CSAT: calculation, scales and reliability of the customer satisfaction score

What is CSAT?

CSAT is an “in the moment” indicator. It covers a recent event, whereas the Net Promoter Score℠ (NPS®) asks about the overall relationship with the company, with a single question rated from 0 to 10 (Bain & Company). The two complement each other; their differences and the limits of NPS are detailed on our Net Promoter Score page.

No public CSAT standard was found among the sources consulted. Customer service software sets it up in its own way. Zendesk, for example, lets the administrator choose the question, the rating scale and the labels (Zendesk help). A CSAT therefore only makes sense together with its question, its scale and its formula.

How do you calculate CSAT?

The easiest method to read counts satisfied customers. On a 1 to 5 scale, these are ratings of 4 and 5: this is called “top 2 box”, the two top boxes.

CSAT (%) = number of 4 or 5 ratings ÷ total number of responses × 100

This is the logic used by Zendesk, which converts any scale into “good” or “bad” ratings. Its documentation on the CSAT score gives the following mapping:

Under this convention, a 3 out of 5, the neutral rating, counts as a bad rating. Other conventions are possible: leaving out the neutral rating, counting only 5s, calculating an average. The gap is far from theoretical, as the following example shows.

Fictitious example: 200 customers answer on a 1 to 5 scale. 90 give 5, 60 give 4, 20 give 3, 15 give 2 and 15 give 1.

These values are calculation results on a fictitious example. The same answers produce four scores: before comparing two CSATs, check that they come from the same formula.

Which question and which scale should you choose?

A good CSAT question covers a single, clearly named event: “Are you satisfied with how your request was resolved?” rather than “Are you satisfied with us?”. A broad question mixes product, price and service, and does not tell you what to fix.

The choice of scale depends on what you want to do with it.

Zendesk presents its survey in the simplest form: a positive or negative rating and an optional comment. For negative ratings, the tool can offer a list of reasons (Zendesk CSAT resources). That list turns a score into an action plan: wait too long, issue not resolved, unclear answer.

Keep the same question and the same scale over time. Changing either one breaks the series: the next rise or fall will no longer mean anything.

When should you send the satisfaction survey?

As close as possible to the event being measured, so the customer remembers it. By default, Zendesk sends the survey 24 hours after a ticket is solved; this delay is set in the corresponding automation (Zendesk help). The customer then has 28 days to respond.

Ask the same customer sparingly. Surveys sent by email follow the rules of your email marketing programme: sending pressure, unsubscribe, content.

How many responses do you need for a reliable CSAT?

A CSAT calculated on few responses varies a lot from one period to the next without satisfaction having changed. The margin of error of a percentage is calculated with the confidence interval formula for a proportion, given in the NIST statistical handbook: p ± 1.96 × √(p × (1 – p) ÷ n) at 95%.

These margins are calculation results. How to read them: with 100 responses a month, a move from 80% to 76% stays within the margin and does not prove a decline. To track small changes, group responses over a longer period or by quarter. NIST notes that this approximation becomes less accurate when the sample is very small.

What is the customer effort score (CES)?

The customer effort score measures the effort a customer had to make to get what they wanted, most often the resolution of a request. It was introduced in 2010 in Harvard Business Review by Matthew Dixon, Karen Freeman and Nick Toman, of the Corporate Executive Board (“Stop Trying to Delight Your Customers”).

The study covers more than 75,000 B2C and B2B customers, surveyed over three years after a contact by phone, voice prompts, web, chat or email. The authors define loyalty as the intention to keep buying, increase spending or say good things about the company. Their conclusion: CES predicts this loyalty better than satisfaction measures or NPS.

In practice, CES is asked at the same time as CSAT, after a contact with customer service. CSAT tells you whether the customer is happy; CES tells you whether the journey cost them. A customer satisfied with an answer obtained after three reminders will give a good CSAT and a poor CES: that signals a process to simplify. The CES question and its original scale are presented on our NPS page.

What is a good CSAT score?

A good CSAT is a score that improves on your own history, measured with the same question, the same scale and the same formula. No industry benchmark published with a verifiable method was found among the sources consulted on November 2, 2026: the figures in circulation mix different scales and formulas.

1. Set a question, a scale and a formula, and document them.
2. Measure for at least a quarter to know your starting level and its normal variation.
3. Break the score down by channel, contact reason or product to find where it drops.
4. Read the comments on low ratings and call back the customers concerned.
5. Link CSAT to behaviour: repeat purchase, cancellation, new requests. A score that moves with none of these indicators measures poorly what matters to you.

CSAT complements the customer loyalty indicators that are retention and customer value. Combined with your sales data in your analytics tool, it stops being a reporting figure and becomes a warning signal.

Attribution required by Bain & Company: “Net Promoter®, NPS®, NPS Prism®, and the NPS-related emoticons are registered trademarks of Bain & Company, Inc., NICE Systems, Inc., and Fred Reichheld. Net Promoter Score℠ and Net Promoter System℠ are service marks of Bain & Company, Inc., NICE Systems, Inc., and Fred Reichheld.”