How to Build a Customer Health Score from Scratch

Why You Need a Health Score

Every customer success team eventually faces the same problem: you have hundreds (or thousands) of accounts, and you don't know which ones need attention. A customer health score solves this by distilling complex behavioral data into a single, actionable number.

But most health scores fail. They're either too simplistic (based on a single metric like login count) or too complex (a black box that nobody trusts). Here's how to build one that actually works.

Step 1: Define Your Outcome

Before choosing inputs, define what you're predicting. For most SaaS companies, the primary outcome is churn within 90 days. But consider also tracking:

• Contraction risk (downgrade within 60 days)
• Expansion readiness (likely to upgrade within 90 days)
• Advocacy potential (likely to refer or leave a positive review)

Start with churn prediction. You can layer in the others once the foundation is solid.

Step 2: Identify Your Input Signals

Group your signals into four categories:

Product Usage (weight: 40%)

• Login frequency (daily/weekly/monthly trend)
• Core feature adoption (are they using the features that matter?)
• Session depth (how much do they do per visit?)
• Usage trend (increasing, stable, or declining over 30 days)

Customer Engagement (weight: 25%)

• Support ticket volume and sentiment
• NPS/CSAT responses
• Email engagement (opens, clicks)
• Webinar/training attendance

Business Signals (weight: 20%)

• Payment history (failures, late payments)
• Plan changes (upgrades vs. downgrades)
• Active seat count trend
• Contract renewal date proximity

Relationship Quality (weight: 15%)

• Executive sponsor identified and engaged
• Multi-threaded contacts (more than one champion)
• QBR attendance and engagement
• Feature request activity (engaged customers ask for things)

Step 3: Score Each Signal

For each signal, create a 0-100 scale. Here's an example for login frequency:

• Daily active (90-100): Customer logs in 5+ days/week
• Regularly active (70-89): 2-4 days/week
• Occasionally active (40-69): A few times per month
• Rarely active (10-39): Once a month or less
• Inactive (0-9): No logins in 30+ days

Apply the same logic to every signal. The key is making scoring criteria explicit and objective — no gut feelings.

Step 4: Weight and Combine

Multiply each signal score by its category weight, then sum to get the composite health score. Start with the weights above and adjust based on what your data tells you.

Example calculation:

• Product Usage: 72/100 x 0.40 = 28.8
• Engagement: 55/100 x 0.25 = 13.75
• Business Signals: 80/100 x 0.20 = 16.0
• Relationship: 65/100 x 0.15 = 9.75
• Composite Score: 68.3 / 100

Step 5: Validate Against Outcomes

This is where most teams stop — and where the real work begins. Compare your health scores against actual outcomes:

• Did low-scoring customers actually churn?
• Did high-scoring customers stay and expand?
• Which signals were most predictive?

Run this analysis monthly. Adjust weights based on predictive power. A good health score should correctly identify 70%+ of churn cases.

Step 6: Operationalize

A health score is useless if nobody acts on it. Create automated workflows:

• Score drops below 40: Alert CSM, trigger personal outreach
• Score drops below 25: Escalate to leadership, executive touch
• Score rises above 80: Flag for expansion opportunity
• Score increases 20+ points: Celebrate with the team (reinforces good behaviors)

Common Mistakes to Avoid

1. Too many signals. Start with 6-8 strong signals. You can add more later.
2. Equal weighting. Not all signals matter equally. Usage almost always matters most.
3. No validation. If you don't check predictions against outcomes, you're guessing.
4. Static scores. Health changes daily. Recalculate at least weekly.
5. Ignoring context. A startup with 3 users will look different from an enterprise with 300. Segment your scoring.

Or Let ChurnRate.io Do It For You

Building and maintaining a health score is a significant investment. ChurnRate.io does all of this automatically — we ingest your data, build a custom scoring model for your business, validate it against outcomes, and improve it continuously. You get actionable health scores on day one.