Customer Health Scores
Understand customer health scores, AI-powered intelligence, payment failure predictions, and churn risk analysis
Customer Health Score
RevKeen automatically calculates a health score (0--100) for each customer based on their payment behaviour. The score helps you identify at-risk customers before they churn and prioritise outreach to accounts that need attention.
How the Score Works
The health score is a weighted composite of five components:
| Component | Weight | What it Measures |
|---|---|---|
| Punctuality | 30% | Ratio of invoices paid on time and average days to pay |
| Dunning | 25% | How deep a customer goes into dunning and whether they self-recover |
| Payment Failures | 20% | Failed transaction rate over the last 6 months, with extra penalty for hard declines and fraud flags |
| Tenure | 15% | How long the customer has been active, plus score trend over time |
| Method Health | 10% | Whether the customer's default payment method is valid and not expiring soon |
Risk Levels
Each customer is assigned a risk level based on their overall score:
| Score Range | Risk Level | Meaning |
|---|---|---|
| 80--100 | Excellent | Reliable payer, no action needed |
| 60--79 | Good | Generally healthy, minor issues possible |
| 40--59 | Needs Monitoring | Some payment problems -- consider proactive outreach |
| 0--39 | High Risk | Significant payment issues -- immediate attention recommended |
Where to View Health Scores
- Dashboard home -- The Customer Health summary card shows the average score across all customers and a breakdown by risk level
- Customer list -- Each customer displays a colour-coded health score pill next to their name
- Customer detail -- The Health Score widget shows the overall score, risk level, trend direction, and a breakdown of all five component scores with a Recalculate button for on-demand updates
Recalculation Schedule
Health scores are automatically recalculated daily at 03:00 UTC for all customers with invoice history. You can also trigger an immediate recalculation for any individual customer from their detail page.
Trend Tracking
RevKeen tracks score changes over time and labels each customer with a trend:
- Improving -- Score increased by 5+ points since last calculation
- Stable -- Score within 5 points of last calculation
- Declining -- Score dropped by 5+ points since last calculation
Use the trend indicator to catch customers whose payment behaviour is deteriorating before they reach the high-risk threshold.
Customer Intelligence
Customer Intelligence is an AI-powered layer built on top of Health Scores. While Health Scores give you a numeric snapshot, Customer Intelligence provides natural-language explanations, forward-looking predictions, and actionable recommendations for every customer.
AI-Powered Narratives
Each customer receives an automatically generated narrative that explains their health score in plain English. The narrative summarises payment behaviour, highlights recent changes, and calls out specific risk factors or positive trends.
Narratives appear in the Intelligence Panel on the customer detail page. You can click Regenerate to request a fresh narrative at any time -- useful after a customer makes a large payment or resolves an overdue invoice.
Payment Failure Predictions
RevKeen analyses a 20-feature vector for each customer to predict the likelihood of their next payment failing. Features include:
- Recent failure rate and hard-decline history
- Payment method age and expiration proximity
- Average days to pay and dunning depth
- Transaction amount volatility and frequency patterns
Each prediction returns a failure probability (0--100%), a risk level (low, medium, or high), the top contributing risk factors, and a recommended action such as updating the payment method or switching to manual invoicing.
Churn Predictions
A separate model assesses the probability of each customer churning within the next billing cycle. The churn signal vector includes:
- Health score trend direction and velocity
- Subscription downgrade or cancellation signals
- Payment failure frequency over the last 90 days
- Engagement patterns (portal logins, support interactions)
Churn predictions include a churn probability, risk level, risk signals, and a recommended action to retain the customer.
Unified Intelligence Panel
The customer detail page includes an Intelligence Panel that brings together all three outputs:
- Narrative -- the AI-generated health summary
- Payment risk -- failure probability, risk level, and top factors
- Churn risk -- churn probability, risk level, and key signals
The panel updates automatically after each scheduled prediction run.
Intelligence Summary Card
The dashboard home page includes an Intelligence Summary card that aggregates predictions across all customers:
- Average health score and distribution breakdown
- Number of customers at risk
- Predicted payment failures in the next 7 days
- Customers with high churn probability
Use this card to quickly identify which customers need attention.
Prediction Schedule
| Prediction | Schedule | Notes |
|---|---|---|
| Health narratives | Daily at 03:30 UTC | Runs after health score recalculation |
| Payment failure | Daily at 04:00 UTC | Covers all customers with active payment methods |
| Churn | Monthly at 05:00 UTC (1st) | Assessed once per billing cycle |
All predictions can also be triggered on-demand from the customer detail page.
Related
- Customers -- Customer profiles, payment methods, and custom fields
- Dunning -- Automated payment recovery that feeds into health scores
- Customer Portal -- Self-service portal for customers