What it means
A business cannot manage what it does not measure, and customer satisfaction was for a long time something managers felt rather than measured. CSAT gives it a number.
The customer is asked a single question, usually "How satisfied were you with [the product, the service, this interaction]?", and answers on a scale, most often 1 to 5 from very dissatisfied to very satisfied. The responses are aggregated and reported either as the percentage of respondents who chose the top two points, which is the most common convention, or as the mean score.
The question can be asked about the relationship as a whole, in a periodic survey, or about a single transaction, in a survey triggered immediately after it. Transactional CSAT is the more operationally useful of the two.
Asked within minutes of a support call or a delivery, it captures the customer's reaction to that specific event while it is fresh, and it can be attributed to the agent, the branch, the product line or the process step involved. A support team can see its CSAT by agent and by issue type; a logistics operation can see it by carrier and by region; a hotel can see it by property and by department.
This granularity turns the score from a headline into a diagnostic tool, because it shows where dissatisfaction is generated and lets the business act on causes rather than symptoms. The measure has well-understood limitations.
It captures a moment, not loyalty: a customer can be satisfied with every interaction and still switch to a cheaper competitor, which is why CSAT is often paired with net promoter score (willingness to recommend) and with behavioural measures such as retention and repeat purchase. It suffers from response bias: the customers who answer are not a random sample, and the happiest and the angriest are over-represented.
It is sensitive to how and when the question is asked, so scores are not comparable between companies that ask differently. And it can be gamed, by staff who ask only satisfied customers to complete the survey or who coach customers on the score.
Interpreting the number requires context. A CSAT of 80% means eight in ten respondents were satisfied, but the 20% who were not represent the opportunity, and the comments they leave in the optional free-text box are usually worth more than the score itself.
Driver analysis, which correlates CSAT with attributes of the interaction (wait time, first-contact resolution, delivery lateness, agent tenure), identifies which factors move the score most, and therefore where investment will have the most effect. A common finding is that a small number of operational failures, such as long waits or repeat contacts, account for most dissatisfaction.
The financial connection is through retention and cost to serve. Companies that segment their customers by satisfaction consistently find that dissatisfied customers churn at several times the rate of satisfied ones, complain more, cost more to handle and say negative things to others.
Putting a value on moving a customer from dissatisfied to satisfied, through the difference in retention and margin, converts CSAT from a soft measure into a business case, and allows service improvements to be evaluated like any other investment.
In practice
Real-world examples.
Example
A telecoms company finds that CSAT for support calls resolved on first contact is 88% and for calls requiring a second contact is 51%, and makes first-contact resolution its primary support metric.
Example
A software company's CSAT after onboarding calls averages 4.6 out of 5, but its CSAT after billing queries averages 3.1, which points to a billing process problem rather than a people problem.
Example
A restaurant chain's app-based survey shows a 90% CSAT while its online reviews average 3.2 stars, and an investigation finds that the survey is only shown to customers who used the loyalty scheme, a biased sample.
Think of it
“CSAT simply asks customers 'Are you satisfied?'-a direct measure of happiness with your service.
Formula
Calculation
CSAT (percentage method) = Number of respondents rating 4 or 5 on a 5-point scale / Total respondents x 100
CSAT (mean method) = Sum of all ratings / Total respondents
Response rate = Responses received / Surveys sent x 100
Margin of error (95% confidence) = 1.96 x square root of (p x (1 minus p) / n), where p is the CSAT proportion and n the number of responses
Worked example. A retailer sends a one-question survey after each online order. In a month it sends 8,000 surveys and receives 1,000 responses (12.5% response rate) distributed as follows: 5 stars 420; 4 stars 310; 3 stars 150; 2 stars 70; 1 star 50.
