Customer Health Scoring for SaaS: How to Build a System That Predicts Churn

Most SaaS founders find out a customer is churning when they see the cancellation email. By that point, the customer had already decided weeks ago — the signals were there, just not organized into anything actionable.

A customer health score converts scattered product signals into a single number per account that tells you, in real time, which customers are thriving and which ones are trending toward the exit.

What Is a Customer Health Score?

A customer health score is a composite metric — typically on a scale of 0–100 — that aggregates multiple behavioral and relationship signals into a single indicator of retention risk or expansion potential.

It answers one question: based on what this customer is doing (and not doing) right now, how likely are they to stay, expand, or churn?

Unlike churn rate, which measures what already happened, a health score is a leading indicator — a prediction engine. The goal is to surface at-risk accounts early enough to intervene.

The 5 Core Components of a Health Score

No single signal is reliable in isolation. Health scoring works by combining signals across five dimensions:

1. Product Engagement

The single most predictive input for retention. Customers who use your product regularly are getting value; customers who don't are candidates for churn.

Key sub-signals:

• Login frequency (sessions per week vs. baseline for their cohort)
• Core feature adoption (are they using the 1-3 features that drive retention?)
• Breadth of usage (number of features used — not as important as depth in core features)
• Recency (last login date — sudden drop-off is an early warning)

Weight recommendation: 35–45% of total score

2. Activation Completion

Customers who never completed onboarding churn at 3–5x the rate of fully activated users. If your activation rate is tracked at the cohort level, activation status per account belongs in the health score.

Key sub-signals:

• Onboarding checklist completion %
• Aha moment achieved (yes/no)
• Integration connected (yes/no — integrations correlate strongly with retention)
• Team seats used vs. licensed (adoption breadth for multi-seat plans)

Weight recommendation: 15–20% of total score

3. Support and Sentiment Signals

Support tickets are a double-edged signal: some are a sign of engagement (customers who care enough to ask for help), but unresolved or high-volume support is a churn predictor.

Key sub-signals:

• Open support tickets (especially unresolved > 5 days)
• Support ticket volume trend (increasing = risk)
• NPS score and trend (most reliable for detecting dissatisfaction)
• Last NPS response recency (stale NPS is a risk signal itself)

Weight recommendation: 15–20% of total score

4. Commercial Health

Payment behavior and plan-level data predict churn in ways product data misses.

Key sub-signals:

• Payment status (failed payment = immediate red flag)
• Days until renewal (combined with engagement score, predicts renewal risk)
• Plan tier (customers on minimum plan with low usage are highest churn risk)
• Discount applied (customers who only joined during a promo have higher price sensitivity)

Weight recommendation: 10–15% of total score

5. Relationship and Success Signals

Especially relevant for accounts with a named CSM or AE relationship.

Key sub-signals:

• Last customer interaction date (CSM call, email response)
• Executive sponsor engaged (yes/no for enterprise)
• QBR/EBR completed in last 90 days
• Champion turnover (contact changed recently = risk)

Weight recommendation: 10–15% of total score

Building the Score: A Practical Framework

Step 1: Define Your Score Inputs

Start with data you actually have, not data you wish you had. For most early-stage SaaS teams, this means: login frequency, activation status, support ticket count, NPS, and payment status.

Don't build for six months. Build with today's data and iterate.

Step 2: Assign Weights

Sample weighting model for a mid-market B2B SaaS:

Adjust weights based on what actually correlates with churn in your data. If NPS has strong predictive power in your customer base, weight it higher.

Step 3: Score Each Component (0–10)

Score each input on a normalized 0–10 scale. Example for login frequency:

Step 4: Calculate Composite Score

Health Score = (Engagement Score × 0.40) + (Activation Score × 0.20) + (Support Score × 0.15) + (Commercial Score × 0.15) + (Relationship Score × 0.10)

Multiply by 10 to get to 100.

Step 5: Map to Red/Yellow/Green

Health Score Benchmarks

What constitutes a "good" score distribution depends on your product and customer mix. As a reference:

For a new cohort (first 60 days), expect more yellow/red. The score's value is in the trend per account, not the absolute number.

Triggering Actions from Health Scores

A health score that doesn't trigger an action is a dashboard metric, not a retention system. Build automated triggers:

Yellow accounts (score 50–74):

• Trigger in-app message: "You haven't used [core feature] in 2 weeks — here's a quick refresher"
• Send CSM an alert to schedule a check-in email
• Offer a 15-min onboarding call if activation is incomplete

Red accounts (score 25–49):

• Assign to CSM for personal outreach within 48 hours
• Trigger NPS survey if not collected in 60 days
• Offer a product review call or health review

Critical accounts (score 0–24):

• Immediate CSM escalation
• Consider save offer proactively before they reach the cancel button
• Prepare for exit survey if intervention fails

Green accounts (score 75–100):

• Flag as expansion candidates for expansion revenue scoring
• Request NPS response or G2/Capterra review
• Invite to beta features or customer advisory board

Health Score and Your Growth Ceiling

Customer health scoring directly impacts your Growth Ceiling. Every point of churn you prevent through early intervention is worth significantly more than acquiring a new customer to replace it.

At 5% monthly churn, your Growth Ceiling at $10K new MRR is $200,000. Reduce churn to 3% through systematic health scoring interventions, and the ceiling rises to $333,000 — a 67% increase without touching acquisition.

The math always favors fixing retention before scaling acquisition.

Common Health Scoring Mistakes

1. Too many inputs. Start with 5–7 signals. Adding 20 inputs without validating predictive power adds noise, not signal.

2. Static weights. Recalibrate weights quarterly by running a correlation analysis between health score and actual churn 90 days later.

3. No action loop. A score without automated triggers is a vanity metric. Every health segment needs at least one automatic action.

4. Ignoring leading edge of cohorts. Day 1–14 health signals are the most predictive of 90-day churn. Weight early activation signals more heavily for new accounts.

5. One-size-fits-all. Enterprise accounts and self-serve SMB accounts should have separate health models. Behavior patterns differ significantly.

Conclusion

A customer health score turns churn from a reactive fire drill into a proactive system. By combining product engagement, activation status, support sentiment, commercial health, and relationship signals into a single score, you create a predictive layer your SaaS metrics dashboard can act on automatically.

Start simple. Build with the data you have. Validate against actual churn. Iterate the weights.

The teams that win at retention aren't the ones with the most sophisticated models — they're the ones who act on signals fast enough to matter.