Customer Success Metrics and Health Scoring: Predicting and Preventing Churn
Master customer success metrics. Build health scores to predict churn and prioritize retention efforts before customers leave.
Key Takeaways
- Health score = predictive model of churn risk; build using: usage metrics (active users, feature adoption, login frequency), engagement (tickets opened, support responses, training attended), business metrics (contract value, expansion likelihood), satisfaction (NPS, support sentiment); example: score 0-100, >80 = healthy, 50-80 = at-risk, <50 = churn risk; score should predict actual churn with 80%+ accuracy
- Usage metrics drive churn: customers who use product daily have <1% monthly churn, customers who use <1x/week have 5-10% monthly churn; implement usage tracking (dashboard logins, feature usage, API calls) to identify inactive accounts; engage inactive users: email campaigns, in-app notifications, support outreach; reactivation rate: if catch users early, can prevent 20-30% of churn
- NPS (Net Promoter Score) survey: ask 'would you recommend us?' on 1-10 scale; 9-10 = promoters, 7-8 = passives, 0-6 = detractors; NPS = (promoters − detractors) / total; benchmark: >50 is excellent, 0-30 healthy, <0 concerning; follow up with open question 'why?' to understand issues; use NPS + health score to identify at-risk segments, prioritize CS resources
Building a Customer Health Score
A health score predicts likelihood of churn, allowing you to intervene before customers leave. **Why Health Scores Matter** Example: SaaS company with 1000 customers Without health scoring: - Discover churn after customer cancels - No way to prevent it - Monthly churn: 5% (50 customers lost) - Cost: 50 × £20K LTV = £1M lost revenue With health scoring: - Identify at-risk customers 30 days before churn - Intervene with CS outreach, product improvements, discounts - Prevent 50% of at-risk churn (25 customers saved) - Monthly churn: 2.5% (25 customers lost) - Savings: 25 × £20K = £500K saved annually This is massive ROI. Building health score costs £20-50K but saves £500K+. **Components of a Health Score** Build health score from 4 categories: 1. Usage Metrics (40% weight) - Daily active users (% of user licenses using product) - Feature adoption (% using advanced features) - Login frequency (days since last login) - Session duration (minutes per session) Example scoring: - Active daily: +25 points - Active weekly: +15 points - Active monthly: +5 points - Inactive 30+ days: -10 points (churn risk) 2. Support and Engagement (20% weight) - Support tickets (quality of tickets, resolution time) - Training attendance (did customer attend onboarding?) - Feedback surveys (positive/negative sentiment) - Executive engagement (does customer meet with your team?) Example scoring: - No support tickets in 3 months: -5 points (low engagement) - Support tickets resolved >95%: +10 points (success) - Positive NPS: +15 points - Negative NPS: -20 points 3. Business Metrics (25% weight) - Contract value (higher value = more churn risk monitoring) - Growth trajectory (is customer expanding?) - Usage growth (increasing usage month-over-month) - Payment health (on-time payment, not declining) Example scoring: - Payment always on time: +10 points - Payment late 30+ days: -15 points (financial stress, churn risk) - Expansion revenue this quarter: +15 points - Declining usage trend: -20 points 4. Competitive/Market Factors (15% weight) - Competitor mentions (did customer mention competitor in survey?) - Contract renewal date (imminent renewals = risk) - Market disruption (is customer's industry changing?) - Reference-ability (willing to be customer reference?) Example scoring: - Within 30 days of renewal: +5 points (attention needed) - Within 7 days of renewal: -10 points (critical) - Competitor mentioned: -20 points - Willing to be reference: +10 points **Calculating Health Score** Total score: Sum of weighted components (0-100) Example customer calculation: Usage (40% weight): - Active daily (+25), feature adoption (+15) = 40 points × 40% = 16 points Engagement (20% weight): - Recent support tickets (+10), positive feedback (+10) = 20 points × 20% = 4 points Business (25% weight): - On-time payment (+10), expansion revenue (+10) = 20 points × 25% = 5 points Competitive (15% weight): - 45 days to renewal (+5), no competitor mention (0) = 5 points × 15% = 0.75 points Total score: 16 + 4 + 5 + 0.75 = 25.75 (out of 100) Wait, this seems low. Let me recalculate with better baseline scoring. Better approach: Start at 50 (neutral) and adjust: Base score: 50 Usage factors: - Active last 7 days: +20 - Active last 30 days: +10 - Inactive 30+ days: -20 - Total adjustment: +20 Engagement