Zendesk CSAT, Measured Honestly
One question after a ticket closes, and a number that either tells you something or flatters you. Which one depends on how you set it up.
How the survey works
Zendesk sends a satisfaction survey after a ticket is solved, asking whether the experience was good or bad, with an optional comment. The rating comes back on the ticket, so you can report by agent, group, channel, tag and time.
The survey is triggered by an automation, which means you control the timing. That single setting affects your score more than anything your agents do.
Timing, and why it decides your score
Send too fast and you catch people who have not tried the fix yet. Send too slowly and you catch people who have forgotten the interaction and are rating their mood.
A few hours after solve is the usual sweet spot for straightforward tickets, and next morning works better for anything the customer needs to verify.
Whatever you choose, keep it consistent. A score that moves because you changed the delay is a score that measures your automation, not your support.
Reading the number
Most teams land between 75% and 90%, and comparing across companies is close to meaningless because survey design moves the number more than performance does.
What matters is the comment field. Ten bad ratings with comments teach you more than a thousand good ones without. Read them weekly, tag the themes, and take the top theme to whoever owns the underlying problem.
Segment before you conclude anything: by channel, by group, by first response time band. An aggregate CSAT hides the one queue that is failing.
The pattern that drags CSAT down invisibly
A customer writes in twice about one problem. Two tickets, two agents, two answers, sometimes contradictory. Both tickets get solved. Both send a survey.
So one confused customer produces two survey invitations, and any dissatisfaction gets counted twice. Worse, the interaction that caused it was avoidable: the second ticket should never have reached an agent.
If your CSAT has an unexplained soft patch, check your duplicate rate before you coach anybody. Two answers to one question is a quality failure that no amount of agent training prevents.
Frequently asked questions
When should Zendesk send the CSAT survey?+
A few hours after solve for simple tickets, the next morning for anything the customer has to verify. Consistency matters more than the exact delay.
What is a good Zendesk CSAT score?+
Most teams sit between 75% and 90%. Track your own trend rather than benchmarking against a number from a different survey design.
Should merged tickets send a survey?+
No. Exclude tickets tagged `closed_by_merge` from your survey automation, otherwise you are surveying people about a ticket that was administratively closed.
Remove the interactions that earn bad scores
Two agents answering one customer differently is a CSAT problem you can delete rather than coach.
Start free trial14-day free trial. No credit card required.