CSAT Calculator
Know what your CSAT score really means.
Enter your survey responses and get the score, the statistical confidence behind it, and a plain-English read on what to do next.
460 of 547 responses rated you 4 or 5 out of 5.
33 more satisfied responses in a row — with no dissatisfied ones in between.
Free for 3 months. No credit card required.
How to calculate CSAT
The CSAT calculator above does this for you, but the arithmetic is worth knowing. Customer Satisfaction Score is the percentage of survey respondents who said they were satisfied. That is the whole idea — everything else is a decision about what counts as “satisfied.”
Count satisfied ratings, divide by every response, then convert to a percentage.
“Satisfied” almost always means the top two options on your scale — a 4 or a 5 on a 1–5 scale. That is the top-2-box method, and it is what most published benchmarks assume.
You send a post-resolution survey and get 547 replies: 312 rated you 5, 148 rated 4, 47 rated 3, 28 rated 2, and 12 rated 1.
- 1.Add the top two boxes: 312 + 148 = 460 satisfied responses.
- 2.Divide by the total: 460 ÷ 547 = 0.841.
- 3.Multiply by 100: CSAT = 84.1%.
With that sample size, the 95% confidence interval runs from 80.8% to 86.9% — so the honest reading is “84.1%, give or take three points.”
Neutral responses stay in the denominator. Dropping them is the most common way teams accidentally inflate their own score by five to ten points without anything actually improving.
Which CSAT formula should you use?
Three methods are in common use. They are not interchangeable, and quietly switching between them is how a support team ends up reporting an improvement that never happened.
| Method | What it counts | Use it when |
|---|---|---|
| Top-2-boxDefault | Ratings 4 and 5 on a 1–5 scale | You want a number comparable to published benchmarks and to other teams. |
| Top-box only | Only the highest rating | You are chasing delight rather than adequacy and want a harsher signal. |
| Average score | The mean of every rating | You report a rating out of 5 rather than a percentage. |
The trap worth naming. The same 547 responses above give a 84.1% top-2-box score and a 4.32 / 5 average — which converts to 86.3%. Those are different measurements of the same data, not two attempts at one number. Comparing a top-2-box score against an average-based benchmark is meaningless, and switching methods mid-quarter will show a jump that no customer ever felt.
What is a good CSAT score?
Most healthy SaaS support teams sit somewhere in the low-to-mid eighties. Here is the working rubric this calculator uses — a rule of thumb for reading your own number, not an industry standard.
Genuinely rare. Worth checking your sample isn't self-selecting before you celebrate.
Where a healthy SaaS support team usually lands. Protect it.
Usually a speed or consistency problem rather than a people problem.
Something structural is breaking often enough for customers to notice.
Customers are actively telling you to fix this. Treat it as an outage.
Cross-industry averages move a lot between sectors, and a software CSAT is not comparable to an airline's. If you need an external reference point for your specific industry, the American Customer Satisfaction Index publishes sector-level scores each year. Failing that, the most useful benchmark you have is your own score last quarter.
Why your CSAT score is probably lying to you
CSAT is a useful number that is easy to over-read. Four things distort it in almost every support team.
Your sample is too small
At 40 responses, an 85% score carries roughly ±11 points of uncertainty. Last month's drop from 88% to 82% may be nothing at all. The calculator above shows this interval live for your own numbers.
Only the extremes answer
Response rates on support surveys are low, and the people who bother to reply skew delighted or furious. The quietly satisfied middle is systematically underrepresented.
You're measuring the agent, not the experience
Customers rate the last human they spoke to. That quietly hides everything that happened in the queue beforehand — the wait, the bounce between inboxes, the repeated question.
Timing skews everything
A survey sent the moment a ticket is resolved scores higher than the same survey sent 48 hours later. If you change when you ask, you change the score without changing the service.
CSAT vs NPS vs CES
These three get used interchangeably in board decks and they answer completely different questions. Most support teams need CSAT and CES; NPS belongs to the company, not the queue.
| Metric | Question | Scale | Tells you |
|---|---|---|---|
| CSAT | How satisfied were you? | 1–5 typical | Whether one interaction went well |
| NPS | How likely are you to recommend us? | 0–10 | How the overall relationship is trending |
| CES | How easy was it to get this resolved? | 1–7 typical | How much friction your process added |
CSAT is the right default for support because it is tied to a specific ticket, which makes it actionable. NPS moves too slowly to tell you whether yesterday's queue was handled well.