factors: - NPS >7: +15 - NPS <5: -20 - Support tickets in good health: +10 - Total adjustment: +15 Business factors: - On-time payment: +10 - Contract renewal <30 days: -5 - Expansion revenue: +10 - Total adjustment: +15 Final score: 50 + 20 + 15 + 15 = 100 This customer is very healthy (100/100). Another customer: - Inactive 30+ days: -20 - No support tickets (low engagement): -10 - Payment late: -15 - Renewal in 10 days: -10 - Final score: 50 - 20 - 10 - 15 - 10 = -5 (below 0, should cap at 0) This customer is at very high churn risk. **Health Score Thresholds** Define action levels: Score >80: Healthy - No action needed - Monitor quarterly - Consider for upsell/expansion Score 60-80: At-risk - Monitor monthly - Check in via email/call - Identify pain points - Offer additional training Score 40-60: High-risk - Weekly monitoring - Immediate CS outreach - Escalate to leadership - Offer customer success call Score <40: Critical - Daily monitoring - Emergency CS outreach - Offer immediate support - Consider discounts/extended trial to save **Implementing Health Scoring** Step 1: Choose metrics (2-3 months data collection) - Pull historical data on usage, engagement, churn - Test which metrics predict churn best - Iterate until accuracy >80% Step 2: Assign weights (1-2 weeks) - Rank metrics by churn correlation - Highest correlation = highest weight - Test different weight combinations Step 3: Automate calculation (2-4 weeks engineering) - Build dashboard or use integration (many tools have health scores) - Calculate score weekly or monthly - Set alerts when score drops Step 4: Implement CS processes (ongoing) - Define actions for each score level - Train CS team on thresholds - Create playbooks for at-risk customers Step 5: Monitor and improve (quarterly) - Compare predicted churn vs. actual churn - Adjust weights if accuracy drifts - Add new metrics if discovered Cost: - Tools (Gainsight, Totangi, Planhat): £5-20K/month - Or custom build: £20-50K one-time + £2K/month maintenance - Payoff: If prevent 20-30% of churn, saves £500K+ annually **NPS (Net Promoter Score)** Definition: Measures customer satisfaction with one question. Survey: "On a scale of 0-10, how likely are you to recommend [Company] to a colleague?" Scoring: - 9-10 = Promoters (loyal, will refer) - 7-8 = Passives (satisfied but not loyal) - 0-6 = Detractors (unhappy, will discourage others) NPS = (# Promoters − # Detractors) / Total responses × 100 Example: - 100 responses - 50 promoters (9-10) - 30 passives (7-8) - 20 detractors (0-6) - NPS = (50 − 20) / 100 × 100 = 30 Benchmarks: - <0: Poor (high churn) - 0-30: Good (acceptable) - 30-50: Very good (healthy) - >50: Excellent (industry-leading) Use NPS in health score: - Promoters: +15 points - Passives: +0 points - Detractors: -20 points Follow-up questions: - "Why did you give this score?" - "What could we improve?" - "What do you value most?" Use feedback to improve product and identify issues. **At-Risk Customer Playbook** When health score drops or customer is at-risk: Week 1: Triage - Call customer immediately (within 48 hours) - Ask: "How are things going?" - Listen for issues - Document findings Week 2: Assess - Understand root cause (product issue? Cost? Competitor?) - Identify if fixable or not - Get buy-in from leadership Week 3: Action - If fixable: Propose solution (extra training, features, custom work) - If competitor threat: Demonstrate differentiation, offer trial of new features - If cost: Discuss pricing options, discounts, extended trial Week 4: Follow-up - Check if issues resolved - Offer additional support - Schedule check-in for future Success rate: - If intervene when score drops from 70 to 60: 70% can be saved - If intervene when score at 40: 30% can be saved - If intervene when cancellation requested: 10% can be saved Early intervention is key. **Predicting Churn with Cohort Analysis** Build churn prediction by customer cohort: Cohort = customers acquired in same month Example: January 2024 cohort (100 customers): - Month 0 (Jan): 100 customers (base) - Month 1 (Feb): 98 customers (-2% churn) - Month 2 (Mar): 94 customers (-2% churn) - Month 3 (Apr): 92 customers (-2% churn) - Month 12 (Dec): 77 customers (-15% cumulative churn) This cohort has 2% monthly churn. Compare across cohorts: January 2024 cohort: 2% monthly churn February 2024 cohort: 3% monthly churn (worse) March 2024 cohort: 4% monthly churn (even worse) Question: Why is March cohort churning faster? - Product change in March? (feature removed, UX change) - Sales quality changed? (lower-quality leads) - Market change? (economic downturn) Investigate and fix before April cohort becomes worse. This is how you use data to improve retention.