Monday's leadership meeting. This quarter's NPS comes up at +42, six points higher than last time, and everyone congratulates each other. Three weeks later, a customer who gave you 9/10 in the spring tells you they won't be renewing. The score didn't lie to you: it was simply answering a different question from the one you were asking. Here's how to choose the right metric at the right moment, so you never get caught out again!
NPS, CSAT, CES. Three acronyms that turn up on every Customer Success dashboard, often lumped together in the same "satisfaction" column and sometimes used interchangeably. That's a mistake! Each one measures something different: the overall relationship, a specific interaction, the effort the customer had to make. Mixing them up is like steering with a thermometer when what you need is a barometer.
And no, the real question isn't "which one is best". The real question is: what do you want to ask your customer, and at what point in their journey?
Three metrics, three different questions
NPS: the relationship question
The Net Promoter Score was popularised in the early 2000s by Fred Reichheld, a consultant at Bain & Company. The principle is well known: you ask the customer, on a scale of 0 to 10, whether they would recommend your solution to a peer.
Respondents are then sorted into three groups:
- Promoters: scores of 9 and 10;
- Passives: scores of 7 and 8;
- Detractors: scores of 0 to 6.
NPS equals the percentage of promoters minus the percentage of detractors. A fictional example: out of 100 respondents, with 50 promoters, 30 passives and 20 detractors, you get an NPS of +30. The score therefore ranges from -100 to +100.
What it measures is an intention to recommend. It's often used as a proxy for loyalty, and that's precisely where things get complicated (more on that below). NPS takes the temperature of the relationship as a whole, not of a specific moment. That's both its strength (a trend you can compare over time) and its limitation: it tells you neither what to fix nor who's going to leave.
CSAT: the in-the-moment question
The Customer Satisfaction Score is the simplest of the three. You ask the customer whether they were satisfied with a given interaction: a conversation with support, a training session, a Business Review. They usually answer on a scale of 1 to 5, from "very dissatisfied" to "very satisfied".
CSAT is the percentage of 4 and 5 responses out of all responses. If 80 customers out of 100 answer 4 or 5, your CSAT is 80%.
CSAT is an "in the moment" metric: it judges an episode, not the relationship. A customer can be delighted with support's answer and still be seriously considering leaving, because the underlying problem lies elsewhere. That's where you need to be careful about how it's interpreted.
Incidentally, a ticket isn't bad news in itself. According to ChurnRX data, customers who have opened at least one ticket have a median lifetime of 89 months, compared with 15 months for those who never have: a customer who reaches out to support is a customer who's invested. Post-ticket CSAT measures the quality of your response, not the health of the account.
CES: the effort question
The Customer Effort Score is the least well known of the three, and probably the most underrated in B2B SaaS. It was introduced in 2008 by Matt Dixon and his team at the Corporate Executive Board (CEB, since acquired by Gartner), then popularised by the article "Stop Trying to Delight Your Customers", published in Harvard Business Review in 2010.
Their argument went against the grain: loyalty depends less on moments of delight than on how easy you make things. According to CEB research, 96% of customers who had a high-effort interaction became more disloyal, compared with just 9% after a low-effort one.
A word of caution, though: these figures relate to stated loyalty intentions after customer service interactions, mostly in B2C. They show that effort damages the relationship. They don't say that CES predicts your renewals.
In its most common form, CES asks the customer whether they agree with a statement such as "The company made it easy for me to handle my issue", on a scale of 1 to 7. You then take an average, or the percentage of responses between 5 and 7.
That might seem trivial. The thing is, in B2B SaaS, effort is everywhere. Configuring the tool, connecting an integration, finding a feature, getting a clear answer from support. Every point of friction is a small debt the customer will make you pay at renewal.
What really sets them apart
If you need to sum up all three in one line each:
- NPS: "What does this customer think of us overall?"
- CSAT: "Was this customer satisfied with this interaction?"
- CES: "Did this customer have to fight to get what they wanted?"
Beyond the question asked, three differences shape how they're used.
- Timing. NPS is measured with hindsight, periodically. CSAT and CES are measured in the moment, straight after an event. Sending an NPS survey the day after a major incident measures the incident, not the relationship.
- Granularity. NPS lumps everything together: product, price, support, the relationship with the CSM, the customer's own circumstances. CSAT and CES isolate a single touchpoint. That's exactly what Gartner criticised NPS for in a customer service context: in May 2021, the firm predicted that more than 75% of organisations would have abandoned NPS as a measure of service performance by 2025, for lack of actionable insight. The prediction should be taken with a pinch of salt given how widespread NPS still is, but the underlying diagnosis holds.
- Actionability. A low CES on onboarding points to a step you need to simplify. A low CSAT on support points to a ticket to review. A falling NPS tells you there's a problem without telling you which one.
Put simply: CSAT and CES explain, NPS observes. They aren't competitors; they serve different purposes.
Pitfalls specific to B2B SaaS
Most content on this topic is written with B2C logic in mind: one customer, one person, one score. In B2B SaaS, reality is messier.
