CSAT (customer satisfaction score) means the percentage of surveyed customers who say they were satisfied with a specific interaction, product or service. If 80 of 100 respondents pick a positive rating, your CSAT is 80%.
In short
- It is usually asked right after a support interaction closes.
- Most teams count the top two ratings, like 4 and 5 out of 5.
- Your own trend matters more than any industry benchmark.
- CSAT only hears from customers who answer. QA covers the rest.
What does CSAT mean?
CSAT stands for customer satisfaction score. It is a survey-based metric that tells you what share of customers were happy with one specific experience, most often a support conversation, a delivery or an onboarding step.
The key word is specific. CSAT is transactional: it asks about the thing that just happened, not about the customer’s whole relationship with your company. That makes it fast to collect and easy to tie back to a team, a channel or an agent. It also makes it narrow, which we come back to below.
In customer support, CSAT is usually the first metric a team adopts, because every help desk (Zendesk, Salesforce, Intercom and the rest) can send the survey automatically when a ticket is solved.
How is CSAT measured?
A CSAT survey is one question, sent straight after the interaction, with a short rating scale. The classic wording is:
“How satisfied were you with your experience today?”
Variations are fine as long as you keep the same wording over time. Changing the question changes the score, and then your trend line stops meaning anything.
The most common scales:
| Scale | What counts as positive | Where you see it |
|---|---|---|
| 1 to 5 (very unsatisfied to very satisfied) | 4 and 5 | The default in most help desks and survey tools |
| 1 to 3, or sad, neutral, happy faces | 3 only | Chat widgets and quick in-app surveys |
| Good or bad, thumbs up or down | Good | Zendesk’s built-in satisfaction survey and many email footers |
| 1 to 7 or 1 to 10 | Top two or three ratings | Research-style surveys, less common in support |
Most teams add an optional free-text box (“What could we have done better?”). The score tells you how many customers were unhappy. The comments are where you find out why, at least for the ones who bother to write.
Timing matters too. Send the survey when the issue is actually resolved, not when the first reply goes out. A survey that arrives while the customer is still waiting measures their patience, not your service.
How to calculate CSAT: formula and worked example
The formula:
CSAT = (number of positive responses / total number of responses) x 100
Positive means the top ratings on your scale, usually 4 and 5 on a five-point scale. Neutral and negative responses count in the total but not in the top.
Worked example. Last month your team solved 2,000 tickets and sent a five-point survey on each. You received 400 responses:
| Rating | Responses |
|---|---|
| 5 (very satisfied) | 230 |
| 4 (satisfied) | 90 |
| 3 (neutral) | 40 |
| 2 (unsatisfied) | 25 |
| 1 (very unsatisfied) | 15 |
| Total | 400 |
Positive responses are 230 + 90 = 320. CSAT = 320 / 400 x 100 = 80%.
Two things to note from the same numbers. First, the response rate was 400 / 2,000 = 20%, so the score describes one conversation in five. Second, some teams report an average rating instead (here, 4.2 out of 5). Both are legitimate, but they are different numbers, so say which one you use and never compare one to the other.
What is a good CSAT score?
A good CSAT score is one that is stable or rising against your own baseline, measured with the same question and scale. There is no universal target, because scale, wording, channel, industry and response rate all move the number.
If you want an external reference, the best-known one is the American Customer Satisfaction Index (ACSI), which put overall US customer satisfaction at 77.0 on a 0 to 100 scale in the first quarter of 2025. Treat it as context, not a target. The ACSI is an index built from several survey questions about a company as a whole, so it is not the same calculation as a support team’s percentage of 4s and 5s.
Some practical ways to judge your own score:
- Compare like with like. Chat CSAT tends to differ from email CSAT. Benchmark each channel against itself.
- Watch the response rate. A CSAT of 95% from a 3% response rate tells you less than 85% from a 30% response rate.
- Look at the bottom, not just the average. A steady 80% can hide one queue or one issue type that has collapsed.
- Investigate drops early. A score drifting down for three months is a signal, even if it still looks healthy on paper.
