What is a good CSAT score?

These figures come from survey vendors, from their own customers' surveys.

Survicate's report covers 1,259 CSAT surveys from 370 companies. Half of them scored between 64.8% and 90.0%, so a score of 70% sits well inside the normal range. Simplesat's 2026 report puts its benchmark at 87.8%, but its summary page does not say which answers count as satisfied. Use a benchmark only if it says what it counted and divided.

Most tools ask "How satisfied were you with...?" on a 1 to 5 scale, then turn the answers into a score in one of two ways:

Take 100 answers: 50 fives, 30 fours, 10 threes and 10 ones. The share is 80% and the average is 4.1, so a benchmark of "4.3" and one of "85%" cannot be put side by side. The two also move differently: if the readers who gave 1 start giving 3, the average rises and the share stays at 80%, because a 3 is still not a 4.

Scales add another layer. SurveyMonkey and Survicate count 5 to 7 as satisfied on a 7-point scale and 6 to 10 on a 10-point scale, so half the points count on a 10-point scale against two of five on a 5-point one. Report the share of 4 and 5 answers as your CSAT, put the average beside it, and only compare 1 to 5 scores with other 1 to 5 scores.

The ACSI publishes satisfaction scores for US industries and companies that look a lot like CSAT figures: scores out of 100 that read like percentages. They are a different kind of number.

An ACSI score is an index. A US government report that used the ACSI method (prepared by CFI Group for US Citizenship and Immigration Services) describes it as a weighted average of three questions: overall satisfaction, satisfaction compared with expectations, and satisfaction compared with the ideal. Each is answered on a 1 to 10 scale, and the result is converted to a 0 to 100 scale for reporting.

So an ACSI score of, say, 75 means the weighted answers sit three quarters of the way up the scale, not that 75% of customers were satisfied. It is closer to an average rating than to a share of 4 and 5 answers.

The mix-up is common because both numbers tend to land in the 70s and 80s. SurveyMonkey's CSAT guide, for example, uses ACSI industry scores as its industry benchmarks and writes them with a percent sign. If an industry benchmark comes from the ACSI, leave it out of any comparison with your share of 4 and 5 answers.

Even with the same calculation, scores vary a lot by industry. Survicate's 2025 medians:

Where the survey appears matters too. In the same data, surveys opened from a page or a shared link had a median of 86.2%, and surveys shown in a website widget 70.1%. The report does not explain the gap. Part of it is likely who sees each one: a link usually goes to someone who has just finished a purchase or a support case, while a widget is shown to anyone who visits the page.

So before you call a score good or bad, check that the benchmark matches yours on calculation (share or average), scale, the moment you ask (after a support case, an order or at the end of a page), placement (widget or link) and industry. If no outside figure matches, compare your own pages, products or teams with each other.

CSAT comes from customer service and e-commerce: a rating after a support case or an order. The usual question at the end of an article or docs page is "Was this helpful?", answered Yes or No. The share of Yes answers is calculated like CSAT, but it is a different question, so do not set it against CSAT benchmarks.

We did not find a published CSAT benchmark for article pages. The nearest figures are the media and marketing median and the widget median above, and both mix many kinds of survey. A 1 to 5 rating on an article still shows how strongly readers feel: a page that averages 3.2 tells you something a Yes / No split does not. If you use one, report the share of 4 and 5 answers and the average for each page, and rank pages against each other.

Asklet shows a short survey on your pages and groups the answers by page path. Its Rating question asks for 1 to 5 stars, and the results show the average and how many answers each star got. It does not report the share of 4 and 5 answers itself: add up the 4s and 5s, or enter the totals in the CSAT calculator.

For the content-page question, a Yes / No first question gives a "found it useful" figure for each page. Answers are anonymous: no cookies, no tracking identifiers and no sign-in for readers.