Churn rarely happens without warning. Before a customer cancels, there are almost always signals — declining product usage, unresolved support issues, missed QBRs, falling NPS scores. The problem is that these signals are scattered across different tools and systems, making them easy to miss until it’s too late.

A customer health score aggregates those signals into a single number that gives you — and your account managers — a quick, reliable read on which accounts are thriving, which are at risk, and which need immediate intervention. When it’s built into your CRM, it becomes part of everyday account management rather than a separate exercise.

This guide walks you through how to build a health score: what inputs to use, how to weight them, how to set up the score in your CRM, and — critically — how to act on it.

What a Customer Health Score Actually Is

A health score is a composite metric that combines multiple signals from different data sources into a single number or color-coded status (green, yellow, red). The goal is to reduce the complexity of managing many accounts by surfacing which ones need attention, without requiring an account manager to manually review every data point for every account every week.

A good health score is:

• Predictive: It correlates with future outcomes (renewal vs. churn, expansion vs. contraction)
• Actionable: It points to specific things your team can do differently
• Practical: It can be calculated with data you actually have, not data you wish you had

Choosing Your Health Score Inputs

The first step is identifying which signals are most predictive of customer outcomes in your specific business. Not every input listed below applies to every business — choose the ones that fit your product and customer base.

Product Usage

Product usage is typically the most powerful health signal, especially for software businesses. Customers who use the product actively are more likely to renew and expand. Customers who have logged in rarely in the past 60 days are classic churn candidates.

Useful usage signals:

• Monthly active users (actual vs. licensed seats)
• Login frequency over the past 30/60/90 days
• Feature adoption (are customers using core features, or only logging in superficially?)
• Last active date for the primary contact

Support Ticket Volume and Sentiment

High ticket volume can signal two very different things: engaged users who have questions (positive), or frustrated users who keep hitting the same problems (negative). Context matters.

Useful support signals:

• Ticket volume over the past 60 days (high volume = investigate)
• Open tickets beyond SLA (any open tickets sitting beyond promised resolution time = red flag)
• CSAT scores on closed tickets (if your support tool captures satisfaction)
• Number of escalated or high-priority tickets

Engagement with Your Team

How often is the customer actually interacting with their account manager or customer success manager? Customers who don’t respond to emails, skip scheduled calls, or haven’t had a QBR in two quarters are showing disengagement signals.

Useful engagement signals from your CRM:

• Date of last logged activity (call, meeting, email)
• QBR completion status in the current period
• Email response rate (if your CRM tracks opens and replies)
• Champion or primary contact’s tenure (a departed champion is a significant risk signal)

NPS and Survey Scores

If you conduct Net Promoter Score surveys or periodic satisfaction surveys, these are direct signals of customer sentiment.

A Promoter (9-10) is a healthy customer. A Passive (7-8) is stable but not secure. A Detractor (0-6) is at genuine risk, regardless of how their usage data looks.

Contract and Commercial Signals

The structure of a customer’s contract can signal health risk:

• Contract renewal date (accounts with renewals in the next 90 days need heightened attention)
• Year-over-year contract value change (declining contracts are a red flag)
• Whether the customer is on a month-to-month versus annual contract

Weighting Your Health Score Inputs

Once you’ve identified your inputs, you need to assign each one a weight that reflects its relative importance as a predictor of customer health.

There are two ways to approach this:

Judgment-based weighting: Based on your team’s experience, assign weights that reflect what you believe matters most. Product usage is most important? Give it 40% of the score. Support issues matter but aren’t as critical? Give them 15%.

Data-driven weighting: Analyze your churn data. Which signals appeared most consistently in the accounts that churned last year? Which were most common among accounts that expanded? Use those patterns to weight the inputs empirically.

For most teams starting out, judgment-based weighting is practical enough. You can refine it over time as you collect outcome data.

Example Health Score Framework

Input	Weight	Scoring Scale
Product usage (MAU vs. licensed)	35%	0-100 based on % of seats active
Support ticket health	20%	100 = no open tickets; −20 per open ticket beyond SLA
Team engagement recency	20%	100 = activity within 2 weeks; decays over time
NPS / last survey score	15%	Mapped to 0-100 (Promoter = 100, Passive = 60, Detractor = 0)
Contract signals	10%	100 = annual contract >1 yr from renewal; 50 = <90 days to renewal

Apply weights:

Health Score = (Usage × 0.35) + (Support × 0.20) + (Engagement × 0.20) + (NPS × 0.15) + (Contract × 0.10)

A customer who scores 100 on all inputs would have a health score of 100. Practically, most customers will fall in the 40-80 range, with outliers at both ends.

