Customer satisfaction score for support

Customer satisfaction score (CSAT) is a direct question about a support experience, usually sent after a conversation or ticket reaches an outcome. It can reveal poor explanations and unresolved problems that a speed metric misses. It cannot tell you what every customer thinks: only some people receive, open and answer a survey.

Freshdesk describes its CSAT rating as the percentage of respondents choosing a positive answer. Zendesk's Support metric definition likewise divides good-rated tickets by all rated tickets. Other survey products may normalise a multi-point score differently, so document the question and scale before comparing numbers.

Calculate the positive-response rate

CSAT = positive rated responses ÷ all rated responses × 100. If 80 customers responded and 64 selected a positive choice, CSAT is 80%. Do not divide by every ticket or every survey sent unless you explicitly mean to calculate a different measure. If your survey has five points, define whether the top two count as positive and whether the middle option stays in the denominator. Keep the rule unchanged when comparing periods.

Beside that 80%, show “64 positive of 80 ratings” and the survey response rate. If 800 surveys were sent, only one in ten recipients answered. An apparent two-point change may be noise or a change in who replied. Small samples need especially cautious interpretation; do not turn one bad survey into an agent performance verdict.

Choose timing and coverage deliberately

Send the survey after a meaningful resolution, not merely after an automatic closure. If the workflow surveys only email but many complaints arrive by chat, the score is not the whole support experience. Check whether a customer can answer after a reopen and whether duplicate surveys are sent for one issue. Keep the question short, accessible and specific: satisfaction with the support received, not a broad question that could be about product price or delivery policy.

Freshdesk's new CSAT reporting documentation includes positive, neutral and negative distributions plus surveys sent and responses received; availability depends on account configuration. Use the distribution rather than hiding neutral feedback inside a single number.

Turn feedback into an improvement

Review low ratings with the ticket and reason category. Was the answer wrong, the process slow, the policy frustrating, or the customer's desired outcome impossible? A negative score attributed to one agent may reflect a refund rule outside their control. Record a root-cause tag and an action owner, then check whether similar cases improve. Close the loop with the customer when appropriate, without pressuring them to change a rating.

Compare CSAT with resolution time, reopens and the weekly report. A quicker queue with worse feedback might mean cases are being rushed; a lower score during a product outage may call for better incident communication rather than faster typing.

Avoid false comparisons

Do not publish an unsourced universal “good CSAT” benchmark. Survey question, scale, channel, sector and respondent mix all change the number. Your own stable trend by issue type is more useful. Keep comments and survey response rate visible alongside the percentage. If you change the rating scale or sending rule, mark a new baseline rather than presenting the next result as like-for-like growth.

Consider a small practical example. Suppose a team improves its score from 78% to 85%, but only five customers answered the later survey and the response rate fell sharply. That is a reason to investigate, not announce a decisive win. Check which channels received the survey, which issue types were represented and whether unresolved customers were excluded by the sending rule. A short read of the five conversations may reveal more than the apparent seven-point rise.

Sources

• Freshdesk: setting up customer satisfaction surveys
• Zendesk: Support CSAT metric definitions
• Freshdesk: new CSAT reporting

Survey design and improvement recommendations are Support Sorted's editorial judgement.