KPIs and metrics

Measuring CSAT After Calls: Questions, Timing and Reading the Results

How to measure customer satisfaction after phone calls: choosing the question, when to ask, handling low response, reading scores fairly and acting on feedback.

Call editorial team · Published · 5 min read

Customer satisfaction, or CSAT, is the simplest way to ask callers how it went. One question, a short scale, and a place to say more. It is also easy to do badly: ask too late, ask too much or read too much into the number. This guide shows how to set up a post-call measure that tells you something true and useful.

A customer finishing a phone call and tapping a rating on a screen
Illustrative image

CSAT sits alongside other measures. Net promoter score asks about loyalty rather than one interaction, and customer effort score asks how hard it was. A KPI list places all of them in context.

What CSAT measures, and what it does not

CSAT measures how satisfied a customer felt about a specific interaction, at that moment. It does not measure loyalty, the quality of your product or the agent's skill on its own. A caller can be happy with a kind agent and still unhappy with the policy the agent had to explain. A good analysis separates those.

Which question for which job

  • CSATHow satisfied were you with this call? Suited to judging single interactions.
  • EffortHow easy was it to get your problem solved? Suited to finding friction.
  • Net promoterHow likely are you to recommend us? Suited to the overall relationship.
  • Open commentWhat could we have done better? Suited to understanding the reason behind any score.

Designing the question

Use one plain question, such as "How satisfied were you with today's call?", and a scale people can answer in a second. A five-point scale from very unsatisfied to very satisfied is common. Whatever you choose, keep it fixed so results are comparable over time.

When and how to ask

Ask soon after the call, while the memory is fresh, and tie the question to a clear moment: when the call ends, or when a case is resolved. Asking days later gives vague answers and a lower response. Choose the channel by what suits your callers: a keypad rating at the end of the call, a short spoken question, an email link or a message. Each has its own response pattern and bias, so stay consistent.

A simple post-call survey flow

  1. Call endsThe conversation finishes and the agent's work is done.
  2. QuestionThe caller is asked one short satisfaction question.
  3. Optional commentThe caller may leave a few words or a voice note.
  4. ThanksA brief acknowledgement, with no further questions.
  5. ReviewScores and comments appear next to the call they belong to.

Decide whether every call is asked or only a share. Asking everyone gives more data but can irritate frequent callers. Asking a sample is gentler and often enough for a small team. Never ask callers in the middle of a sensitive situation, such as a complaint they are still raising.

Response rates and bias

Only some callers answer, and those who do are not typical. People with strong feelings, good or bad, respond more often than people with neutral experiences. Short calls may be rated differently from long ones. Callers who reached an AI agent may respond differently from callers who spoke to a person.

A small number of responses can swing an average widely. Treat any score built on a handful of ratings as a prompt to listen to the calls, not as a verdict.

Reading results fairly

Two ways to read the same score

Shallow readingUseful reading
Average onlyTeam scored lower this week, so coach themLook at response count, call types and comments first
Individual scoresRank agents by ratingReview a few calls per agent and coach on specifics
Low scoresTreat as the agent's faultCheck whether the cause was policy, wait time or product
High scoresAssume all is wellCheck comments for friction the score hides

Always open the calls behind the unhappy scores. A rating alone cannot tell you whether the problem was a long wait, a missing answer, a rude moment or a rule the agent could not bend. Listening takes minutes and usually reveals a pattern. If three low-scored calls in a row were about the same delivery problem, the fix lies outside the phone team.

Using scores fairly with staff

Ratings are noisy for individuals. A person who takes hard calls will get lower scores than someone who takes easy ones. Use results to start a conversation, not to end one. Share them openly, explain what they measure and combine them with call reviews. Quality management for calls and a coaching framework show how.

Turning feedback into action

A survey that changes nothing teaches callers that answering is pointless. Close the loop in two ways. Internally, review comments weekly, tag them by theme and choose one or two fixes to make. Externally, when a caller leaves a low score and asks for contact, follow up with a callback; this is the quickest way to recover a poor experience.

Weekly CSAT routine

  • Check the count of responses next to each score
  • Read all comments from low scores
  • Listen to a few of the calls behind them
  • Tag the cause: wait, answer, tone, policy or technical
  • Choose one fix and one owner
  • Follow up with callers who asked for contact
  • Share a short summary with the team

Common pitfalls

One more pitfall is chasing the score itself. When agents feel measured on it, they may ask for good ratings or avoid difficult callers. Focus on the behaviours that tend to make people satisfied: being understood, getting an answer and not having to repeat themselves. First call resolution is a close companion to satisfaction, and empathy on calls covers the human side.

Doing it in Call

In Call, the review area under AI agents shows what customers said over the last thirty days: a satisfaction score out of five, a net promoter score, how many replied and how many left comments. Unhappy answers link to the call, so you can play the recording and see whether a reviewer scored it. You can also create a survey by naming it, choosing when it is asked, either when every call ends or when a case is resolved, and setting what share of those calls receive it. Results can be broken down by agent. Call does not text callers by itself, so choose the channel you use for the survey with that in mind.

Frequently asked questions

What is a good CSAT score?

There is no universal target. Compare against your own history and your own call types, and watch the response count. A score from a few ratings means little.

Should I ask after every call?

Asking every call gives more data but can annoy frequent callers. A sample is gentler and is often enough. Whatever you choose, keep it consistent.

How is CSAT different from NPS?

CSAT rates one interaction. Net promoter score asks about overall loyalty. Use CSAT to judge calls and NPS to judge the relationship.

Why are my response numbers low?

Surveys that are long, late or hard to answer attract few responses. Shorten the question, ask sooner and make answering take seconds.

Can I use CSAT to rank agents?

Be careful. Individual scores are noisy and depend on the calls people take. Use them to start conversations and combine them with call reviews.

What should I do with a low score?

Open the call, find the cause and, if the caller asked for contact, call back. Then record the theme so repeated causes are fixed.

Create an after-call survey in Call

In the demo console you can try incoming calls in a simulation.

Related guides