# CSAT vs NPS: How to Measure and Improve Customer Satisfaction > CSAT vs NPS explained: what each score measures, the exact formulas with a worked example, when to send each survey, and how to improve CSAT for real. _Source: https://professionalstoolkit.com/articles/csat-vs-nps — The Professional's Toolkit · updated November 2026_ --- CSAT (customer satisfaction score) asks how happy a customer was with one interaction, right after it happened, and reports the share of people who answered 4 or 5 on a five-point scale. NPS (net promoter score) asks how likely they are to recommend you, on a zero-to-ten scale, and subtracts detractors from promoters. The short method: **send CSAT after every resolved support conversation, send NPS a few times a year**, and read the comments behind each low score before you read the number. **9 or 10** — the only NPS answers that count as promoters, per Bain - **4 or 5** — the ratings that count as satisfied on a five-point CSAT scale - **0 to 6** — the NPS answers that count as detractors - **-100 to 100** — the range an NPS can take ## What a CSAT score measures, and what NPS measures A CSAT score is a reading on one moment. The customer just got an answer about a late order, a refund or a login problem, and you ask a single question: how satisfied were you with this? Because it sits so close to the conversation, it tells you about the agent, the reply and the speed. It says little about whether that person will buy again next month. NPS works at a different height. Bain, which created it, frames it as one question: how likely are you to recommend us to a friend or colleague, scored zero to ten [1]. The answer reflects the product, the price, the shipping and every past support contact together. That makes it a loyalty signal for the whole relationship. So the two do not compete. One grades the support desk this week. The other grades the business this quarter. ## CSAT vs NPS: which question to ask, and when The practical difference is timing and ownership. CSAT belongs to the support team, because the team can move it within days by changing how it answers. NPS belongs to the founder or the leadership group, because a bad score usually traces to the product, the delivery promise or the price, not to one email. | | CSAT | NPS | |---|---|---| | Question | How satisfied were you with this conversation? | How likely are you to recommend us? | | Scale | 1 to 5 (or Good / Okay / Bad, where only Good counts as satisfied) | 0 to 10 | | When to send | Right after a ticket is solved or a chat ends | A few times a year, or after a milestone | | Who acts on it | Support lead and agents | Founder, product, operations | A small online shop answering customers by email usually starts with CSAT alone. It is one click for the customer, and every answer maps to a ticket and an agent. **NPS earns its place once you have repeat customers** and want to know whether they would bring friends. A SaaS startup with monthly subscribers often needs both early, since subscribers who leave quietly never fill in a CSAT survey. Asking both in the same message is a common shortcut. Avoid it. A customer who just got a fast refund may rate the conversation a 5 and still give the company a 6, and the two answers then blur in the reader's mind. > 💡 **Ask one question at a time:** Send CSAT after the conversation and NPS on its own schedule. Mixing them in one message blurs a happy-with-the-agent answer with an unsure-about-the-company one. ## Net promoter score calculation and the CSAT formula, worked through Both formulas are definitions, not opinions, so get them exactly right. ### The NPS formula Bain sorts answers into three groups. Promoters scored 9 or 10. Passives scored 7 or 8. Detractors scored 0 to 6. Your score is the percentage of promoters minus the percentage of detractors [1]. Passives never enter the subtraction, but they still count in the total, so they pull both percentages down. The result is a number, not a percentage, and it runs from minus 100 to plus 100. ### The CSAT formula SurveyMonkey's CSAT calculator page, a survey vendor's own definition, divides the positive responses, usually the top two boxes, by all responses and multiplies by 100 [2]. On a five-point scale that means 4s and 5s. A neutral 3 counts against you. ### One worked example Here is an illustration with round, invented inputs. A shop handles 400 support conversations in a month and 120 customers answer the CSAT question. Sixty give a 5, thirty give a 4, twelve give a 3, eight give a 2 and ten give a 1. Satisfied answers are 60 plus 30, so 90. Divided by 120 and multiplied by 100, the CSAT is 75%. The same shop runs a quarterly NPS survey and gets 200 answers. Ninety are promoters (45%), sixty are passives (30%) and fifty are detractors (25%). The NPS is 45 minus 25, which is 20. Notice what the averages would have said. The CSAT answers above average 4.0 out of 5, which sounds strong. The top-two-box share of 75% tells a sturdier story: one customer in four was not happy. Pick one method, write it down, and never switch midyear. > ⚠ **Averages hide unhappy customers:** An average of 4.0 out of 5 can hide one unhappy customer in four. Report the top-two-box share and the number of responses, and keep the method fixed. ## How to improve CSAT once the number is in Raising the score is about closing loops, not about chasing the number. Work in this order. 1. **Read every low rating within a day.** A 1 or a 2 usually comes with a comment or an obvious reason in the thread. Read the thread before you judge the agent. 2. **Reopen and reply.** A short follow-up that fixes the actual problem often turns an unhappy customer into one who stays. Many help desks can reopen a ticket automatically when a bad rating arrives. 3. **Tag the reason.** Late delivery, wrong answer, slow reply, policy the customer disliked. After a month the tags show where the score really leaks. 4. **Fix the top cause, not the symptom.