A customer replies to thank you, and a survey link at the bottom of the email asks how the exchange went. Their click is one answer, and the share of answers that come back positive is the number your report calls CSAT.
CSAT stands for customer satisfaction score. It is the percentage of survey answers that land at the top of the rating scale, and on the common 1 to 5 scale the top is 4 and 5. A score of 80 percent means four in five of the answers were 4s or 5s.
That 80 percent is a share of the people who answered, not of the customers you helped. The question is about one exchange rather than the relationship.
What the score counts
One help desk’s reporting documentation defines the satisfaction score as the percentage of rated surveys that received a 4 or 5, which is the definition to compare your own report against.
The threshold does more work than it looks like. A rating of 3 sits in the denominator and not in the top box, so a month of 3s reads as zero percent. Lifting every rating from a 4 to a 5 does not move the score at all.
| What the customer did | In the score |
|---|---|
| Rated the exchange 4 or 5 | Yes. It counts as a satisfied answer |
| Rated it 1, 2 or 3 | In the denominator, not the top box |
| Left a comment and no rating | No. The score reads ratings |
| Never opened the survey | No. Not in the denominator either |
The same product’s survey documentation records that a comment with no star rating does not reach the satisfaction statistics.
When the survey arrives, and what it asks about
Timing decides the question. A rating link in the footer of a reply asks how that reply landed. A separate email sent after the conversation closes asks how the whole exchange went.
In the first design the link rides along with every reply, and the customer can rate each reply for ten days. In the second the survey waits for the close, goes out on a delay that defaults to two hours, and allows one survey per conversation.
The same page lists the conditions that suppress a survey: a snoozed conversation gets none, and neither does one whose last reply from the team is more than a week old when it closes. The exchange least likely to earn a good rating is the one least likely to be asked about.
Two reports that both say CSAT
The label is not a unit. One score counts answers rated 4 or 5 as a share of all answers. Another report starts from three levels, Great, Okay and Not Good, and subtracts Not Good from Great, which is a score that can fall below zero.
So two dashboards can print a CSAT percentage and mean different measurements. Before you compare your number with a figure from a benchmark page, find the question, the scale, the timing and the count behind it.
Why the response rate is the number to read first
The score is a percentage of answers, and the answers are a fraction of the conversations you had. One report prints the number of ratings next to the share of customers who gave one. Another tool reports a response rate as a metric of its own, beside the score.
Who answers is not a random slice of who you helped. A customer with a strong reaction has a reason to click, and one who got a workable answer usually does not. Silence is not a 5.
Read the count first. Six answers and 40 answers are not the same claim, even when both print the same percentage.
Why one bad week moves a five-person score
Small denominators move a long way on one click.
| Week | Answers | Rated 4 or 5 | Score |
|---|---|---|---|
| 1 | 5 | 5 | 100 percent |
| 2 | 6 | 5 | 83 percent |
| 3 | 4 | 4 | 100 percent |
Nothing about the support changed across those three weeks. Week 2 has one extra answer, one of the six rated the exchange a 1, and the score moved 17 points.
At five people a week is noise, so read the score across months, with the count beside it.
What it does not measure
One platform’s documentation describes a CSAT score as a measure of how a customer feels about a single interaction, and that is its honest limit.
The survey goes out when the conversation closes, often well before the fix ships. The customer rates the reply, the wait and the number of times they had to ask. Whether the problem got solved is the question they remember a month later.
The score cannot see the customer who gave up either: somebody who stopped replying or bought elsewhere is not in the survey.
If the question you care about is the wait, that is a different measurement: what is first response time defines it. Do not read one as a proxy for the other.
What a team of five should do with it
- Read the comments, all of them. The percentage says a week went sideways; the comment says which customer.
- Keep the count beside the percentage. Compare months rather than weeks, and compare them with your own earlier numbers.
- Do not attach the score to a person. One rating covers the wait, the reply and the fix, and the person who wrote the reply set only one of the three.
- Close the loop the survey cannot see. If the fix lands a month after the survey, tell the customer when it ships.
Where it lands in a shared inbox
On a shared address a rating attaches to a conversation rather than a person. What decides whether a customer stays is usually the fix, and the fix lives where the survey never looks.
Chirp is a shared inbox for small software teams that track their work in Linear. Press i in a thread to link
it to an issue, create one, or search for one that exists. Snooze the thread until the issue is done, and it comes
back with a reply drafted for you that names the issue.
The Issues view (g i) sorts linked issues by how many customers are waiting, a count of the people still owed an
answer. No rule in Chirp can send mail. A satisfaction survey comes with the ticketing model, and
shared inbox vs help desk covers which of the two your team needs.
FAQ
What is a good CSAT score? A quoted industry average came from other companies’ customers and their survey question. Compare your own score over time, next to the count of answers it came from.
Why is our response rate low? The survey is a second email that arrives after the problem is over, and nothing obliges anyone to answer it. A strong reaction is a reason to click, and a working answer usually is not.
Should CSAT go in a weekly report? Put the comments and the count in, and leave the percentage for a monthly view. Five answers can swing a week by 20 points on one rating.
Next step
The next question is where those replies live and who owns them. What a shared inbox is covers the category and the arrangements that get mistaken for it.
