Support AI Skill

Customer Health Scoring

Build and maintain customer health scores that predict churn risk, expansion potential, and support needs by combining usage, support, and business data into a unified health metric. Use when creating health scoring models, identifying at-risk customers, tr...

Customer Health Scoring for Support

Build and maintain customer health scores that predict churn risk and trigger proactive support intervention — combining support interactions, usage data, and business metrics into actionable health signals.

Workflow

Identify health score components: usage, support, business, engagement data.
Define scoring methodology and weightings for each component.
Build health scoring model (rule-based or ML-powered).
Create health score dashboard with segmentation (green/yellow/red).
Define automated actions for each health tier.
Validate model accuracy against actual churn data.
Iterate model quarterly based on new data and feedback.
Train support and CS teams to act on health score changes.

Health Score Design

HEALTH SCORE COMPONENTS
=========================

Component 1 — Support Health (40% weight):
  ════════════════════════════════════════════════════════════════════════
  Metric                          | Score Range   | Weight
  ════════════════════════════════════════════════════════════════════════
  CSAT (last 90 days avg)         | 1–5 scale     | 25%
  CES (last 90 days avg)          | 1–5 scale     | 20%
  Ticket volume trend             | Decreasing = healthy    | 15%
  Re-contact rate                 | < 15% = healthy        | 15%
  Escalation frequency            | 0 = healthy            | 10%
  SLA compliance                  | > 95% = healthy        | 10%
  NPS score                       | > 50 = healthy         | 5%
  ════════════════════════════════════════════════════════════════════════
  Support Health Score: 0–100 (higher = healthier)

Component 2 — Product Engagement (30% weight):
  ════════════════════════════════════════════════════════════════════════
  Metric                          | Score Range   | Weight
  ════════════════════════════════════════════════════════════════════════
  Login frequency                 | Weekly+ = healthy      | 20%
  Feature adoption breadth        | 5+ features = healthy  | 20%
  Usage trend (MoM)               | Growing = healthy      | 20%
  Time since last activity        | < 7 days = healthy     | 20%
  Team activation rate            | 70%+ active = healthy  | 20%
  ════════════════════════════════════════════════════════════════════════
  Engagement Score: 0–100

Component 3 — Business Health (20% weight):
  ════════════════════════════════════════════════════════════════════════
  Metric                          | Score Range   | Weight
  ════════════════════════════════════════════════════════════════════════
  Contract renewal date           | > 6 months = healthy   | 25%
  Payment history                 | No late = healthy      | 25%
  Usage vs plan capacity          | 50–80% = healthy       | 25%
  Expansion history               | Grew = healthy         | 25%
  ════════════════════════════════════════════════════════════════════════
  Business Score: 0–100

Component 4 — Relationship Health (10% weight):
  ════════════════════════════════════════════════════════════════════════
  Metric                          | Score Range   | Weight
  ════════════════════════════════════════════════════════════════════════
  QBR completion                  | Done = healthy         | 33%
  Stakeholder engagement          | 3+ contacts = healthy  | 33%
  Feedback responsiveness         | Responds = healthy     | 34%
  ════════════════════════════════════════════════════════════════════════
  Relationship Score: 0–100

COMBINED HEALTH SCORE:
  Overall = (Support × 0.40) + (Engagement × 0.30) + (Business × 0.20) + (Relationship × 0.10)
  Scale: 0–100

Health Tiers and Actions

HEALTH TIER DEFINITIONS
=========================

GREEN — Healthy (70–100):
  ════════════════════════════════════════════════════════════════════════
  Characteristics:
    → High engagement, low ticket volume, positive CSAT
    → Growing usage, paying on time, expanding account
    → Active stakeholders, responsive to communication
  
  Actions:
    → Standard support SLAs
    → Quarterly proactive check-in
    → Expansion opportunity identification
    → Case study / reference customer outreach
  
  Volume: ~60% of customer base
  Churn risk: < 5% annual

YELLOW — At-Risk (40–69):
  ════════════════════════════════════════════════════════════════════════
  Characteristics:
    → Declining engagement, increasing ticket volume, dropping CSAT
    → Stagnant usage, approaching renewal window
    → Reduced stakeholder engagement
  
  Actions:
    → Support: Proactive outreach within 48 hours
    → CS: Health review call scheduled
    → Escalation: Account flagged to CS manager
    → Investigation: Identify root cause (product issue? budget? competition?)
    → Action plan: Customized intervention based on root cause
  
  Volume: ~25% of customer base
  Churn risk: 15–30% annual

RED — Critical (0–39):
  ════════════════════════════════════════════════════════════════════════
  Characteristics:
    → Very low engagement, multiple escalations, very low CSAT
    → Usage declining sharply, payment issues, key stakeholder left
    → Contract renewal within 90 days
  
  Actions:
    → Immediate: CS + Support manager outreach within 24 hours
    → Executive: VP-level outreach for strategic accounts
    → Retention plan: Customized (pricing, features, dedicated resources)
    → Weekly check-ins: Until health improves or churn confirmed
    → Win-back: If churned, structured win-back program
  