** If a third of the low scores mention waiting, the fix is staffing or first-reply speed, not friendlier wording. 5. **Coach with examples.** Share one great and one poor conversation per week with the team. Scores move faster when people see what good looks like. Speed matters, but resolution matters more. A fast reply that does not solve anything still earns a low score. A customer whose problem is fixed on the second reply is usually satisfied. > 💡 **Reopen every bad rating:** A reply that fixes the real problem after a low score often keeps the customer. Make reopening low-rated conversations a daily habit, not a monthly review. Separate what support controls from what it does not. A rating that blames a refund policy belongs in a report to the owner, not in the agent's review. If you mix them, agents learn that the score is unfair and stop trusting it. ## Mistakes that make both scores lie Most bad CSAT and NPS programs fail in the same few ways. - **Counting averages one month and top-two-box the next.** The trend line then shows a change that never happened. - **Surveying only happy endings.** Some teams send the survey only when a ticket closes cleanly. Every conversation that ends should get the same question. - **Tiny samples.** Ten answers can swing the score by ten points. Report the number of responses beside the score every time. - **Asking too often.** A customer who writes in weekly should not get an NPS survey every time. Cap NPS at one request per person per quarter or so. - **Comparing with someone else's benchmark.** A good score depends on channel and industry. Chat ratings, email ratings and phone ratings rarely line up. Compare yourself with last month. - **Ignoring the comment box.** The number tells you that something is wrong. The comment tells you what. ## When CSAT answers outgrow a form and a spreadsheet A form plus a spreadsheet works for a founder answering the inbox alone. The guide to [building a customer satisfaction survey with free tools](/articles/customer-satisfaction-survey) walks through that setup step by step. It breaks once two or more people answer, because the sheet cannot tell you which ticket, which agent or which channel produced a rating. Matching answers to conversations by hand stops being realistic past a few dozen a week. Help desks solve that by sending the survey from the ticket itself, so every rating lands on the right conversation and agent. Help Scout covers both metrics in one place: it sends [CSAT and NPS surveys as Messages](/articles/help-scout-beacon) from the Standard plan. Zoho Desk keeps it simple with a three-option rating, and its [happiness ratings on tickets](/articles/zoho-desk-ticketing) start on Standard. Zendesk sends its survey automatically when a ticket is marked solved. Freshdesk sends a CSAT survey when a ticket closes, but its surveys have no NPS question. ### Help Scout — solves this Sends CSAT, NPS and microsurvey Messages to customers, and chat ratings of Great, Okay or Not Good feed the Happiness report. A workflow can reopen a poorly rated conversation and alert the team. - Where it stops: The Free plan has no surveys. - Plan: Standard - How Help Scout does it: [Set up CSAT and NPS Messages](https://professionalstoolkit.com/articles/help-scout-beacon) - More: [Help Scout](https://professionalstoolkit.com/tools/help-scout) ### Zoho Desk — solves this Adds a Good, Okay or Bad happiness rating link to every reply or sends it when a ticket closes; a negative rating can reopen the ticket, and the Radar app alerts on customer happiness. - Where it stops: No happiness ratings on Free or Express. - Plan: Standard - How Zoho Desk does it: [How happiness ratings work on tickets](https://professionalstoolkit.com/articles/zoho-desk-ticketing) - More: [Zoho Desk](https://professionalstoolkit.com/tools/zoho-desk) ### Zendesk — solves this Sends a satisfaction survey when a ticket is marked solved, mandatory on email and optional on messaging, so every rating is tied to one ticket. - Where it stops: Team plans get prebuilt dashboards; custom reports and dashboards need Suite Professional. - More: [Zendesk](https://professionalstoolkit.com/tools/zendesk) For the full field of options, including chat-first tools and store-focused help desks, see our comparison of [customer feedback tools](/articles/customer-feedback-tools). ## FAQ **Can I calculate NPS in Google Sheets?** Yes. Count the answers of 9 or 10 and the answers of 0 to 6 with COUNTIF, divide each count by the total number of answers, subtract the second share from the first, and multiply by 100. **Can NPS be negative?** Yes. If more customers score 0 to 6 than 9 or 10, the score drops below zero. It can run all the way down to minus 100. **Can a CSAT survey use a 1-to-10 or emoji scale?** Yes. Some businesses use 1 to 7, 1 to 10 or emoji ratings. Decide in advance which answers count as satisfied and keep that rule fixed, or the trend stops meaning anything. **How many responses do I need before trusting a CSAT score?** There is no fixed number, but a handful of answers can swing the score widely. Always show the response count next to the score and wait for a steady trend before acting on small moves. ## The bottom line CSAT and NPS answer different questions, so use both on purpose rather than together by habit. Send CSAT after every solved conversation and treat it as the support team's weekly instrument. Send NPS to customers a few times a year and treat it as the owner's loyalty check. Keep the formulas fixed, show the response count, read the comments behind every low score and reopen the conversations that earned them. Once more than one person answers customers, move the survey into the help desk so each rating ties to a ticket. ## Sources [1] Bain & Company (Net Promoter System) — https://nps.bain.com/about/measuring-your-net-promoter-score/ (Nov 2026) [2] SurveyMonkey — https://www.surveymonkey.com/learn/customer-feedback/csat-calculator (Nov 2026)