CSAT vs NPS vs CES: Which Customer Satisfaction Metric Should You Track?

CSAT, NPS and CES measure three different things, and the mistake most teams make is picking one and treating it as "the" satisfaction number. CSAT measures satisfaction with a single interaction, usually on a 1 to 5 scale. NPS measures overall loyalty and willingness to recommend, on a 0 to 10 question reported from -100 to +100. CES measures how much effort a customer had to spend to get something done. They answer different questions, they move for different reasons, and the score you should track depends entirely on what you are trying to fix.

Here is how each one actually works, when to reach for it, and the thing that matters far more than which one you choose.

The three metrics at a glance

CSAT: the touchpoint metric

Customer Satisfaction Score is the most direct of the three. You ask someone how satisfied they were with a specific thing, right after they experienced it, and you count the share who answered positively. Most teams take the top two options on a 1 to 5 scale and report that as a percentage, so 80 percent CSAT means four in five respondents were satisfied or very satisfied.

Its strength is precision. Because it is tied to a moment, a CSAT score points at something you can actually go and fix. Support CSAT dropping after a tooling change is a clear, actionable signal in a way that a blended company-wide number never is.

Its weakness is that it is easily gamed by when you ask. Survey people immediately after a successful action and your CSAT will look wonderful, because you have selected for the happy path. The score is only as honest as the moments you choose to measure. A blended 80 percent routinely hides a support experience at 95 percent and an onboarding flow at 55 percent, which is why segmenting by touchpoint matters more than the headline figure.

NPS: the relationship metric

Net Promoter Score asks one question: how likely are you to recommend us to a friend or colleague, from 0 to 10. Scores of 9 and 10 are promoters, 7 and 8 are passives, and 0 through 6 are detractors. The score is the percentage of promoters minus the percentage of detractors, so it runs from -100 to +100. Passives count toward the total but not toward either side, quietly dragging the number down.

NPS is the most criticized of the three, and some of that criticism is deserved. It is coarse, it swings hard on small samples because it is a difference of percentages rather than an average, and the moment an executive turns it into a target, Goodhart's law takes over and it stops measuring anything useful.

It survives anyway, for one reason worth being honest about: it is a cheap, universally understood prompt that gets customers to explain themselves. The score is the least interesting part of NPS. The free-text follow-up is the point. If you are running it, run it continuously on a sampled basis rather than as one big quarterly blast, and read what detractors wrote. Our NPS software page covers how to do that at scale and what a good score actually looks like in context.

CES: the friction metric

Customer Effort Score asks how easy it was to get something done, usually as an agreement statement ("the company made it easy for me to handle my issue") on a 1 to 7 scale. It emerged from research suggesting that reducing effort predicts loyalty better than delighting customers does, and in support contexts it has held up well. The mechanics are simple enough to run by hand, and we walk through how to calculate customer effort score including the scale, the timing and what counts as a good result.

CES is the one most product teams underuse. If your problem is that users are dropping out of a flow, or that support tickets keep arriving about the same task, CES points straight at the friction. It is also the least ambiguous to act on: a high-effort step is a step to redesign. Nobody has to debate what a bad CES means.

The catch is scope. CES only tells you about the effort of a specific task. It says nothing about whether the customer values your product, or whether they would recommend it. It is a scalpel, not a thermometer.

So which one should you track?

Track the one that matches the decision you are trying to make.

• You want to know if a specific experience is working: CSAT, asked right after that experience.
• You want an early warning on friction: CES, on the flows where users drop.
• You want a long-run health signal and a stream of honest qualitative feedback: NPS, sampled continuously.

Most mature teams end up running all three, and that is fine, because they cost almost nothing to ask. What actually breaks is not the choice of metric. It is that nobody reads the answers.

The thing that matters more than the metric

Every one of these three surveys has an open-text follow-up, and in most companies that text is the single most valuable and least used dataset they own. A CSAT of 72 percent tells you there is a problem. The 400 sentences customers wrote underneath tell you what the problem is. Teams obsess over which score to put on the dashboard and then leave the explanation sitting unread in a spreadsheet.

The score tells you that something moved. Only the comments tell you why, and only the why is actionable.

This is a solvable problem now in a way it was not five years ago. Language models read open-ended responses, work out what each one is about, and cluster them into themes without anyone building a tagging taxonomy in advance. What used to be a quarterly manual coding exercise can run continuously. The one thing to insist on is traceability: a theme labeled "pricing confusion, 214 responses" is only worth acting on if you can click it and read those 214 responses.

The bigger win is joining the survey answer to what that customer actually did. A detractor who wrote "too complicated" and never completed onboarding is telling you something very specific, and very different from a detractor who wrote "too complicated" after two years of heavy use. Behavior gives the comment its meaning. That join, between the score, the words and the usage, is what turns a satisfaction program from a reporting exercise into a roadmap. It is exactly what customer satisfaction survey software should be doing for you, and what most of the category still leaves as manual work.

Do not forget the feedback you never asked for

Surveys only capture the customers who agreed to answer a question you decided to ask. A large share of the most candid feedback about your product is written somewhere you did not send a survey at all: in support tickets, in app store reviews, in G2 comments, and in public posts. Those channels are unprompted, which makes them harsher and often more useful, and they cover the people who would never fill in a form.

Any serious satisfaction program eventually needs to pull those in alongside the survey data, and to keep an eye on what customers are saying about you in public as well as what they tell you directly. Analyzed together, the prompted and the unprompted feedback usually agree, and where they disagree is itself informative: it usually means your survey is asking the wrong question, or asking it at the wrong moment.

A practical setup

If you are starting from nothing, this sequence works well:

• Put CES on your two worst flows. One question, in-app, right after the task. Fix what it surfaces.
• Put CSAT on your support interactions. It is the industry norm, it is easy to benchmark internally, and it catches regressions fast.
• Run NPS continuously and sampled, not quarterly and blasted. Aim to have a rolling score and a steady flow of comments rather than a once-a-quarter pile nobody reads.
• Analyze all the open text together, along with your support tickets and reviews, and cluster it into themes with counts attached.
• Segment everything. A blended score across plans, cohorts and touchpoints averages away the exact signal you need.

Do that and the question of CSAT versus NPS versus CES stops being interesting, which is the correct outcome. The metrics are just prompts. The answers are the asset.