CSAT = positive responses ÷ total responses, for a specific interaction
The last four words are the whole distinction.
Transaction, not relationship
CSAT asks about a moment: this ticket, this call, this onboarding session. NPS asks about the relationship.
Using CSAT as a health measure produces a number driven by who happened to open a ticket. A customer quietly deciding not to renew never files one, so they never appear, and CSAT stays high through the quarter they leave.
For relationship health use behaviour: retention rate and health score.
The scale decides the number
Common scales: 1 to 5, 1 to 7, and a three-point satisfied/neutral/unsatisfied.
What counts as positive is a choice. On a 5-point scale, counting 4 and 5 is standard; counting 3 as positive inflates by a lot. On a 7-point, the top two or top three both appear in practice.
None of this is comparable across scales. Fix yours, write down what counts as positive, and treat any external benchmark as decorative.
When you ask changes what you get
- Immediately after resolution: highest response rate, and it measures the
interaction, which is what CSAT is for.
- Delayed by days: the answer drifts toward general sentiment and stops
being a transaction measure.
- Only after resolved tickets: excludes the customers whose issue was never
resolved, which is systematic and severe bias.
Send after every closure, resolved or not, or state the exclusion.
A worked example
800 tickets closed in the month, 240 responses. On a 5-point scale: 96 fives, 78 fours, 34 threes, 20 twos, 12 ones.
CSAT (4 and 5) = 174 ÷ 240 = 72.5%
Counting 3 as positive = 208 ÷ 240 = 86.7%
Same month, fourteen points apart, purely from the positive-response definition. Response rate is 30%, which should be published beside it.
The traps
- Using it as a relationship or churn predictor.
- Changing the scale or the positive threshold mid-series.
- Surveying only resolved tickets.
- Publishing without the response rate.
- Averaging the raw score and calling it CSAT, which is a different statistic.
Where this sits
CSAT is Measurement in the Tenbound Pipeline Architecture Standard and it is the narrowest metric in the library. It is genuinely useful for grading a support function and genuinely misleading for anything else.