.30) + (Support × 0.20) + (Billing × 0.20) + (Engagement × 0.15) + (Growth × 0.15)

Health Tiers:
  🟢 Green (70–100): Healthy — expansion opportunities, low churn risk
  🟡 Yellow (40–69): At-risk — proactive engagement needed
  🔴 Red (0–39): Critical — immediate intervention required

Example Calculation:
  Usage: 65 (login down 30%, features at 55%)
  Support: 50 (ticket spike, CSAT 3.2)
  Billing: 80 (current, no issues)
  Engagement: 40 (NPS 6, non-responsive to surveys)
  Growth: 30 (user count down 15%, seat utilization 35%)

  Health = (65 × 0.30) + (50 × 0.20) + (80 × 0.20) + (40 × 0.15) + (30 × 0.15)
  Health = 19.5 + 10 + 16 + 6 + 4.5 = 56 → 🟡 YELLOW (At-risk)
```

### Health-Based Playbook Triggers

```
HEALTH SCORE — AUTOMATED TRIGGERS
===================================

Trigger 1: Green → Yellow (score drops below 70)
  Action: CSM notified (Slack + email)
  Timeline: 7 days to outreach
  Playbook: "Check-in call — understand what changed, offer help"
  Data provided: Score breakdown, trend chart, specific signal changes
  Success metric: Score returns to Green within 30 days

Trigger 2: Yellow → Red (score drops below 40)
  Action: CSM + Account Executive notified (Slack + email + call)
  Timeline: 48 hours to outreach
  Playbook: "At-risk intervention — identify root cause, create action plan"
  Data provided: Full account analysis, churn risk factors, competitive intel
  Success metric: Score returns to Yellow within 14 days; churn prevented

Trigger 3: Red → Any (score rises above 40)
  Action: CSM notified (email)
  Timeline: Monitor for 30 days (score may fluctuate)
  Playbook: "Recovery monitoring — ensure improvement is sustainable"
  Data provided: Score trend, improved signals, remaining risks
  Success metric: Score stable in Yellow or Green for 60+ days

Trigger 4: Green for 90+ days (consistently healthy)
  Action: AE + CSM notified (email)
  Timeline: Quarterly business review
  Playbook: "Expansion opportunity — upsell, cross-sell, reference"
  Data provided: Usage growth, feature adoption, success metrics
  Success metric: Expansion revenue, reference provision, case study

Trigger 5: Score unchanged in Red for 30+ days
  Action: VP Customer Success + Sales notified (executive escalation)
  Timeline: Executive call within 7 days
  Playbook: "Executive intervention — personal relationship, special offer"
  Data provided: Full account history, intervention attempts, churn probability
  Success metric: Account retained or graceful offboarding completed
```

## Edge Cases

- **New customer health scoring** (insufficient data for meaningful score in first 30–60 days):
  - Challenge: New accounts have limited usage history, no support tickets, no survey responses
  - Approach: Use onboarding-specific metrics — setup completion %, first key action time, training attendance, initial engagement
  - Scoring: Separate "Onboarding Health" score (first 60 days) → transitions to standard health score
  - Thresholds: More lenient thresholds for new accounts; focus on trajectory (improving vs. declining)
  - CSM action: Proactive check-in at day 7, 14, 30 regardless of score

- **Enterprise vs. SMB health differences** (different signals matter for different account sizes):
  - Enterprise: More signals available (multiple users, complex usage, dedicated CSM); health score more granular
  - SMB: Fewer signals (single user, limited usage data, self-service); health score more binary
  - Approach: Segment health models by account tier; different weights and thresholds per segment
  - Example: Login frequency matters more for SMB (single user); user count growth matters more for Enterprise
  - Metric: Model accuracy measured separately per segment; target > 70% churn prediction accuracy

- **Seasonal usage patterns** (legitimate usage drops during holidays/season):
  - Detection: Historical pattern comparison (same period last year); industry seasonality data
  - Approach: Seasonal adjustment in scoring — compare to same period last year, not absolute baseline
  - Communication: "Usage is down but consistent with seasonal pattern — score adjusted"
  - Prevention: Annual seasonality profile per customer; automatic score adjustment
  - Example: Retail customers — low usage in January, high in November/December

- **Health score disputes** (CSM believes score doesn't reflect reality):
  - Cause: Data lag; missing context (customer on vacation, planned migration); signal misinterpretation
  - Resolution: Allow manual score override with justification; log override reason; review overrides monthly
  - Prevention: Improve data freshness; add "context" field for CSM to note account-specific factors
  - Governance: Quarterly review of override patterns; adjust model if overrides reveal systematic issues
  - Metric: Override rate (target < 10% of accounts); high override rate indicates model needs improvement

- **False positive health alerts** (score drops temporarily but customer is fine):
  - Cause: One-time event (conference, travel, system migration); data sync delay
  - Detection: Alert + CSM feedback loop ("Customer is fine, score is wrong")
  - Prevention: Require 2 consecutive periods of low score before alert (not single-period spike)
  - Mitigation: Alert includes "context requested" — CSM confirms before intervention
  - Metric: Alert accuracy rate (true alerts / total alerts); target > 70%

## Integration Points

- **CRM**: Salesforce, HubSpot — account data, contact information, renewal dates
- **Product analytics**: Mixpanel, Amplitude, Segment — usage data, feature adoption, session metrics
- **Help desk**: Zendesk, Freshdesk, Intercom — support history, CSAT, sentiment analysis
- **Billing**: Stripe, Chargebee, Recurly — payment status, subscription changes, revenue data
- **Customer success**: Gainsight, Totango, Planhat — health score platform, playbook automation
- **Survey tools**: Delighted, SurveyMonkey — NPS, CSAT, engagement surveys
- **Community**: Discourse, Circle.so — community participation, engagement data
- **ML platform**: DataBricks, SageMaker — health score model training, prediction
- **Workflow automation**: Zapier, Make — automated triggers, CSM notifications
- **Data warehouse**: Snowflake, BigQuery — unified data aggregation, historical analysis