Reading your CSAT score
CSAT distils customer satisfaction into one percentage — the share of customers who walked away happy. Here's the formula, worked examples on every common scale, how to do it in a spreadsheet, what a good score looks like, and how you compare to your industry.
How to calculate CSAT
After an interaction you ask “How satisfied were you?” on a 1–5 scale. CSAT is the share who picked the top two ratings — 4 or 5. Most teams report this top-2-box method; the calculator can also take the average rating as a percentage of the maximum.
CSAT % = (satisfied (4–5) ÷ total) × 100
Worked example: you send 400 surveys and 120 people answer. Of those, 78 rated you 4 or 5, 30 rated 3, and 12 rated 1 or 2. Your CSAT is 78 ÷ 120 × 100 = 65%. Note what is not in that sum: the 280 people who ignored the survey never enter the calculation. CSAT measures the satisfaction of people who replied, which is why response rate belongs next to the score whenever you report it.
The one judgement call is where “satisfied” starts. Top-2-box on a 1–5 scale is the near-universal convention, and counting 3s as satisfied inflates the score enough to make it incomparable with anyone else's. Pick a definition, write it down, and never change it mid-year — a CSAT number is only useful against your own history.
CSAT calculation examples by scale
The formula never changes; only the definition of a satisfied response does. These are the three scales teams actually use.
| Scale | Counts as satisfied | Example | CSAT |
|---|---|---|---|
| 1–5 (standard) | Ratings of 4 and 5 (top 2 box) | 78 of 120 responses rated 4 or 5 | 65% |
| 1–3 (simple) | Rating of 3 only (top box) | 44 of 80 responses rated 3 | 55% |
| 1–10 | Ratings of 8, 9 and 10 (top 3 box) | 150 of 200 responses rated 8+ | 75% |
A 1–10 scale is worth avoiding unless you have inherited one. Ten options give respondents more precision than they actually have, and the top-3-box cutoff is arbitrary enough that two teams using the same scale will report different numbers for identical feedback.
How to calculate CSAT in Excel
With one response per row and ratings in column B, the whole calculation is one formula:
=COUNTIFS(B:B,">=4")/COUNT(B:B)*100
COUNT ignores blank rows, so unanswered surveys stay out of the denominator on their own.For CSAT by month or by agent, put that formula behind a pivot table rather than maintaining one sheet per segment. And if you are recalculating a spreadsheet by hand every week, that is the point at which the metric should live in your helpdesk instead — a survey that fires automatically after each resolved conversation collects the same data without anyone exporting anything.
What is a good CSAT score?
CSAT runs from 0 to 100%. Here's the whole range at a glance, and your live result lands on this same scale the moment you calculate.
Excellent — Top-tier. You're ahead of most companies in any industry.
Good — A healthy score most support teams aim for.
Fair — Acceptable, but there's clear room to improve.
Needs work — Below the typical range — worth digging into why.
How you compare by industry
CSAT clusters tighter than NPS — most fields sit in the mid-70s to low-80s, with telecom and travel running lower. Compare against your own field, and against your own past scores.
| Software / SaaS | 78% |
| Ecommerce / Retail | 80% |
| Financial services / Banking | 80% |
| Healthcare | 77% |
| Hospitality / Hotels | 76% |
| Travel / Airlines | 75% |
| Telecom / Utilities | 70% |
Calculating CSAT in a call center
The formula is identical; what changes is when you ask and what you attach the score to. Call centers survey immediately after the call ends — either an IVR prompt before the customer hangs up, or an SMS within a few minutes — because satisfaction with a phone interaction decays fast. Waiting until the next day measures memory rather than the call.
The difference that matters is segmentation. A single centre-wide CSAT is nearly useless for managing a team; the number becomes actionable when you calculate it per agent, per queue, and per call reason, because that is where a low score points at something you can change — a training gap, a mis-routed queue, or a broken process upstream of the call. Watch the sample size when you slice it, though: an agent with nine responses in a month has a CSAT that will swing wildly on one bad call, and treating that as performance data is how you lose people's trust in the metric.
CSAT survey questions
The score is only as good as the question that produced it. A CSAT survey should be one question, asked once, immediately after the thing it is about. Every extra question costs you responses, and a CSAT built on a 4% response rate tells you about the 4% who had strong feelings.
The standard wording is “How satisfied were you with the support you received?” on a 1–5 scale from very dissatisfied to very satisfied. Keep the subject specific — satisfaction with this conversation, not with the company — because a vague question collects opinions about pricing and product that your support team cannot act on. Variants worth using: for onboarding, “How satisfied were you with getting set up?”; after a bug fix, “How satisfied were you with how we handled this?”
Add one optional free-text follow-up — “What could we have done better?” — and make it genuinely optional. That comment box is where the actual value is: the score tells you something changed, the comments tell you what to fix. Never ask for a rating and a comment as two required fields; you will halve your sample to gain a paragraph.
How to improve your CSAT score
Almost every low CSAT traces back to one of three things, and speed is the biggest. First response time correlates with satisfaction more strongly than almost any other variable — people forgive an imperfect answer far more readily than a slow one. If your score is soft and your median first response is measured in hours rather than minutes, fix that before touching anything else.
The second is effort: every repeated explanation, handoff, and “can you confirm your account details again” costs satisfaction. The third is expectation-setting — a customer told honestly that something will take three days is more satisfied at day three than one promised a fix “shortly” and updated on day two.
Practically: read the free-text comments attached to every 1 and 2 rating for a month and tally the causes rather than reacting to individual ones. The tally almost always points at a handful of recurring issues, which is a far better roadmap than a score moving a point either way. Answering the repetitive questions instantly — with a help center customers can search, or an AI agent that replies in seconds — moves both speed and effort at once.
CSAT vs NPS vs CES
The three are calculated differently and answer different questions, which is why most teams end up running more than one. CSAT asks how satisfied someone was with a specific interaction and is scored as a percentage of satisfied responses — it is the right metric immediately after a support conversation. NPS asks how likely someone is to recommend you overall and is scored by subtracting detractors from promoters, giving a number from −100 to +100 — a relationship measure, not an interaction one. CES asks how much effort the customer had to expend, and predicts loyalty better than either for transactional support.
Because the maths differs, the numbers are not comparable: a CSAT of 75% and an NPS of 75 have nothing to do with each other. If you only run one, run CSAT on resolved conversations — it is the easiest to collect and the easiest for a support team to act on. Our NPS calculator and CES calculator handle the other two.