CSAT vs NPS vs CES
CSAT is one of three survey metrics most support teams use. They answer different questions, so most mature teams run more than one. We compare them in depth in CSAT vs NPS vs CES.
| CSAT | NPS | CES | |
|---|---|---|---|
| Question | How satisfied were you with this experience? | How likely are you to recommend us to a friend or colleague? | How easy was it to get your issue resolved? |
| Scale | Usually 1 to 5 | 0 to 10 | Usually 1 to 5 or 1 to 7 |
| Measures | Satisfaction with one interaction | Overall loyalty to the company | Effort the customer had to spend |
| Timing | Right after the interaction | Periodically, for example every quarter | Right after the interaction |
| Score | % of positive responses | % promoters minus % detractors | Average score or % of easy ratings |
| Best for | Judging a specific team, channel or agent | Tracking the relationship over time | Finding friction in processes |
Rule of thumb: use CSAT to see how a conversation landed, CES to see what made it hard, and NPS to see whether the relationship is holding.
The limits of CSAT
CSAT is useful and cheap. It also has well-known blind spots that every support lead should keep in mind before building targets or bonuses on it.
It only covers customers who answer. In the example above, 80% of conversations produced no rating at all. Whatever happened in those 1,600 tickets is invisible to the metric. Response rates in support are often low, so the unseen majority is usually much bigger than the measured minority.
The customers who answer are not typical. People who were delighted or furious are more likely to respond than people who were mildly let down. Mildly let down is exactly the group that quietly leaves.
It mixes up the agent and the company. A customer who hates your refund policy rates the agent who explained it. A low score can mean a bad answer, a bad product, a bad policy or a bad day. CSAT alone cannot tell you which. We cover this in why QA scores and CSAT disagree.
It says how the customer felt, not what happened. A customer can be satisfied with a wrong answer that will cause a second contact next week. Another can be unhappy with a correct, compliant answer they did not like. Neither score tells you whether your team followed the process.
It can be gamed. Agents who know they are judged on CSAT learn to ask for good ratings, avoid surveys on difficult tickets or give away concessions. None of that improves service.
The result: most bad conversations never show up in your CSAT at all. They sit in tickets nobody rated.
How QA fills the gap CSAT leaves
Quality assurance looks at the other side of the conversation. Instead of asking the customer how they felt, it scores what the agent actually did against your own standards: accuracy, tone, process, resolution. Done manually, QA reviews a small sample of tickets, so it shares some of CSAT’s coverage problem.
Scoring every conversation closes that gap. Kaizo’s quality assurance scores 100% of your support conversations, whether a human or an AI agent handled them, against your own scorecard. That includes the large share of tickets with no survey response, so problems surface whether or not the customer chose to tell you.
Kaizo measures internal quality through its Internal Quality Score, or IQS, and connects that conversation quality to outcome metrics like CSAT, so leaders can see how what happens inside a ticket moves what the customer reports afterward. Because Kaizo is native to Zendesk and Salesforce, it reads those conversations straight from the systems your team already runs.
Read together, the two numbers are far more useful than either alone. A CSAT dip becomes coachable when you can trace it to the conversations, and the specific behaviors, behind it.
Frequently asked questions
What does CSAT stand for?
CSAT stands for customer satisfaction score. It is the percentage of surveyed customers who rate a specific interaction, product or service positively, usually 4 or 5 on a five-point scale.
What is the CSAT formula?
CSAT equals the number of positive responses divided by the total number of responses, multiplied by 100. If 320 out of 400 customers give a 4 or 5, your CSAT is 80%. Positive usually means the top one or two ratings on your scale.
What is a good CSAT score?
A good CSAT score is one that holds steady or improves against your own baseline, using the same question and scale. For outside context, the ACSI put overall US customer satisfaction at 77.0 out of 100 in early 2025, but that index is calculated differently from a support team’s CSAT, so use it as a reference point only.
What is the difference between CSAT and NPS?
CSAT measures satisfaction with a specific interaction and is scored as a percentage of positive responses. NPS measures overall loyalty and likelihood to recommend on a 0 to 10 scale. CSAT is transactional and immediate, while NPS looks at the wider relationship.
Why is CSAT not enough on its own?
CSAT only reflects the customers who answer the survey, and those customers lean toward the very happy and the very unhappy. It also cannot separate agent performance from product or policy problems. Pairing it with QA scores on every conversation shows what happened in the tickets nobody rated.
Related terms
- NPS
- CES
- CSAT vs NPS vs CES
- First contact resolution
- Internal Quality Score
- Customer service metrics and KPIs
Connect CSAT to what happens in your conversations
See how Kaizo scores the conversations behind your CSAT, so a satisfaction dip points to the coaching that fixes it.
