Asking "are you satisfied?" sounds simple, but customer satisfaction hides several different questions. One company wants to know whether the customer would buy again; another wants to understand whether the last support contact went well; yet another wants to find out where the product creates unnecessary friction. Each of these questions calls for a different metric — and picking the wrong one leads to the wrong decisions.
In the world of customer experience, three acronyms dominate the conversation: CSAT, NPS and CES. Many teams adopt one of them because it is fashionable, measure it for years, and never understand why the number goes up or down. Others use all three at once without knowing what each one adds.
This article explains, without jargon, what each metric actually measures, when it makes sense to ask, how to calculate it without fooling yourself, and how to combine the three into a coherent system. By the end, you should be able to choose the right metric for the decision in front of you.
What CSAT is and what it measures
CSAT (Customer Satisfaction Score) is the most direct metric: it asks the customer how satisfied they were with something specific. The typical question is "How satisfied are you with...?", answered on a 1-to-5 (or 1-to-7) scale. The result is usually expressed as the percentage of positive answers — normally the two highest levels of the scale.
CSAT's great strength is its closeness to the moment: you ask right after an interaction (a purchase, a support contact, a delivery) and it measures that specific episode. It is intuitive, easy to answer and comparable across touchpoints. Its weakness is that it measures immediate satisfaction, not loyalty: a customer can be happy with one interaction and still switch supplier the following month.
What NPS is and what it measures
NPS (Net Promoter Score) asks a different question: "How likely are you to recommend our company to a friend or colleague?", on a 0-to-10 scale. Answers fall into three groups: promoters (9-10), passives (7-8) and detractors (0-6). The result is the percentage of promoters minus the percentage of detractors, a number that ranges from -100 to +100.
NPS does not measure an episode, it measures the relationship. By relying on recommendation, it captures the customer's willingness to put their own reputation on the line, which correlates with loyalty and word of mouth. That makes it a good relational indicator, measured periodically. The most common criticism is that it is a blunt number: two customers with the same NPS can have opposite motivations, and the metric only gains value when paired with the open "why?" question.
What CES is and what it measures
CES (Customer Effort Score) measures effort: "How easy was it to solve your problem?" or "Did the company make it easy for you to handle your issue?". The core idea, supported by customer-service research, is that reducing effort predicts loyalty better than delighting the customer. People do not reward extraordinary experiences as much as they punish ones that are hard work.
CES shines in transactional and operational contexts: customer support, return processes, activating a service, checkout. A poor CES points directly to friction — a confusing form, transfers between agents, extra steps. It is actionable almost immediately, because it reveals where the process stalls.
Relational vs transactional: when to ask
The most useful distinction is not between metrics, it is between moments. There are relational questions, which assess the overall health of the relationship, and transactional questions, which assess a specific episode. NPS usually lives in the relational register: you ask periodically (for example, quarterly) across the whole base. CSAT and CES live in the transactional register: you ask right after a specific event.
Asking at the wrong moment distorts everything. An NPS sent right after an unresolved problem measures the frustration of that instant, not the relationship. A CSAT sent weeks after a purchase measures memory, not experience. Timing is part of the metric.
How to calculate each one without fooling yourself
The formulas are simple, but there are traps in interpretation:
- CSAT = (positive answers ÷ total answers) × 100. Decide up front what counts as "positive" (for example, 4 and 5 on a 5-point scale) and do not change that criterion halfway through.
- NPS = % of promoters − % of detractors. Never average the answers: NPS is not an average, it is a difference of proportions.
- CES = average (or percentage of "easy" answers) on an effort scale. Always keep the same scale and the same wording over time.
Beyond the formula, watch out for response bias: people who answer surveys are rarely a neutral sample. Low response rates and an over-representation of very happy or very angry customers can skew the number. A small but consistent sample is worth more than a large number measured differently each quarter.
Common mistakes when measuring satisfaction
The first mistake is confusing the three: treating NPS as if it were immediate satisfaction, or CSAT as if it were loyalty. The second is ignoring the free text — the open comment is often worth more than the score, because it says why. The third is optimising the metric instead of the experience: when a team is judged on NPS, it quickly learns to ask for the score only from happy customers. That is Goodhart's law in action: when a measure becomes a target, it stops being a good measure.
There is also the mistake of asking too much. Flooding the customer with surveys creates fatigue and drives down the response rate, degrading the very signal you wanted to measure.
How to combine the three into a measurement system
The three metrics do not compete — they complement each other, as long as each has a clear role. A balanced combination usually works like this:
- Relational NPS, periodic, to track the health of the relationship and the trend over time.
- Transactional CSAT, right after key interactions, to learn which moments go well or badly.
- CES in processes where friction is expensive (support, onboarding, returns), to pinpoint where effort kills the experience.
The value multiplies when these scores connect to operational data: wait time, number of contacts to resolve, channel used. That is where satisfaction stops being a vanity number and starts pointing to concrete causes.
Mini case study: a B2B software company
A B2B software company measured only relational NPS, once a quarter. The number was stable, around +20, but customer churn kept rising and no one could explain why. NPS said that there was dissatisfaction, but not where.
The team added two transactional measurements: a CSAT after every support contact and a CES at the end of onboarding. Within a few weeks the pattern became clear. Support CSAT was healthy (close to 85% positive answers), but onboarding CES was poor: new customers described a long, confusing setup process. Cross-referencing the data, most NPS detractors had gone through a difficult onboarding in their first 90 days.
With the problem located, the company simplified initial setup and began closely following customers with high CES. The next quarter, reported onboarding effort fell by half and relational NPS rose from +20 to +34. What unlocked the improvement was not measuring more, it was measuring the right thing at the right moment.
In practice
Choose the metric by the decision you want to make, not by fashion. If you want to track the relationship, use NPS; if you want to assess a moment, use CSAT; if you want to hunt friction, use CES. Ask little, at the right moment, and treat the open comment as the real treasure. A satisfaction metric is only worth it if it leads to an action — otherwise it is just one more number on the dashboard.
