CSAT vs CES vs NPS: which survey to send, and when
What CSAT, CES and NPS each measure, which moment fits each survey, how to read them together, and a worked example where CSAT hides a problem.
By the Paperlink AI teamPublished 6 min read
Short answer
CSAT tells you how one contact felt, CES how much work it took, and NPS how customers feel about the whole relationship. Send CSAT or CES right after a contact and NPS on a schedule, with one survey per contact and a cap per customer. Read them together, because two channels with the same CSAT can have very different effort.
CSAT, CES and NPS are three one-question surveys, and teams often send all three without a plan. Customers then get a survey after every ticket and another one each quarter, answer fewer of them, and the scores tell you less. Each survey answers a different question. This guide shows which one to use at which moment, how to read them together, and why two channels with the same CSAT can be very different to deal with.
Three surveys, three questions
| Survey | The question | What it measures | Usual moment |
|---|---|---|---|
| CSAT | How satisfied were you with the support you got? | How one contact felt | Right after a ticket or chat closes |
| CES | How easy was it to get your issue solved? | How much work one contact took | Right after a ticket or chat closes |
| NPS | How likely are you to recommend us? | The whole relationship | On a schedule, or after a milestone such as renewal |
CSAT and CES are about one contact. A low score points to something your support process can fix. NPS is about the relationship, so a low score can come from price, the product or delivery as much as from support. That is the most useful rule of the three: do not judge a support team on NPS alone, and do not judge the business on CSAT alone.
How to choose
Use CSAT when you want to know how a contact felt
CSAT is the easiest to explain ("three in four customers were satisfied") and the most common in help desks, so it is a good default for a team that has no survey yet. It is sensitive to tone and speed, which makes it useful for coaching. It is less useful for finding process problems, because a friendly agent can earn a high score on a contact that took the customer three emails.
Use CES when you want to find friction
CES asks about work, not feelings. It drops when customers repeat themselves, are transferred, wait for a simple answer or cannot find a help article. Choose it when you are changing a process, such as a new handoff or a new help center, and want to know whether it got easier.
Use NPS when you want the wider picture
NPS belongs to the whole company. Send it on a schedule, not after support contacts, and share the comments with product and sales as well as support. The NPS calculator page covers the scale and the formula.
Same CSAT, different effort: a worked example
One invented team answers both chat and email. In the same month, it sent CSAT and CES after every closed conversation. The numbers below are computed with the same code as our CSAT calculator:
| Channel | CSAT answers | CSAT | CES answers | CES (1 to 7) | Difficult (1 to 3) |
|---|---|---|---|---|---|
| Chat | 100 | 75% (±8.5) | 100 | 5.53 (±0.29) | 10% |
| 120 | 75% (±7.7) | 120 | 4.63 (±0.33) | 30% |
On CSAT alone, the two channels look the same: 75% in both. On effort, they are not. Email scores 4.63 against 5.53 for chat, a gap of 0.9 points, which is much larger than the margin of error of either channel. 30% of email customers called the contact difficult, against 10% on chat.
The likely story is that email agents are polite and helpful, so customers are satisfied with them, but the email process makes customers work: long waits between replies, or a second message to give details the first reply asked for. CSAT would never show it. The next step is to read the comments and the tickets behind the low effort scores, and to count how many email conversations needed more than one reply from the customer.
Reading the scores together
| Pattern | What it often means | Where to look |
|---|---|---|
| High CSAT, poor CES | Kind agents, hard process | Handoffs, wait times, missing help articles |
| Poor CSAT, good CES | Fast answers that felt cold or unwelcome | Tone and policy wording in replies |
| Good CSAT and CES, poor NPS | Support works; something else does not | Comments about price, product or delivery |
| Poor on all three | A problem customers meet before they contact you | The most common ticket tags this month |
These are starting points, not rules. The comments tell you which reading is right for your customers.
How often to ask
- One survey per contact. If you want both CSAT and CES, send one of them, or both as two questions in the same survey, not two separate emails.
- Cap surveys per customer. A customer who opens five tickets a week does not need five surveys. Set a limit, such as one survey per customer each week, in your help desk.
- Keep NPS apart from support contacts. An NPS question that arrives right after a ticket measures the ticket, not the relationship.
- Always add one open question. "What is the main reason for your score?" gives the comments that explain every pattern in the table above.
- Keep the moment fixed. If you change when you send a survey, start a new trend line, because the old and new scores are no longer comparable.
If you can only run one
For a small support team that runs no surveys today, start with CSAT after each closed ticket, plus the open question. It is the simplest to set up and explain. Add CES when you start to change your process and want to measure the change. Leave NPS to whoever owns the customer relationship, and ask for the comments that mention support.
Whatever you choose, check the margin of error before you compare two months. With a few dozen answers, a change of several points can be noise. The CSAT and CES calculator shows the margin for your own counts.
Where a support chatbot fits
Some of the contacts behind poor effort scores are simple questions whose answer is already on your website. Paperlink AI is a customer support chatbot that answers only from the documents you upload and the website pages you import, and says so when your content does not cover a question. Visitors can rate each answer, and an answer they mark not helpful appears in Needs review in your dashboard. Those ratings are about the bot, not your team, so keep them out of your CSAT and read them as a separate signal.