Blog

CSAT vs NPS: what each measures and which to use

CSAT asks how happy a customer was with one interaction, like a support ticket. NPS asks how likely they are to recommend you, which is about the whole company. Support teams should measure CSAT.

By Andreas Du Rietz, founder of Replyful ·

How each is calculated

CSAT

How satisfied were you with how we handled your request?

1–3

Not counted

4–5

Satisfied

Satisfied answers ÷ all answers

NPS

How likely are you to recommend us to a friend or colleague?

0–6

Detractors

7–8

Passives

9–10

Promoters

% promoters − % detractors

CSAT counts the happy answers. NPS subtracts the unhappy ones and ignores the middle, so the same customers give very different numbers.

CSAT is a percentage from 0 to 100. NPS runs from −100 to +100.

A worked example

Say 200 customers answer each survey.

  • CSAT. 170 pick 4 or 5. CSAT is 170 ÷ 200 = 85%.
  • NPS. 120 are promoters (60%), 50 passives and 30 detractors (15%). NPS is 60 − 15 = 45.

When to send each

  • CSAT: right after a ticket is resolved. The customer still remembers the conversation, and the score belongs to that ticket and the person who handled it.
  • NPS: every quarter or so, to all customers. It asks about the relationship, not one moment, and nobody wants that question after every ticket.

Which one support should own

CSAT. Your team decides how fast and how well a ticket gets handled, so the score moves when the team does.

NPS moves with price, product bugs and the sales process. Support affects it, but a support team judged on NPS is judged on things it doesn't control. Use NPS for the company and CSAT for support.

The catch with CSAT

Most customers never answer. The ones who do tend to be either delighted or furious, so 85% from 40 answers says less than it seems.

  • Report the response rate next to the score.
  • Read every low score, not only the average. A 1 or 2 usually points to something you can fix.

A third option: let AI score every ticket

An AI can read each closed conversation and estimate how satisfied the customer was, on the same 1–5 scale. Nobody has to answer a survey.

Pros

  • Every ticket gets a score, not just the few customers who reply.
  • No response bias. The quiet, mildly annoyed customer counts too.
  • No survey emails, so customers aren't asked again and again.

Cons

  • It's an estimate, not the customer's own words. Don't present it as CSAT.
  • It only sees the conversation. If the fix stops working the next day, the score won't know.
  • Polite customers can sound happier than they are.

Use both. Real CSAT tells you what customers say. The AI score fills in the tickets they didn't rate, and a big gap between the two is worth a look.

How Replyful does it

Replyful can email the customer a 1–5 rating request once their ticket is resolved, and the answer shows on the ticket. The AI also scores every closed ticket from the conversation, shown as "Estimated satisfaction" on the dashboard and kept apart from real CSAT. Estimated satisfaction is on the Pro and Enterprise plans.

See how your support is really doing.

Not better in 90 days? Your subscription back.