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.
Retention
Customer health score vs NPS: which one predicts churn?
NPS tells you how customers feel. A health score tells you what they do. Here's why behavior beats sentiment for churn prediction, and how to use both.
If you run a B2B SaaS company, someone has told you to track Net Promoter Score. Someone else has told you to build a customer health score. Both sound like “how happy are my customers?” They are not the same question, and only one of them reliably tells you who’s about to leave.
This article breaks down what each metric measures, where NPS falls short for churn prediction, and how to use both without turning your retention strategy into survey management.
What each metric actually measures
Net Promoter Score (NPS) asks one question: “How likely are you to recommend us to a friend or colleague?” Respondents pick 0-10. Promoters (9-10) minus detractors (0-6) gives you a number between -100 and +100.
A customer health score is a composite number (usually 0-100) calculated from what a customer does in your product: how often they use it, which features they’ve adopted, how recently they were active, and whether they finished onboarding. Nobody fills out a form. The score updates from behavior.
| NPS | Customer health score | |
|---|---|---|
| Source | Self-reported survey | Product activity data |
| Unit | Individual respondent | Customer account |
| Frequency | Quarterly or on trigger | Hourly or daily |
| Coverage | Whoever responds (15-30%) | Every account, always |
| Direction | Lagging (feelings after the fact) | Leading (behavior before the outcome) |
| Best at | Sentiment, referrals, feedback | Churn prediction, expansion targeting |
Neither metric is “wrong.” They answer different questions. The problem starts when you use NPS to answer the question a health score was built for.
Why NPS is a weak churn predictor
It arrives too late
A customer doesn’t wake up unhappy. The sequence is usually: they stop using a feature, then they stop logging in regularly, then their team lead stops caring, then someone answers your NPS survey with a 5, and then they cancel. NPS catches step four. A health score catches step one.
For a customer who churns 8 weeks after their activity starts declining, an NPS survey sent quarterly has roughly a 1-in-3 chance of landing during the decline at all, and even then it only tells you what you could already have seen in the activity data.
It only counts the people who respond
A typical B2B NPS response rate is 15-30%. The customers most likely to skip the survey are the disengaged ones, which is exactly the group you need to hear from. Your NPS ends up over-weighted toward engaged users who like you, and your at-risk accounts stay silent until the cancellation email.
A health score has 100% coverage by definition. Every account that sends events gets a score. The quiet ones get the lowest scores, which is the point.
It measures users, not accounts
In B2B, the account renews, not the user. One enthusiastic power user can give you a 10 while the three managers who approve the budget haven’t logged in for a month. NPS aggregates individual opinions. A health score aggregates account-level behavior: total activity across every seat, features adopted by the company, days since anyone did something meaningful.
The sample-size problem for small SaaS
If you have 80 customers and a 25% response rate, your NPS is built on 20 answers. One detractor moves the score by 5 points. Two promoters moving to passive drops it by 10. You’ll spend meetings explaining noise.
Health scores don’t have this problem because they aren’t averages across a population. Each account is scored on its own data. Whether you have 20 customers or 2,000, the score for account #14 is based on what account #14 did.
What a health score sees that NPS can’t
Here’s a simplified example of two accounts, both of which returned an NPS of 8 last quarter:
| Signal | Account A | Account B |
|---|---|---|
| Orders created, last 7 days | 34 | 2 |
| Orders created, prior 7 days | 31 | 19 |
| Features adopted (of 6 tracked) | 5 | 3 |
| Days since last activity | 0 | 6 |
| Onboarding stage | Power User | Active |
| Health score | 91 | 48 |
| Trend, 30 days | Stable | Declining 22 pts |
Both are “passives” on NPS. One is your healthiest customer. The other dropped 90% in weekly volume and hasn’t touched the product in almost a week. The health score flags Account B today. NPS would have told you nothing until the next survey, and probably nothing then either.
This is what the reduce churn use case is built around: you get the signal while there’s still time to act on it.
Where NPS still earns its place
None of this means you should stop measuring NPS. It does three things a health score can’t:
- It captures sentiment the product data hides. A customer can use your product daily and still resent it because a competitor is cheaper. Behavior says healthy; NPS says “we’re shopping around.”
- It generates referrals. Promoters are your best source of warm leads. You need to know who they are, and behavior alone won’t tell you who’s willing to recommend you.
- It surfaces product feedback. The follow-up question (“What’s the main reason for your score?”) is often the most valuable thing in the survey.
Use NPS for those jobs. Don’t use it as your early-warning system.
How to use both together
The practical setup for a team under 500 customers:
Step 1: Make the health score your primary retention signal. Build it from 3-5 behavioral components: activity frequency, feature adoption, engagement recency, onboarding completion. Weight them by what predicts renewal in your business. The customer health score guide walks through component selection and weighting.
Step 2: Add NPS as a small metadata component. Give it 5-10% of the score. A recent detractor response nudges the score down; a promoter nudges it up. It refines the score without dominating it.
Step 3: Read the two together. The combination is more useful than either alone:
| Health score | NPS | What it means | What to do |
|---|---|---|---|
| High | High | Healthy, happy | Ask for referrals, explore expansion |
| High | Low | Using it, but frustrated | Product feedback call; fix the friction |
| Low | High | Likes you, isn’t using you | Onboarding or enablement problem; re-engage |
| Low | Low | Disengaged and unhappy | Urgent intervention or accept the loss |
Step 4: Trigger surveys from the health score. Instead of blasting NPS quarterly, send it when the score changes: a drop of 15+ points, or crossing into “healthy” after onboarding. Response rates go up because the timing is relevant, and the answers explain what the behavior data showed.
The short version
NPS tells you how customers feel at the moment they bother to answer. A health score tells you what every customer is doing, every day, whether they answer or not. For predicting churn, behavior wins. For understanding why, sentiment helps. Run both, but let the health score set the alarms.
If you want health scores without building the pipeline yourself, LogoPulse calculates them automatically from the events your product already generates. It takes about 10 minutes to connect and is free for your first 20 accounts.