Setting Health Score Thresholds

Once you have a composite score, you need to define thresholds that translate the number into action categories.

Score Range	Status	Meaning	Recommended Action
75-100	Green	Healthy	Identify expansion opportunities
50-74	Yellow	At risk	Schedule proactive check-in
25-49	Red	High risk	Immediate intervention
0-24	Critical	Likely to churn	Escalate to senior CSM or management

These thresholds are starting points — adjust them based on your churn patterns. If you find that Yellow accounts churn at a high rate, lower the Yellow-to-Red threshold so you’re intervening earlier.

Building the Score in Your CRM

The implementation approach depends on your CRM and what data you have available.

Option 1: Manual Score in a Custom Field

The simplest approach: create a custom field on your account record called “Health Score” (numeric, 0-100) and a separate field for “Health Status” (picklist: Green/Yellow/Red). Account managers or CSMs update these manually on a defined cadence (weekly or biweekly).

This is manual and imperfect, but it’s better than no score at all. It forces regular account reviews and creates a documented signal that managers can track over time.

Option 2: CRM Automation Rules

Many CRMs allow you to set up automation rules that update fields based on conditions. You can set rules like:

• If last activity date is more than 30 days ago, decrease health score by 20
• If any ticket is open beyond SLA, flag status as Yellow

This adds some automation without requiring a full integration. The limitation is that CRM automation usually only works with data already in the CRM — it can’t pull in product usage or NPS data automatically.

Option 3: Integration-Based Score

The most powerful approach: integrate your product analytics tool (Mixpanel, Amplitude, Segment, or equivalent) and your support tool (Zendesk, Intercom, etc.) with your CRM. With these integrations, product usage and support data flows into the CRM automatically. You can then build a calculated field or use a customer success platform (Gainsight, ChurnZero, Totango) to automate the health score calculation.

This requires more setup investment but produces a more accurate and timely score with less manual effort.

Acting on Red Health Scores

A health score is only useful if it drives action. A dashboard full of red accounts that nobody acts on is worse than no score at all — it creates the illusion of visibility without the benefit.

Build an “At Risk” Workflow

Define a specific workflow for accounts that drop into Red status:

1. Automatic CRM task created for the account owner to review within 24 hours
2. Required: schedule a call with the customer within one week
3. Required: document what the account owner believes is the root cause
4. Weekly check-in from the account owner’s manager until status improves

Diagnose Before Prescribing

When an account drops to Red, resist the urge to immediately offer a discount or concession. First, understand why the score dropped. Is it a product usage issue? A support frustration? A relationship gap? Different root causes call for different interventions:

Root Cause	Recommended Intervention
Low product usage	Schedule a guided adoption session; identify specific barriers
Unresolved support issue	Escalate to senior support; ensure resolution before next customer conversation
Disengaged champion	Re-engage with new contacts; work to expand relationships at the account
NPS detractor	Have an honest conversation; understand the specific disappointment
Upcoming renewal, no engagement	Proactive executive outreach; QBR before renewal date

Track Interventions in Your CRM

Log every intervention as an activity on the account record. Note what you did and what changed (or didn’t). Over time, this builds a record of what interventions work for which types of health issues — organizational learning that compounds as your team grows.

Frequently Asked Questions

How often should health scores be updated? It depends on your data sources. If you’re updating manually, weekly is a reasonable cadence for high-value accounts and biweekly for lower tiers. If you have automated integrations pulling in usage and support data, the score can update in near-real-time. The key is that the score should reflect the current state of the account, not the state three months ago. Stale health scores create false confidence.

Should all customers be scored? Most teams apply health scoring to their managed accounts — customers above a certain revenue threshold who have a dedicated account manager. For lower-tier or self-serve customers, full health scoring may not be cost-effective to maintain. You can use simpler signals (login activity, payment status) to flag the highest-risk accounts in your lower tiers for periodic review without full health scoring for every account.

What if my health score doesn’t predict churn accurately? This is common in the early stages of building a health score — the weights you chose may not reflect what actually drives churn in your business. Run a retrospective analysis: look at the accounts that churned in the past year and what their health scores were at 90, 60, and 30 days before churn. If churned accounts were showing Green or Yellow scores right before they left, your weights need adjustment. This is normal and expected iteration.

Can a health score be too simple? A single metric like “login frequency” is too simple — it misses too many dimensions of customer health. But a health score with 15 inputs and complex interdependencies can be confusing and hard to act on. The ideal health score is complex enough to be meaningful but simple enough that an account manager can explain it and act on it without a data science degree. Three to five weighted inputs is usually the right range for a starting framework.