Image URL
Do not index
Do not index
AI summary
NPS, CSAT, and CES answer different questions: NPS measures relationship health and advocacy, CSAT evaluates a specific experience, and CES identifies how difficult a task was. Choose the metric based on the decision it will inform, act on the feedback, and skip surveys if the results would not lead to changes.
Title
NPS vs CSAT vs CES: Which One Answers Your Question?
Date
Description
NPS vs CSAT vs CES: when to use each feedback metric, what each hides, and the honest case for measuring none yet. Includes a comparison table.
Status
Current Column
Person
Writer
Your board wants a loyalty number.
Support wants to know if the last ticket felt easy.
Product wants a pulse after onboarding.
Those are three different decisions. NPS, CSAT, and CES are three different tools. Most comparison posts list definitions and leave you to sort it out. This one starts from the decision you are trying to make — and includes the honest case for measuring none of them yet.
Short answer: use NPS when you want relationship health and people who might vouch for you; CSAT when you need a pulse on one recent touchpoint; CES when you need to know how hard a task felt. If you would not change anything based on the number, skip the survey.
For the programme behind the score — what to do with a 9 or a 3 — see our guide to NPS surveys. For the arithmetic, see how to calculate NPS.
One-page cheat sheet
Same dimensions for each metric. Sources dated November 2026.
Dimension | NPS | CSAT | CES |
Question it answers | Would they recommend you? | Were they satisfied with this experience? | How hard was it to get what they needed? |
Typical scale | 0–10 likelihood to recommend | 1–5 or 1–7 satisfaction | 1–5 or 1–7 effort / ease |
Best used for | Relationship health + finding advocates | A single touchpoint (support, onboarding) | Support & self-serve friction after a task |
What it hides | Why they scored that way | Loyalty — satisfied is not a promoter | Overall relationship — only that interaction |
Choose it when… | You want promoters you can ask for proof | You need a pulse on one recent moment | Tickets feel hard and you need the friction |
NPS definitions follow Bain's Net Promoter System measurement page (as of November 2026): promoters 9–10, passives 7–8, detractors 0–6; score = % promoters − % detractors. CES entered the mainstream via Dixon, Freeman and Toman in Harvard Business Review, July–August 2010 — effort after a service interaction. CSAT is the older industry satisfaction question after a touchpoint; there is no single trademarked formula the way NPS has Bain's.
Choose NPS if…
You care about the relationship, not one ticket.
NPS asks how likely someone is to recommend you to a friend or colleague. That is a forward-looking loyalty signal. It is also the only one of the three that naturally splits people into buckets you can act on: promoters you can invite to leave a testimonial, passives you can leave alone for now, detractors you can route to a human.
I'd reach for NPS when:
- You want a quarterly (or lighter) relationship check, not a post-ticket pulse
- You plan to do something with promoters — ask for a testimonial, invite a case study, or at least thank them
- Leadership wants one number they can trend over time
Skip NPS when the only ask is "was this chat helpful?" That is CSAT's job, and stuffing it into a 0–10 recommend question muddies both.
The score alone is thin. How to calculate NPS gets the formula out of the way; the useful part is the bucket mix and the free-text follow-up. If you will never open the comments, pick a lighter metric.
Choose CSAT if…
You need a verdict on one moment.
CSAT asks how satisfied someone was with a specific experience — a support reply, an onboarding call, a checkout flow. It is local. It moves when that touchpoint changes. It does not tell you whether they would still recommend you next quarter.
I'd reach for CSAT when:
- Support or success owns the question and wants a score tied to a ticket or session
- You are testing a change to one workflow and need a before/after pulse
- You already run NPS for the relationship and need something shorter in between
Skip CSAT when you are trying to find advocates. A five-star support rating is not a promoter. Plenty of people are "satisfied" and still would not recommend you — they just got what they needed and moved on.
CSAT is also easy to game with timing. Ask immediately after a resolved ticket and you mostly measure relief. That is fine if relief is the job. It is a bad proxy for product love.
Choose CES if…
The work feels hard and you need to know where.
CES asks how much effort the customer had to put in — or how easy the company made it — to handle a request. The 2010 HBR piece argued that reducing effort in service interactions predicts loyalty better than trying to "delight" people after every contact. Gartner (which acquired CEB, where that research sat) still frames CES around low-effort service.
