A CSAT score is only as good as the survey that produced it. Ask the wrong question at the wrong moment and you get a number that looks great in a board deck and tells you nothing about how to improve. The mechanics of the survey — wording, scale, timing, and what happens after a low score — matter more than the headline percentage.

Ask one question, at the moment of truth

CSAT is a transactional metric: it measures this interaction, not your product overall. So ask it the instant the interaction ends — on ticket resolution — while the experience is fresh. One question:

How would you rate the support you received?

Resist the urge to bolt on "and how do you feel about our product, our pricing, and our roadmap?" Those are different surveys with different owners. A support CSAT that drags in product satisfaction confounds the score and points the blame at the wrong team — the same trap covered in CSAT vs. NPS.

Choose a scale you can act on

The two defensible choices:

  • A 5-point scale (Very dissatisfied → Very satisfied). Rich enough to see a distribution, simple enough to answer in one tap.
  • A binary thumbs up / thumbs down. Maximizes response rate; loses nuance.

Whatever you pick, define your "satisfied" threshold up front and never move it. CSAT% is usually the share of responses in the top one or two boxes. Quietly counting a neutral 3 as "satisfied" is how teams fool themselves.

Design for response rate

A 95% CSAT on a 4% response rate is mostly noise. Push the rate up:

  • Keep the first click inside the email or chat. Make the rating itself the email — one tap to score, then an optional comment box on the landing page.
  • Never make the comment mandatory. Forcing text tanks completion.
  • Send once. One ask per resolved ticket is the ceiling; survey fatigue is real and it poisons the well for every other survey you run.

The follow-up is the whole point

The score is a smoke alarm; the comment is where the fire is. Build the workflow, not just the metric:

  • Every 1 or 2 rating triggers a follow-up — ideally a human reply, automatically reopening or escalating the ticket.
  • Read low-score comments individually. Averages hide the two furious customers behind ten happy ones.
  • Close the loop visibly. A customer who rates you poorly and then hears back is often more loyal than one who never had a problem.

Route those alerts through your escalation workflow so a bad score lands in front of a lead, not in a dashboard nobody opens.

Watch the distribution, not the average

A single average flattens everything. Track the shape: a rising count of 1s alongside a steady average is a problem brewing under a calm number. Segment by agent, by topic, and by channel to find where satisfaction actually breaks — that is where the score turns into a fix.

The honest test

If your CSAT is high but your reopen rate and your churn are climbing, your survey is measuring politeness, not satisfaction. A good survey occasionally tells you something you did not want to hear. If yours never does, it is too easy to answer well — and it is not earning its place in the workflow.