Zendesk Customer Satisfaction

The Zendesk customer satisfaction score is the least useful thing the survey produces. The comments are where the money is, and most teams never read them systematically.

Start free trial All guides

How Zendesk customer satisfaction ratings work

When you enable satisfaction ratings, Zendesk emails the requester after a ticket is solved and asks whether the support they received was good or bad. There is a free text box for a comment. Both the rating and the comment land on the ticket, and both are reportable in Explore.

Your satisfaction score is the share of good ratings out of all ratings received over a rolling window. Nothing more sophisticated than that. Some plans also offer an expanded scale and additional survey options, so check what your tier exposes before you design a programme around a specific format.

One setting most teams should change: satisfaction reasons. Turn it on and a customer who rates a ticket bad is asked to pick why from a list you configure. That single dropdown converts an unusable number into a diagnosis.

Why the headline score misleads

Three reasons, and they compound.

Response rates are low and biased. Typically only a small minority of customers respond, and the ones who do are disproportionately delighted or furious. The quietly satisfied majority never clicks. So your score measures the extremes, not the average experience.

It's dominated by outcome, not by service. A customer told no politely and promptly still rates it bad. A customer told yes by a rude agent rates it good. Managing agents on a number that mostly reflects your refund policy is unfair and it corrodes trust fast.

Small numbers move it wildly. At thirty responses a month, four bad ratings swing your score by more than ten points. Teams then hold meetings about a change that is statistical noise, which is how CSAT programmes lose credibility with the people they measure.

Track it as a trend across a quarter, compare it against itself rather than against an industry benchmark somebody quoted, and never put it in a bonus formula on its own. The CSAT calculator will show you how much a handful of responses moves the number.

Reasons are where diagnosis starts

Configure a short list. Five to seven options, phrased the way a customer thinks rather than the way your org chart does.

Took too long to get a response.

The issue was not resolved.

I had to repeat myself.

The agent wasn't helpful.

The product doesn't do what I need.

Now the score becomes actionable. Bad ratings concentrated in "took too long" is a staffing and routing problem you fix with rotas. Concentrated in "had to repeat myself" it is a handoff and context problem, often duplicate threads or thin internal notes. Concentrated in "product does not do what I need" it's not a support problem at all, and the right response is to take that data to product rather than to coach the agent.

That last distinction is the whole value of reasons. Without them, every bad rating looks like an agent failure, and support teams get blamed for the roadmap.

A weekly routine that works

Reading every comment doesn't scale, and reading none is what actually happens. Here is a middle path that survives contact with a busy week.

Once a week, read every bad rating comment. There will be fewer than you fear. Tag each one with a theme, informally is fine. Then read ten random good ones, because the praise tells you which behaviours to make standard and nobody ever looks at them.

Follow up on the bad ones within a day. Not to argue the rating, and not to ask them to change it. Just: we saw this, here is what happened, here is what we are doing. A meaningful share of customers who rated a ticket bad will engage with that, and some of them become your most loyal accounts precisely because you came back.

Then, once a month, report the themes rather than the score. "Nine of fourteen bad ratings this month were response time on Monday mornings" is a sentence somebody can act on. "CSAT was 87%" is not.

What to do with a bad rating

Reopen the ticket if the issue is genuinely unresolved. A bad rating on a solved ticket often means it was not solved, it was closed, and the customer had nowhere else to go.

Do not make ratings a public leaderboard. It produces two behaviours you do not want: agents avoiding hard tickets, and agents asking customers for good ratings. Both make the data worse and the queue meaner.

And watch the tickets that get a bad rating and then a new ticket from the same person within a day. That pattern is a resolution that did not hold, and it is one of the more reliable early warnings that a category of problem is being closed rather than fixed.

FAQ

Frequently asked questions

How does the Zendesk customer satisfaction survey work?

The Zendesk CSAT survey fires after a ticket is solved and asks one question. The Zendesk satisfaction rating is the good-versus-bad split, and Zendesk satisfaction reasons let a customer pick why, which is far more actionable than the score.

How does Zendesk calculate the customer satisfaction score?

Good ratings as a share of all ratings received over a rolling window. Unanswered surveys aren't counted, which is why response rate matters.

When is the satisfaction survey sent?

After the ticket is solved, to the requester, with timing you can adjust through the automation that sends it.

What are satisfaction reasons?

A configurable dropdown shown to customers who leave a bad rating, asking why. It converts an unusable number into something you can act on.

What is a good CSAT score?

Compare against your own trend rather than a benchmark. Response rate, ticket mix and how you word the survey affect the number more than service quality does.

Should agent bonuses be tied to CSAT?

Not on its own. The score reflects outcomes as much as service, and small sample sizes swing it enough to be genuinely unfair.