The respondent isn't the account
An account is made up of daily users, an administrator, a sponsor and sometimes a buyer. The power user who gives you a 10 isn't the one who signs the renewal. And the finance director who signs it will probably never answer your survey.
An NPS calculated without distinguishing between roles mixes opinions that carry very different weight in the decision. Segment your results by persona (user, admin, decision-maker) and look at coverage: how many accounts have at least one response from a decision-maker?
A good score doesn't predict a renewal
Response rates to NPS surveys in B2B are often low, and that silence is information in itself. A disengaged customer doesn't take the time to rate you. They just leave.
Be wary, too, of the reflex of only following up with detractors. An analysis published by Thematic, which cross-referenced NPS scores with churn data, shows that departures are spread across every score, not just among the most dissatisfied.
ChurnRX goes further. In its study of churn factors, the firm compared the lifetime of SaaS customers according to the NPS score they had given. The result: no correlation at all between NPS and customer lifetime. Not a weak correlation: none.
That's why we take the view that NPS can't predict churn. It's still useful for tracking how your customers' mood changes over time, or for measuring the impact of a change in pricing, product or level of support. Not for knowing who's going to leave. An 8, or even a 9, absolutely does not mean your customer is safe!
The score becomes the goal
As soon as a metric becomes part of the team's objectives, it starts measuring something else. A CSM whose bonus depends on NPS will end up, consciously or not, choosing the timing, the recipient or the wording that flatters the score. Some will go as far as slipping "a 9 or 10 would really help us" into their email.
Don't do it! You'll get a nice number and lose the only thing that mattered: an honest, genuine signal.
The scale doesn't mean the same thing everywhere
If your portfolio is international, your scores aren't directly comparable from one country to another. Cultural attitudes to scoring vary considerably, and a 7 can be a compliment in one market and a warning sign in another. Compare each segment with itself over time rather than with a global average.
How to combine them across the customer lifecycle
The right approach is to pair each metric with the point in the journey where it tells you something the others can't. Here's a starting framework to adapt to your own lifecycle:
- End of onboarding or go-live: CES ("Setting up the solution was easy"), to spot the steps where things get stuck.
- After first use of a key feature: a one-question in-app CES.
- After a support ticket is closed: CSAT, or even CES on the resolution.
- After a training session or a Business Review: CSAT, to check that the time spent was useful to the customer.
- Twice a year, on the relationship: NPS, outside periods of incident, to track a trend and open a conversation with every respondent.
Two rules complete this framework. First, always pair the score with an open question ("What's the main reason for your score?"). That's often where the useful information lies. Second, keep an eye on survey fatigue: a contact who receives three surveys in a month will stop answering all three.
A score is useless if nobody acts on it
Measuring satisfaction without closing the loop achieves nothing beyond a feel-good moment in the leadership meeting. Every response should trigger something.
Responses that trigger action
In practice, it might look like this:
- Detractor: a call from the CSM within a few days to understand the issue and propose a plan.
- Promoter: a request for a review, a testimonial or a referral.
- Low CSAT: the support manager reviews the ticket.
- Low CES: escalated to the Product team, with the verbatim.
And tell the customer what you did with their feedback. A customer who sees that their opinion changed something will answer the next survey. One who feels they're talking into the void won't.
A conversation starter, not a health indicator
NPS, CSAT and CES have one thing in common: they're declarative. They depend on the customer's willingness to respond, and on what they're prepared to tell you.
Take a fictional example. An account gave you a 9 in the last NPS survey. But since then, usage has dropped by 40%, your sponsor has left the company and conversations have become less frequent. Which of these signals do you believe?
That's why we advise against including NPS as a component of your health score. Without solid data on their link to retention in B2B SaaS, we apply the same caution to CSAT and CES. Your health score should be built on signals with a proven link to retention: usage, measured outcomes, stakeholder engagement and the quality of interactions.
Surveys have their place alongside it. They trigger an action (calling back a detractor, understanding a low CES, fixing a step where users get stuck) and they shed light on the context when the health score moves. But they don't drive it. To measure the quality of the relationship without relying on a response rate, sentiment analysis of customer conversations is a far more solid option.
So which one should you choose?
Don't choose, but don't launch them all at once either: a survey whose responses nobody deals with does more harm than good.
If you're starting from scratch, here's an order that works:
- Start with support CSAT, which is often built into your ticketing tool. It's the easiest to roll out and use.
- Add CES to onboarding, where time to value and a large share of early churn are decided.
- Launch relationship NPS once you have the capacity to handle every response and get back to every detractor.
And if you could only keep one in B2B SaaS? Our view: CES, asked at key moments in the journey, is the most actionable of the three. It points to specific friction you can remove. NPS remains useful for tracking a trend and talking to your leadership team, as long as you never mistake it for a renewal forecast.
Conclusion
NPS, CSAT and CES aren't in competition. They answer three different questions: what does this customer think of us, how did this interaction go, did they have to struggle to get there. A good set-up places each one at the right point in the journey, pairs it with an open question and, above all, links it to action. The score is never the goal. What matters is what it prompts you to do.
Further reading
