A customer health score that cannot tell you who to call this week is a decoration. A health score has one job: tell you which customers need attention this week. Three signals combined into red, yellow, green. Act on every red account within forty-eight hours. Three signals, three colors, reviewed weekly, with a forty-eight hour rule on every red account.
Churn signals appear weeks before cancellation
Customers do not churn suddenly. They disengage gradually. The signals are there weeks before the cancellation email: login frequency drops, feature usage narrows, support tickets increase or stop entirely, and the champion goes quiet. If you are tracking these signals, you can intervene before the customer decides to leave.
Build a simple health score: logins per week, features used, support tickets open, and days since last meaningful interaction. Score each customer red, yellow, or green. Review reds weekly and yellows biweekly. The intervention for a red account is a personal call from the founder or CS lead, not an automated email. Automated emails to disengaged customers accelerate the churn they are trying to prevent.
Health scores should be simple and actionable
A customer health score has one job: tell you which customers need attention this week. If your health score requires a data scientist to calculate, it is too complex. If it does not lead to a specific action, it is too abstract. The best health score is three signals combined into red, yellow, green.
The three signals that matter for most B2B products: usage frequency, breadth of feature adoption, and relationship strength. Usage frequency is how often they log in. Breadth is how many of your core features they use. Relationship is whether your champion is engaged and responsive. Weight them equally, review weekly, and act on every red account within forty-eight hours.
QBRs are for the customer, not for you
The quarterly business review is not a report on your product's usage statistics. It is a strategic conversation about the customer's business and how you are helping them achieve their goals. If your QBR is a slide deck of login counts and feature adoption rates, you are doing it wrong.
The QBR that works: thirty minutes, three topics. Topic one is the customer's goals for the quarter and how you contributed. Topic two is what is not working and what you are doing about it. Topic three is what is next on your roadmap that maps to their needs. The customer should talk more than you do. If they are not engaged in the conversation, the QBR is a waste of both your time and theirs.
Support is a product feedback channel
Every support ticket is a product decision waiting to be made. A bug report is a quality issue. A how-to question is a UX issue. A feature request is a roadmap signal. If you are only resolving tickets without categorizing and analyzing them, you are throwing away your cheapest source of product intelligence.
Categorize every ticket: bug, UX confusion, feature request, or account issue. Review the categories weekly. If twenty percent of tickets are the same UX confusion, fix the UX. If ten customers request the same feature, consider building it. Support volume is a product health metric. Rising volume means your product is getting harder to use, not that your customers are getting needier.
Customer feedback should change your roadmap
If your roadmap looks the same after three months of customer feedback, you are not listening. Customer feedback should be the primary input to your product prioritization. Not the only input, but the primary one. The customers who use your product every day know things about it that you do not.
The system: collect feedback from support tickets, sales calls, customer success conversations, and NPS surveys. Categorize it by theme. Count the mentions. When a theme reaches ten mentions from different customers, it goes on the roadmap. This is not scientific, but it is better than building what the loudest customer asked for last week.
Frequently asked questions
What is a customer health score?
A few usage and relationship signals combined into red, yellow, or green per account. Its only job is telling you which customers need attention this week. If it cannot do that, it is trivia.
What signals belong in a health score?
Usage frequency, breadth of feature adoption, and relationship strength. Weight them equally. If calculating the score needs a data scientist, it is too complex to act on, and action is the point.
How often should health scores be reviewed?
Weekly, by a person, with authority to act. A score nobody reviews is worse than none because it documents the warnings you ignored. Twenty minutes a week covers most early customer bases.
What should happen when an account goes red?
A human calls within forty-eight hours. Not an automated email: a call. Red means the pattern broke, and patterns break for reasons only a conversation surfaces. Speed matters more than the script.
Can a health score predict churn?
It predicts risk, not dates. Accounts drift yellow weeks before they consider leaving, which is exactly when a save is cheap. The score earns its keep by moving the intervention earlier.