Quick answer
CSAT stands for customer satisfaction score. It is based on one question asked right after a specific experience, such as a support chat, and is usually reported as the share of customers who rated it satisfied (4 or 5 on a 1–5 scale). It measures one interaction, not loyalty.
Formula: CSAT (%) = satisfied responses ÷ total responses × 100
What does CSAT mean?
CSAT stands for customer satisfaction score. It comes from one short question asked right after a specific experience, such as a support chat, a purchase or an onboarding session: "How satisfied were you with this?" The customer answers on a scale, usually 1 to 5, and the result is reported as the share of customers who were satisfied.
Two things set CSAT apart. It measures a moment, not the whole relationship, so it is a good tool for judging a conversation, a feature or a delivery and a poor one for judging loyalty. And it is only as good as the people who answer: a score built from the 5% of customers who bothered to reply says something different from one built from 60%.
How to calculate CSAT
There are two common methods, and they give different numbers from the same answers, so say which one you use.
| Method | Formula | What counts as satisfied |
|---|---|---|
| Share satisfied (top-two box) | satisfied responses ÷ all responses × 100 | 4 and 5 on a 1–5 scale |
| Average score | sum of all scores ÷ number of responses (or ÷ 5 × 100 as a percentage) | Every score contributes |
The first method is the one most people mean when they quote a CSAT percentage. The second keeps more information, because a 3 counts for something, but it hides how many customers were unhappy. Use the share method for a headline number and look at the full distribution beside it.
A worked example
Sample data: 200 customers rate a support conversation on a 1–5 scale.
| Rating | Responses | Share | Points |
|---|---|---|---|
| 5 | 92 | 46% | 460 |
| 4 | 68 | 34% | 272 |
| 3 | 20 | 10% | 60 |
| 2 | 12 | 6% | 24 |
| 1 | 8 | 4% | 8 |
| Total | 200 | all | 824 |
- CSAT, share method: (92 + 68) ÷ 200 = 160 ÷ 200 = 80%.
- Average score: 824 ÷ 200 = 4.12.
- Average as a percentage of the maximum: 824 ÷ (200 × 5) = 824 ÷ 1,000 = 82.4%.
Same 200 answers, three numbers: 80%, 4.12 and 82.4%. None is wrong; they answer different questions. The only mistake is comparing a share-method figure from one team with an average-based figure from another.
How many responses do you need?
A satisfaction score from a few answers is noise. A rough margin of error for a share is 1.96 × √(p × (1 − p) ÷ n), where p is the share satisfied and n the number of responses.
| Responses (n) | Share satisfied | Margin of error (95%) |
|---|---|---|
| 200 | 80% | ±5.5 points |
| 30 | 80% | ±14.3 points |
With 200 answers, "80%" is somewhere between about 74% and 86%. With 30 answers it could be anywhere from 66% to 94%, and you cannot call a move from 80% to 75% a decline.
The same arithmetic stops you ranking people on thin data. Sample data: agent A has 45 satisfied out of 50 ratings (90%, ±8.3 points) and agent B has 30 of 40 (75%, ±13.4 points). Their ranges overlap (about 82–98% and 62–88%), so the 15-point gap is not proof that A is better. Collect more ratings before drawing conclusions about individuals.
CSAT vs NPS vs CES
| CSAT | NPS | CES | |
|---|---|---|---|
| Asks | How satisfied were you with this? | How likely are you to recommend us? | How easy was it to get this done? |
| Scale | Usually 1–5 | 0–10 | Usually 1–7 |
| Measures | Satisfaction with one interaction | Overall loyalty and advocacy | Effort of one interaction |
| When to ask | Right after the interaction | Periodically, on the relationship | Right after a task or support case |
| Reported as | % satisfied or average | Promoters minus detractors | Average or % easy |
CSAT is the one to use for quality of a single touchpoint. For the effort side, see customer effort score. For whether customers would recommend you, the Net Promoter Score was introduced in a 2003 Harvard Business Review article.
What is a good CSAT score?
There is no universal answer. Guides from vendors often call roughly 75 to 85% good for support, but surveys differ in scale, wording, timing and who answers, so an outside figure is a weak target. A better test is your own trend measured the same way: is the score for the same kind of interaction rising, flat or falling, and does it move when you change something? Compare like with like: chat with chat, one product with the same product.
A survey question you can copy
Keep it to one question and one optional comment box. Wording: "How satisfied were you with the help you received today?" Labels: 1 Very dissatisfied, 2 Dissatisfied, 3 Neutral, 4 Satisfied, 5 Very satisfied. Follow-up, shown only after a 1, 2 or 3: "What could we have done better?" Send it right after the conversation ends, in the same window where it took place, so that answering costs one tap. Every extra field lowers the response rate, and a lower response rate makes the score less trustworthy.
Common mistakes
- Surveying only the happy path. If the survey is sent only after a ticket is resolved, you never hear from people whose problem was not solved. Include unresolved and abandoned cases in some way.
- Ignoring the response rate. Report the number of responses and the share of customers who answered next to the score.
- Changing the scale between periods. Moving from 1–5 to thumbs up or down breaks the trend line.
- Averaging across unlike things. A billing question and a technical outage are different experiences; segment before you compare.
- Treating CSAT as loyalty. A satisfied customer can still leave, which is why effort and recommendation are measured separately.
- Waiting too long to ask. Ask while the experience is fresh, as Zendesk's guide also advises.
To raise the score rather than measure it, see improve CSAT. For the bigger KPI picture, see customer success KPIs. Related terms: customer sentiment and resolution rate.
Reading CSAT in TapCub
TapCub's live chat collects a satisfaction rating after a conversation, and the report shows an average. Here is how the example above maps onto it.
- Find the number. Open Live chat › Reports. Satisfaction is a score from 1 to 5 that the visitor gives within 24 hours after the conversation ends, averaged to one decimal. Each conversation can be rated once. In the example, the report would show 4.1.
- Convert it if you need a percentage. The average divided by 5 gives the average-method percentage (4.12 ÷ 5 = 82.4%). The share-satisfied figure (80%) needs the count of 4s and 5s, which the report does not give; the report gives the average.
- Look per person. The agent workload table shows satisfaction next to handled and resolved counts, first response and transfers. Apply the sample-size rule before comparing agents.
- Do not forget the AI. If the AI answered and no agent ever replied, the rating is credited to the AI, not to any agent.
- Put it next to the other numbers. Read satisfaction with the resolution rate (resolved conversations ÷ all conversations) and first response from the same report.
TapCub does not run NPS or CES surveys; it reports the chat satisfaction rating only.
Sources
- Zendesk, Customer satisfaction score (CSAT): what it is and how to calculate it.
- Frederick F. Reichheld, The One Number You Need to Grow, Harvard Business Review, December 2003.
- Matthew Dixon, Karen Freeman and Nicholas Toman, Stop Trying to Delight Your Customers, Harvard Business Review, July–August 2010.
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.

Product Analyst
Focuses on event analysis, conversion funnels, and retention, using clear metric definitions and structured analysis to explain product performance.