- Satisfied responses (4 and 5) = 420 + 310 = 730
- CSAT = 730 / 1,000 = 73%
- Mean score = (5 x 420 + 4 x 310 + 3 x 150 + 2 x 70 + 1 x 50) / 1,000 = (2,100 + 1,240 + 450 + 140 + 50) / 1,000 = 3,980 / 1,000 = 3.98
- Margin of error = 1.96 x square root of (0.73 x 0.27 / 1,000) = 1.96 x 0.014 = about 2.8 percentage points, so the true figure is likely between 70% and 76%
Segmented view. Splitting the responses: orders with no support contact scored 81%; orders where the customer contacted support scored 68%; orders delivered late scored 44%. Late delivery is the largest driver. A single store with 50 responses reporting 73% has a margin of error of about 12 points (1.96 x square root of (0.73 x 0.27 / 50) = 1.96 x 0.063), so store-level monthly scores should be read over several months, not one.
Financial linkage. The retailer has 20,000 active customers, of whom about 27% (5,400) would rate themselves dissatisfied or neutral. Satisfied customers are retained at 92% a year and dissatisfied ones at 60%. Each retained customer generates $400 of annual gross margin. If service improvements moved 1,000 customers from the dissatisfied to the satisfied group, retention among those 1,000 would rise from 600 to 920: 320 additional customers retained, worth 320 x $400 = $128,000 of gross margin in the first year, and more thereafter as the effect compounds.Case study
Seen in the real world.
A hotel group of 40 properties reported a group CSAT of 84% to its board each quarter and treated the figure as evidence that service was under control, even as online review scores drifted down and repeat bookings fell. A new chief operating officer asked how the survey was distributed and found that it was sent through the group's mobile app, which was used mainly by loyalty members, and that guests who had complained during their stay were routinely excluded "to avoid upsetting them further". The measured 84% described the group's happiest guests, not its guests.
The survey was rebuilt: sent by email to every guest 24 hours after checkout, one question plus a free-text box, with no exclusions. The first full quarter's CSAT was 71%, a shock to the board but the first honest number it had seen. Driver analysis on 9,000 responses showed that check-in waiting time explained more of the variation than any other factor: guests who waited under five minutes scored 83%, those who waited over fifteen scored 52%.
The average wait at the busiest properties was nine minutes at peak. The group invested $600,000 in mobile check-in and reallocated front-desk rosters to match arrival patterns, cutting the average peak wait to three minutes.
Two quarters later group CSAT was 79% on the honest basis, review scores had risen, and repeat booking rates at the eight properties with the worst original waits had improved by four percentage points. On the group's average guest value, the retention improvement alone was worth about $1,400,000 a year against the $600,000 spent. The board's lesson was twofold: a satisfaction score is only as good as the sample it is drawn from, and the score becomes useful only when it is connected to the operational causes that move it.
Watch out
Common mistakes.
- Surveying a biased sample, such as loyalty members only or customers who did not complain, which produces a flattering score that hides the problems the measure exists to find.
- Comparing CSAT figures between companies or business units that ask different questions, on different scales, at different moments; the number is only comparable on a consistent method.
- Reading small differences as meaningful when the response count is low; a branch with 50 responses has a margin of error of more than ten points.
Questions
People also ask.
What is the difference between CSAT and net promoter score?
CSAT measures satisfaction with a product, service or interaction, usually immediately afterwards; NPS measures willingness to recommend the company, which relates more to loyalty. They answer different questions and are often used together.
What is a good CSAT score?
Percentages above 80% are generally considered strong, but the useful benchmark is the company's own trend and its competitors on a like-for-like method. A score should be read alongside response rate and sample composition.
Should CSAT be linked to staff pay?
With care. Linking it directly encourages staff to select who is surveyed or to lobby for scores. Using it as one input among several, with sample integrity checks, is the safer approach.
From the founder's library
Take it further with the book.
Build your financial confidence beyond this definition. Shihan's full-length guide, Accounting Fundamentals, takes the same plain-English approach and turns it into a complete, practical playbook for non-finance managers, business owners and students - with chapter-end quiz answers and presentation slides included.
25% off with code MMHQ25, applied at checkout. Priced in USD - checkout may show the equivalent in your local currency.
View the book and save 25%