Low CSAT is usually a queue problem, not an empathy problem
When CSAT slips, the instinct is to coach agents on tone. But the most consistent drivers of support satisfaction are structural, and none of them are about how warmly the reply was written.
How long the customer waited
Response time tracks satisfaction more reliably than almost anything else you directly control.
Whether they had to repeat themselves
Every handoff and every "can you confirm your account email?" costs you points that no apology recovers.
Whether the answer was consistent
Two agents giving two different answers to the same question reads as incompetence, however polite both were.
Here is the uncomfortable part. In most SaaS teams a large majority of the queue is the same twenty questions rephrased. That repetition consumes the hours that would otherwise go into fast, careful replies on the genuinely hard tickets. Your agents are not slow — they are re-typing yesterday's answer.
DelDesk answers the repeats automatically. Your team answers a question once, approves the resolution with one click, and every future customer asking it gets an instant, consistent answer. The hard tickets get the time they deserve — which is where CSAT is actually won.
Stop re-answering the tickets that drag your score down
DelDesk turns every solved ticket into permanent automation, so your team spends its hours where satisfaction is actually won.
No credit card required · Unlimited teammates · $49/mo flat after trial
Want the volume side of this? See what repeat tickets cost you.
Frequently asked questions
What is a CSAT score?
CSAT (Customer Satisfaction Score) is the percentage of survey respondents who said they were satisfied with an interaction, product or service. It is usually collected with a single question — "How satisfied were you?" — answered on a 1-5 scale, and reported as a percentage between 0 and 100.
How do you calculate CSAT?
Divide the number of satisfied responses by the total number of responses, then multiply by 100. "Satisfied" normally means the top two ratings on your scale, so a 4 or a 5 on a 1-5 scale. If 460 of 547 people rated you 4 or 5, your CSAT is (460 ÷ 547) × 100 = 84.1%.
What is a good CSAT score?
Most healthy SaaS support teams land between 80% and 89%. Above 90% is exceptional and worth checking for sample bias. Below 70% usually points at a structural problem — slow responses or inconsistent answers — rather than a people problem. Benchmarks vary widely by industry, so compare against your own trend first.
What's the difference between CSAT and NPS?
CSAT measures satisfaction with a specific interaction and is usually asked immediately after it. NPS measures how likely someone is to recommend your company overall, on a 0-10 scale, and is asked periodically. CSAT tells you whether a ticket went well; NPS tells you how the relationship is trending.
Should neutral responses count in CSAT?
They stay in the denominator but do not count as satisfied. Dropping neutral responses entirely is the most common way teams accidentally inflate their own score — it can add five to ten points without anything actually improving. Keep every response in the total.
How many responses do I need for a reliable CSAT score?
Below about 30 responses a CSAT score is closer to noise than signal. At 40 responses an 85% score carries roughly ±11 points of uncertainty, so a drop to 80% may mean nothing at all. At 500 responses that uncertainty falls to about ±3 points. This calculator shows the interval for your own sample.
Can I calculate CSAT on a 1-10 scale?
Yes. On a 1-10 scale the top-2-box method counts ratings of 9 and 10 as satisfied. This calculator supports 1-3, 1-5, 1-7 and 1-10 scales. The important thing is to pick one scale and keep it — changing scales mid-quarter makes your trend meaningless.
What's the difference between top-2-box and average CSAT?
Top-2-box reports the percentage of people who chose one of the top two ratings. Average CSAT reports the mean rating, sometimes converted to a percentage of the scale. They are different measurements: a 4.32 average out of 5 (86.3%) and an 84.1% top-2-box score describe the same data but are not comparable to each other.
How often should you measure CSAT?
Send the survey continuously — triggered after each resolved ticket — but review the score on a rolling window large enough to be meaningful, usually monthly. Reading a weekly CSAT number off a few dozen responses mostly means reacting to sampling noise.
Other support calculators
Ticket Deflection Rate CalculatorSoon
Work out what share of your support queue never needs a human, and what that is worth in hours.
CES CalculatorSoon
Score customer effort on a 1-7 scale and see how much friction your resolution path is adding.
NPS CalculatorSoon
Turn promoter, passive and detractor counts into a Net Promoter Score you can actually act on.
First Response Time CalculatorSoon
Average your first-reply times and compare them against the thresholds where satisfaction starts to drop.
Cost Per Ticket CalculatorSoon
Break down what a single support ticket really costs once salaries, tooling and overhead are counted.