I'd reach for CES when:
- Support volume is high and you suspect handoffs, repeats, or self-serve dead ends
- You just shipped a help centre or in-app flow and need to know if it actually reduced effort
- The complaint you keep hearing is "it took forever," not "I don't like the product"
Skip CES when you want relationship health or advocates. Effort on one ticket says nothing about whether they would introduce you to a peer. Also skip it if you will not fix the friction you find — a high-effort score with no backlog item attached is theatre.
CES is the one I would recommend even when Testimonial is not in the picture. We do not ship a CES product. If effort is your only question, use your support tool's survey or a simple form after the ticket closes. That honesty is the point of a comparison.
What each one hides
Every metric lies about something. Name it before you pick.
NPS hides the reason. A 6 and a 9 can look the same in a dashboard until you read the comment. Without a follow-up question, you have a trend line and no diagnosis. The score also compresses very different relationships into one number — the calculation post covers why bucket mix matters more than the headline.
CSAT hides loyalty. Satisfied today ≠ staying next year. It also hides comparison: people rate against the last thing they remember, not against your competitors.
CES hides the relationship. A smooth refund does not mean they love the product. A hard onboarding might still lead to a promoter six months later if the product pays off. CES is a microscope, not a map.
If you only have room for one follow-up habit, put it on NPS: ask why, then route the bucket. That is the path from score to proof.
The honest case for measuring none of them yet
You do not need a metric programme to learn from customers.
Skip all three when:
- You would not change anything based on the number. If the survey lands in a folder nobody opens, the ask is spam. Talk to five customers instead.
- You have fewer than a few dozen active customers. Sample sizes that small make trend lines noisy. A monthly call or a short email beats a dashboard.
- You have not fixed the last round of feedback. Sending another survey while known issues sit open trains people to ignore you.
- You only want social proof. Then ask for a testimonial directly. Do not run NPS as a roundabout way to collect quotes — unless you will actually route promoters, which is a real workflow (see below).
"Measure nothing" is a temporary state. The moment you have a decision a number would change — ship the help article, hire another support person, ask promoters for proof — pick the metric that answers that decision. Not all three at once.
Doing this with Testimonial
Testimonial ships NPS, not CSAT or CES. That is a real limit. If your only need is a post-ticket satisfaction or effort score, use your support tool. We are not the right product for that job.
What we do ship is NPS with score-bucket routing on the Ultimate + plan ($95/month per space as of November 2026 — see pricing). One 0–10 question, adjustable thresholds, and rules for what happens to promoters, passives and detractors. Promoters can get a testimonial request or a case study invite automatically. Detractors can alert support. The feature page is NPS with auto-routing.
That is the reason to pick NPS in our stack: the score is the start of a proof workflow, not the end of a report. How to turn NPS promoters into testimonials walks through the promoter rule. How to create an NPS survey is the twenty-minute setup.
Where ours stops
- No CSAT or CES. If those are the questions you need, we will not stretch NPS to fake them. Use the right tool.
- Ultimate + only (per space). Below that there is a one-time NPS preview for the account — enough to try routing, not a substitute for a full programme. Confirm current limits on pricing.
- The number still needs a human. Routing sends the next email. It does not fix the product issue a detractor named, and it does not write the testimonial for a promoter.
Can you run more than one?
Yes — in sequence, not as a mash-up.
A common pattern that works:
- CES or CSAT on the ticket. Owned by support. Closes the loop on that interaction.
- NPS on a slower cadence. Owned by product or CX. Looks at the relationship.
Do not ask all three in the same email. You will train people to skim, and you will not know which question drove the reply. If you already send a post-ticket CSAT, do not also drop an NPS link in the same thread.
When the metrics disagree — high CSAT, soft NPS — believe both. The ticket went fine; the relationship is thin. That is useful. It is not a reason to average them into one fake score.
A simple way to decide this week
Ask one sentence in standup:
"What would we do differently if this number moved?"
- If the answer is "ask happy customers for a public quote" → NPS
- If the answer is "change how we close tickets" → CSAT or CES, depending on whether you care about satisfaction or effort
- If the answer is "nothing, we just want a slide" → measure none until the answer changes
Write the answer down before you pick the tool. Then run one metric well. Add a second only when the first has a habit behind it — comments read, buckets routed, fixes shipped.
If you pick NPS and you use Testimonial, the next concrete step is creating the survey and ticking the promoter rule. Everything else in this post was the decision. The setup is short on purpose.
Wrapping up
NPS, CSAT and CES are not rivals for the same job. They answer different questions. Pick by the decision you will actually make — and if you would not make one, skip the survey until you would.
If NPS is the fit and you want the score to feed proof, set promoter routing once and let every 9 and 10 get the ask on the day they answer.
Written by
