NPS vs CSAT vs CES: which number answers your question

NPS was introduced by Fred Reichheld in the December 2003 Harvard Business Review article "The One Number You Need to Grow". Reichheld's team linked survey answers from more than 4,000 customers to what those customers later bought and referred, built 14 case studies, and found that a single recommendation question tracked growth in most of the industries studied.

The question is "How likely are you to recommend [company] to a friend or colleague?", answered on a 0–10 scale. 9–10 are promoters, 7–8 passives, 0–6 detractors. The score is promoters minus detractors, as percentages of everyone who answered. Net Promoter and NPS are registered trademarks of Bain & Company, NICE Systems and Fred Reichheld.

NPS is a relationship metric. It is meant to be asked of a representative sample on a regular cycle, such as quarterly, and read as a trend.

CSAT has no single inventor. It is the common-practice name for asking "How satisfied were you with [the specific thing]?" and reporting the share of satisfied answers. The usual version is a 1–5 scale from very dissatisfied to very satisfied, with 4 and 5 counted as satisfied ("top-two-box").

Don't confuse it with the American Customer Satisfaction Index. The ACSI began in 1994 at the University of Michigan's National Quality Research Center, developed by Claes Fornell. It publishes a 0–100 score estimated from several weighted questions per company. An ACSI score and your helpdesk CSAT percentage are different measurements and should never share a chart.

CSAT is a transactional metric. It works best when it is attached to one moment and asked within hours or days of it.

CES comes from "Stop Trying to Delight Your Customers" by Matthew Dixon, Karen Freeman and Nick Toman (Harvard Business Review, July–August 2010). Their study of more than 75,000 people using contact centers or self-service found that going above and beyond made little difference to loyalty, while reducing effort did.

The original question was "How much effort did you personally have to put forth to handle your request?", scored from 1 (very low effort) to 5 (very high effort). On that version, lower is better. The authors reported that CES predicted loyalty better than CSAT or NPS in customer service interactions, while NPS did better at the company level. That detail is often dropped when CES is sold as a replacement for NPS.

CEB, where the research was done, is now part of Gartner, which still promotes effort as the main driver of service loyalty. A later, widely used version is an agreement statement, "The company made it easy for me to handle my issue", on a 1–7 scale where higher is better. The two versions run in opposite directions, so record which one you ask.

More than most dashboards suggest. With the 200-response examples above, the approximate 95% margin of error is about ±5.5 points for CSAT (80%) and about ±10 points for NPS (+45). NPS is a difference between two proportions, so it carries roughly twice the noise of a single percentage at the same sample size. For CES, the error depends on how spread out the answers are. With an illustrative standard deviation of 1.4 on a 7-point scale, 200 responses give about ±0.2.

That means an NPS moving from +45 to +40 between two waves of 200 is well within noise. Before reacting, compare the change with the margin of error. Our margin of error calculator at /tools/margin-of-error-calculator gives the figure for a CSAT percentage directly, and the sample size calculator at /tools/sample-size-calculator tells you how many responses a target margin needs.

Response bias matters as much as sample size. People with strong feelings answer more often than the contented middle, so voluntary-response scores lean toward the extremes. That is tolerable if your channel, timing and wording stay the same every wave.

The strongest challenge is Keiningham, Cooil, Andreassen and Aksoy's "A Longitudinal Examination of Net Promoter and Firm Revenue Growth" (Journal of Marketing, 2007). Using 21 firms and more than 15,500 interviews from the Norwegian Customer Satisfaction Barometer, they could not support the claims that NPS is the single most reliable predictor of growth or that it beats customer satisfaction. The paper won the Marketing Science Institute/H. Paul Root Award.

Fisher and Kordupleski's "Good and bad market research: A critical review of Net Promoter Score" (Applied Stochastic Models in Business and Industry, 2019) concluded that NPS had not lived up to its claimed benefits, and argued for measuring the value customers perceive and what drives it.

Two practical criticisms follow from the arithmetic. Collapsing 11 points into three buckets throws information away: a customer moving from 0 to 6 changes nothing, while one moving from 8 to 9 changes the score. And the score cannot tell you why. The 2010 CES authors' own data also point to NPS being stronger as a company-level gauge than as a verdict on individual service contacts.

None of this makes NPS useless. It is a simple, comparable-over-time signal that many teams track consistently. Treat it as one input, don't use it as a bonus target for frontline staff, and never report it without the response count and the comments.

Yes, as long as each has its own moment. A common setup is CES after onboarding or support contacts, CSAT after deliveries or sessions, and NPS once a quarter to active customers. The rule that keeps this manageable: never ask two of them in the same survey about the same moment, and cap how often any one person is surveyed. A customer who gets three surveys in a week stops answering all of them.

Whatever the mix, add one open question after the score, such as "What is the main reason for your score?", and read low-score comments first. That question usually produces the list of things to fix. The score only tells you whether things are getting better.

CSAT and CES are ordinary rating questions in Zunoform: set the scale length and label both ends. Rating questions are on the Free plan, which includes 500 responses a month. That is enough for a small support team's CSAT, but tight for a high-volume helpdesk. Google Forms, Tally and Youform have no monthly response cap on their free plans.

NPS is a native question type with the 0–10 scale fixed, and the form's summary tab shows the score and the promoter, passive and detractor split. It is one of three field types that need the Pro plan ($14.99/month) to publish. You can build and preview it on Free, or collect the same data on Free with a 0–10 rating question and score it yourself in Google Sheets.

Zunoform is a form builder, not a customer-experience platform. It does not trigger surveys from your helpdesk or CRM events, run closed-loop case management, or benchmark you against other companies. If you need those, a dedicated CX tool is the better fit.