  Volume: ~15% of customer base
  Churn risk: 50–80% annual

HEALTH SCORE TRIGGERS:
  ════════════════════════════════════════════════════════════════════════
  Trigger                           | Action
  ════════════════════════════════════════════════════════════════════════
  Score drops > 15 points in 7 days | Alert CS + Support manager
  Score drops to YELLOW             | Proactive outreach within 48 hours
  Score drops to RED                | Executive outreach within 24 hours
  CSAT drops below 3.0              | Immediate support manager follow-up
  Login gap > 14 days               | Re-engagement email sequence
  Ticket volume spikes > 50%        | Investigation + proactive call offer
  Payment failure                   | Billing specialist outreach + grace period
  Key stakeholder departure         | Identify new stakeholder + relationship building
  ════════════════════════════════════════════════════════════════════════

Health Score Dashboard

HEALTH SCORE DASHBOARD
========================

Portfolio View:
  ════════════════════════════════════════════════════════════════════════
  Health Tier   | Count  | % Total | ARR at Risk  | Trend
  ════════════════════════════════════════════════════════════════════════
  Green         | 1,200  | 60%     | —            | ↑ +3%
  Yellow        | 500    | 25%     | $750K        | ↓ -2%
  Red           | 300    | 15%     | $1.2M        | ↑ +1%
  ════════════════════════════════════════════════════════════════════════
  Total ARR     | 2,000  | 100%    | $1.95M at risk| —
  ════════════════════════════════════════════════════════════════════════

Top 10 At-Risk Accounts (Yellow → Red):
  ════════════════════════════════════════════════════════════════════════
  Account        | Score | Trend | Primary Driver          | Assigned To    | Last Action
  ════════════════════════════════════════════════════════════════════════
  Acme Corp      | 42    | ↓ 18  | Usage declining 40%     | CSM: Jane      | Call scheduled
  TechStart Inc  | 38    | ↓ 12  | CSAT dropped to 2.8     | ASM: Mike      | Manager follow-up
  GlobalRetail   | 45    | ↓ 8   | 3 escalations this month| Support: Lead  | Investigation started
  ════════════════════════════════════════════════════════════════════════

Health Score Distribution (Monthly Trend):
  ════════════════════════════════════════════════════════════════════════
  Month    | Avg Score | Green % | Yellow % | Red % | Churned
  ════════════════════════════════════════════════════════════════════════
  Jan      | 72        | 58%     | 27%      | 15%   | 12
  Feb      | 73        | 59%     | 26%      | 15%   | 10
  Mar      | 71        | 57%     | 28%      | 15%   | 14
  Apr      | 72        | 60%     | 25%      | 15%   | 9
  ════════════════════════════════════════════════════════════════════════

HEALTH IMPROVEMENT TRACKING:
  → Customers moved from Red → Yellow: 45 this quarter (target: 50)
  → Customers moved from Yellow → Green: 82 this quarter (target: 80)
  → Avg time to improvement: 34 days
  → Prevention: 68% of Yellow accounts recovered (target: 70%)

Integration Points

Customer Success Platform (Gainsight, Totango, HubSpot CS): Health score calculation, dashboard, alerting
Help Desk (Zendesk, Freshdesk): Support interaction data, CSAT, ticket volume, escalation data
CRM (Salesforce, HubSpot): Business data, contract info, payment history, stakeholder records
Product Analytics (Amplitude, Mixpanel, Pendo): Usage data, feature adoption, engagement metrics
Data Warehouse (Snowflake, BigQuery): Unified data model, health score computation
BI/Analytics (Tableau, Power BI, Looker): Health score dashboards, trend analysis, reporting
Communication (Slack, Teams): Health score alerts, escalation notifications
Email (SendGrid, Mailchimp): Automated outreach based on health tier
Billing (Stripe, Chargebee): Payment data, contract details, revenue tracking

Edge Cases

Health score is inaccurate: Model flags healthy account as at-risk (or vice versa)
Validation: Quarterly comparison of health predictions vs actual outcomes
Precision target: 80%+ accuracy (health score correctly predicts churn/non-churn)
Adjustment: Retrain model with new data; adjust weightings
Human override: CS/Support can flag false positives; reason logged for model improvement
Multi-signal: Don't rely on single metric; use combined score

Health score changes too frequently: Score fluctuates daily, causing alert fatigue
Smoothing: Use 30-day rolling average (not daily snapshot)
Threshold: Trigger only on > 10-point change (not 1-point)
Frequency: Health tier alerts sent at most once per week per account
Trending: Focus on trend (consistent decline) rather than single-day drop
Baseline: Account for seasonal patterns (lower usage during holidays)

New accounts lack history: Can't calculate health score for accounts < 30 days old
Placeholder: Assign "neutral" score (50) for first 14 days
Onboarding score: Separate onboarding health score for first 90 days
Minimal data: Use available data (fewer components) with adjusted confidence
Ramp-up: Score becomes more reliable as data accumulates (30, 60, 90 day milestones)

Health score doesn't predict actual churn: Model accuracy declining
Root cause: External factors not captured (competition, budget cuts, M&A)
Enrichment: Add external signals (news alerts, LinkedIn changes, funding news)
Feedback loop: CS team inputs qualitative signals (customer sentiment in calls)
Model refresh: Quarterly model retraining with latest outcomes
Acceptance: No model is 100% accurate; use as signal, not definitive prediction

Support actions don't improve health: Proactive outreach sent but no engagement
Channel switch: If email ignored, try phone call or in-app message
Escalation: If no response after 3 attempts, executive outreach
Value proposition: "Here's what you're missing" instead of "Are you OK?"
Acceptance: Some churn is unavoidable despite intervention
Post-churn analysis: Why didn't support intervention work?