Customer Success
Customer success without a CS team: a playbook for small SaaS
You don't need a customer success hire to keep customers. A 30-minute weekly routine, three signals to watch, outreach templates, and when to actually hire.
Best Practices
How to build a customer health score that actually predicts churn
Step-by-step guide to building a customer health score. Components, weights, thresholds, and common mistakes. With examples for SaaS teams at every stage.
A customer health score is a single number (typically 0-100) that tells you how well a customer is doing. A high score means they’re engaged, getting value, and likely to renew. A low score means they’re disengaging and at risk.
The concept is simple. Getting it right is not.
Most SaaS companies either don’t have a health score at all, or they have one that nobody trusts because it doesn’t match reality. This guide covers how to build one that works — from component selection to weight tuning.
The three properties of a useful health score
1. It predicts outcomes
A good health score correlates with retention. If your highest-scoring customers churn at the same rate as your lowest-scoring customers, your score is noise.
2. It’s actionable
“Health score: 42” is useless unless you know why it’s 42. A good health score breaks down into components you can act on: “Activity: 8/30, Feature adoption: 20/30, Engagement: 14/40.”
3. It updates automatically
A health score that requires manual input is a health score that’s always outdated. Automate the inputs so the score reflects reality in real-time.
What to include in your health score
Component 1: Product activity frequency
What it measures: How often the customer uses your core product functionality.
How to measure it: Count key actions per time period. For example:
- Orders created per week
- Documents uploaded per week
- API calls per day
Why it matters: Active customers renew. Inactive customers churn. Activity frequency is the strongest predictor of retention for most SaaS products.
Example thresholds:
| Activity level | Score | What it means |
|---|---|---|
| 5+ actions/week | Full points | Power user, highly engaged |
| 2-4 actions/week | 60% of points | Regular user, healthy |
| 1 action/week | 30% of points | Low usage, monitor closely |
| 0 actions/week | 0 points | Inactive, at risk |
Component 2: Feature adoption
What it measures: How many of your key features the customer has adopted.
How to measure it: Binary check — has the customer used feature X? For features that can be “disabled,” track enabled/disabled state.
Why it matters: Customers who use more features are stickier. They’ve invested more time learning your product and get more value from it. Single-feature customers are easy to replace.
Example scoring:
| Feature | Points | Rationale |
|---|---|---|
| Core integration connected | 10 | Required for value |
| First entity created | 10 | Proof of activation |
| Bulk import used | 5 | Advanced usage |
| API key generated | 5 | Developer engagement |
Component 3: Engagement recency
What it measures: When was the last time the customer did something?
How to measure it: Days since last meaningful activity.
Why it matters: A customer who was active 2 days ago is different from a customer who was active 30 days ago — even if their total activity count is the same.
Example thresholds:
| Days since last activity | Score | What it means |
|---|---|---|
| 0-3 days | Full points | Recently active |
| 4-7 days | 70% of points | Getting quiet |
| 8-14 days | 30% of points | Warning sign |
| 15+ days | 0 points | Gone quiet |
Component 4: Onboarding completion
What it measures: How far through your setup/onboarding process the customer has progressed.
Why it matters: Customers who don’t complete onboarding are 3-5x more likely to churn in the first 90 days. If they never get to value, they never renew.
Component 5: Metadata and subscription signals
What it measures: Non-product signals that indicate health.
Examples:
- Subscription status (active, trial, past_due)
- Last login date
- Support ticket frequency (high volume can indicate frustration)
- NPS score (if available)
How to assign weights
There’s no universal formula. Weights depend on your product and what predicts retention in your business. Start here and iterate:
| Component | Suggested starting weight | Rationale |
|---|---|---|
| Product activity frequency | 30-40% | Strongest predictor for most SaaS |
| Feature adoption | 20-30% | Stickiness indicator |
| Engagement recency | 15-20% | Timeliness signal |
| Onboarding completion | 10-15% | Critical for first 90 days |
| Metadata signals | 5-10% | Supporting context |
| Total | 100% |
How to iterate
- Start with the weights above
- Run for 30 days
- Check: do the lowest-scoring accounts match the ones you know are at risk?
- Adjust weights. If activity matters more than you thought, increase it. If feature adoption doesn’t predict churn, reduce it.
- Repeat quarterly.
Five health score mistakes to avoid
1. Too many components
10+ components dilute the signal. Stick to 3-5 that actually predict outcomes. Every component you add makes the score harder to interpret.
2. Equal weights
Not all signals are equal. If you weight everything at 20%, you’re saying login frequency matters as much as product usage. It doesn’t.
3. Lagging indicators only
NPS scores arrive quarterly. Support tickets happen after the problem. Use leading indicators (activity frequency, feature adoption) that show problems before the customer notices.
4. Binary thresholds
“Active = 100 points, Inactive = 0 points” creates a score with no gradient. Use tiered thresholds so the score responds to changes in behavior, not just on/off states.
5. Never recalibrating
Your product changes. Your customers change. A health score model that was accurate 6 months ago might not be accurate today. Review and adjust quarterly.
What the numbers mean
| Range | Label | What it means | Action |
|---|---|---|---|
| 75-100 | Healthy | Engaged, getting value, low churn risk | Expansion opportunities, advocacy |
| 50-74 | Moderate | Using product but not fully engaged | Monitor, encourage deeper adoption |
| 25-49 | At Risk | Declining engagement, warning signs | Proactive outreach, intervention |
| 0-24 | Critical | Barely using product, likely to churn | Urgent intervention or acceptance |
Your options
| Approach | Setup time | Cost | Pros | Cons |
|---|---|---|---|---|
| Spreadsheet | 1 day | Free | Simple | Manual, always outdated, doesn’t scale |
| SQL queries | 1-2 weeks | Free (if you have data) | Flexible | Manual runs, no dashboard, maintenance burden |
| LogoPulse | 10 minutes | $0-349/mo | Auto-calculated, configurable, dashboard included | Requires SDK integration |
| Enterprise CS platform | 2-6 weeks | $500-2,500+/mo | Full CS workflow | Overkill for < 500 accounts |
| Custom build | 3-6 months | $30,000-90,000 | Total control | Engineering cost, maintenance, opportunity cost |
For a deeper comparison of these options, see LogoPulse vs the alternatives. If you’re weighing a health score against survey-based metrics, read customer health score vs NPS.
Build your health score in 10 minutes
LogoPulse automates everything in this guide: component scoring, weight configuration, automatic calculation, and trend visualization. See how health scores work in LogoPulse or start free with your first 20 accounts.