CSAT survey examples: what to ask, when to ask it, and why length is the thing that breaks

A CSAT survey asks a customer to rate satisfaction with one specific thing, then gives them room to explain the rating. Every example below follows that pattern. The rating question changes based on what you are measuring, and the follow-up question stays open-ended.

This is the original use for CSAT, and the most common one. The survey goes out immediately after a ticket closes, while the interaction is still fresh.

The value here is attribution. Because the survey is tied to a single closed ticket, a low score can be traced to a specific agent, queue, or issue type rather than to support in general.

Onboarding satisfaction is worth measuring separately because early friction is disproportionately predictive. A customer who struggles in week one carries that impression into every later interaction.

The follow-up wording matters. Asking which part was hardest assumes something was, which gives permission to answer honestly and produces more specific comments than a neutral prompt.

Post-purchase surveys measure the fulfilment experience rather than the product. Sending too early captures excitement instead of reality.

Release surveys tell you whether a change landed. They work best triggered in-app once someone has actually used the new thing, rather than emailed to everyone on announcement day.

Sometimes you need a read on the whole relationship rather than one moment. This survey runs on a schedule and is not triggered by an event.

The customer satisfaction survey template is built around this last pattern, combining a satisfaction scale, product perception attributes, and a likelihood-to-return question.

Every rating question above asks about exactly one thing. That constraint is the single most important design rule in a CSAT survey, and the easiest to break by accident.

Consider a question like "Did you enjoy our service and new menu?" A customer who loved the service and disliked the food has no honest way to answer. Whatever they choose, you cannot tell which half of the question they were rating. Splitting compound questions into two is almost always the fix.

If your answer options do not include the response a customer wants to give, they will answer inaccurately, skip the question, or abandon the survey. None of those outcomes produces usable data.

The practical safeguards are straightforward:

Satisfaction scales can be numbers, stars, or emoji, and a 1 to 5 or 1 to 7 range are both defensible. What matters is not switching between them mid-survey. Mixed scales make a survey harder to complete and the resulting data harder to compare. Our explanation of Likert scales covers how to build balanced options.

Every example on this page is two or three questions. That is deliberate, and Greyhound demonstrates why better than any hypothetical.

Greyhound's original post-trip survey ran 57 questions and took an hour to finish. Its completion rate sat below 18%, and the reports it produced were long enough that station managers stopped reading them. The survey was eventually shelved because it was not delivering anything usable.

The rebuild was a five-question survey, including one Net Promoter Score question and one open-ended question. Completion climbed to 94%, Net Promoter Score rose by nearly 15 points within a few months, and the time station managers spent reviewing comments dropped from three hours a week to three minutes.

The company later traced a drop in scores at one station to restroom complaints, then found through open-ended feedback that the real cause was the women's restroom being closed for cleaning during the busiest part of the day.

That last detail is the argument for the open-ended follow-up. The rating told them something was wrong at that station. Only the comments explained what.

A rating on its own gives you a trend line with no cause. The open-ended follow-up is what makes a score diagnosable, and one is usually enough. Making a text field mandatory tends to backfire, producing filler from customers who had nothing to add and irritating the ones who were already unhappy.

At volume, reading comments individually stops being realistic. Ryanair collects roughly 500,000 CSAT survey responses a month across booking, check-in, the in-flight experience, and customer service interactions. Grouping that many comments into themes is what makes them usable, which is what text analysis is for.

Individual surveys are easy. Running several without producing contradictory data takes a little structure.

Sort your surveys along two axes:

Then apply three filtering rules. Trigger each transactional survey from the event itself rather than a calendar, so timing stays consistent. Keep the rating question identical across time for any given touchpoint, because changing the wording resets your trend. And do not send a customer two surveys in the same week, since survey fatigue depresses response rates and skews who answers.

The customer satisfaction survey template library covers narrower cases if your touchpoints do not map cleanly onto the five above.

Once your surveys are running, these cover what comes next:

The examples on this page work because they ask one clear question, cover every answer a customer might want to give, and stay short enough that people finish them. Greyhound went from 18% completion to 94% by cutting 52 questions, which is a bigger gain than any clever wording would have produced.

Every example here can be built in minutes using expert-written templates and scales. Use these examples in a survey and start collecting satisfaction data your team can actually act on.

NPS, Net Promoter & Net Promoter Score are registered trademarks of Satmetrix Systems, Inc., Bain & Company and Fred Reichheld.