A CSAT survey asks one question after one experience. How satisfied were you? It is the oldest of the three common feedback scores. In 2010, the Corporate Executive Board found that 80% of service organizations used it as their main gauge (HBR).

What is CSAT?

CSAT is short for customer satisfaction score. It is the share of surveyed customers who say they were satisfied with a purchase, a contact or a stay. IBM describes the survey as one question: “How satisfied were you with your experience?” Customers answer on a scale of 1 to 3, 1 to 5 or 1 to 10 (IBM).

The score has no single inventor. The theory comes from Richard Oliver. His 1980 paper modeled satisfaction as a result of expectations and of how far an experience beat or missed them. He tested it in a field study of a flu shot (Oliver, 1980).

Measurement at scale came next. In 1992, Claes Fornell wrote up Sweden’s national survey. Sweden was the first country to measure satisfaction for a whole nation (Fornell, 1992).

The ACSI began in 1994 at the University of Michigan. The name is short for American Customer Satisfaction Index. It now covers over 400 firms in 40 industries. About 200,000 interviews a year feed it. Scores run on a 0 to 100 scale (ACSI).

How do you calculate CSAT?

Add the positive responses and divide by all responses. Then multiply by 100. On a five-point scale, a positive response is a 4 or a 5 (IBM). NiCE gives the same rule (NiCE). Teams call the result a top-2-box score.

For illustration, a program surveys 400 members after a points claim. Of those, 212 answer 5 and 104 answer 4. Another 52 answer 3, 20 answer 2 and 12 answer 1. The top-2-box CSAT is 316 divided by 400, or 79%. The average rating is 1,684 divided by 400, or 4.21.

Researchers have tested both versions. Neil Morgan and Lopo Rego used ACSI data from 1994 to 2000. They found that the average score best predicted later business results. The top-2-box score also had good value (Morgan and Rego, 2006).

A second study looked at 93 firms in 18 industries. Top-2-box satisfaction did best at predicting retention (de Haan et al., 2015). So report both figures, and report how many people replied.

Pair the score with a driver question. The US government sets a standard for its own surveys. Agencies must ask an overall question on a Likert scale, and five points are preferred. They add driver statements, such as “My issue was resolved,” to find the cause. The base statement reads “I am satisfied with the service I received from [program/service name]” (OMB Circular A-11).

What is a good CSAT score?

No one number is good everywhere, because scales and survey methods differ. Public figures show the spread. The ACSI travel study for 2026 scored airlines at 76 and lodging at 77 on its 0 to 100 index. Airbnb and Hilton tied at 79 (ACSI). Navan reported a CSAT of 96% for its overall platform and 81% for its virtual agent (Navan 10-K).

Those two sets of figures do not compare. The ACSI is a modeled index. It estimates satisfaction from survey inputs on expectations, quality and value (ACSI). Navan’s is a survey percentage. So set your own baseline. Ask the same question the same way, and track it by channel and member segment.

How does CSAT compare with NPS and customer effort score?

The three scores ask different questions. CSAT rates one experience. Net Promoter Score, created by Fred Reichheld of Bain in 2003, asks about the relationship (Bain). Customer effort score came from CEB researchers in 2010. It asks how hard a service request was (HBR).

ScoreThe questionScale and mathBest used for
CSATHow satisfied were you with this experience?1 to 5, share of 4s and 5sOne interaction, such as a claim or a redemption
NPSHow likely are you to recommend us to a friend or colleague?0 to 10, promoters (9 to 10) minus detractors (0 to 6)The overall relationship over time
Customer effort scoreHow much effort did you personally have to put forth to handle your request?Original scale 1 (very low effort) to 5 (very high effort)A service contact

The research crowns no single winner. In ACSI data, metrics built on recommendation intent had little or no predictive value (Morgan and Rego, 2006). In the 93-firm study, effort score had little to no predictive power alone. Adding it to satisfaction metrics improved retention forecasts (de Haan et al., 2015).

The CEB team found the reverse for service contacts. It surveyed more than 75,000 customers. Effort beat satisfaction and NPS at predicting loyalty, which it defined as intent to keep buying, spend more or speak well of the firm. It called CSAT a poor predictor (HBR).

At the firm level, ACSI satisfaction is tied to market value. The link is significant (Fornell et al., 2006). So treat each score as a read on a different moment.

How do loyalty programs use CSAT?

J.D. Power rates program satisfaction on a 1,000-point scale. Its 2016 hotel study found that 83% of highly satisfied members said they definitely will recommend the brand. These were members who scored 900 or higher. Among displeased members, who scored 649 or lower, the share was 10% (J.D. Power, 2016).

The same report weighted six factors. Account maintenance carried 23%, ease of redeeming 22% and ease of earning 18%. Variety of benefits and program terms carried 16% each. Customer service carried just 5% (J.D. Power, 2016).

In the airline studies, the basics move the score. Roughly half of members do not understand how to earn or redeem. Satisfaction climbs 123 points when members understand redemption and 131 points when they understand earning (J.D. Power, 2018). It runs 99 points higher among members who have had no problems with their program (J.D. Power, 2017).

American Airlines rose 7% to 78 in 2026. The ACSI tied this in part to customer appreciation for its loyalty program, mostly among business travelers (ACSI). That is the same link at the brand level.

A loyalty team can send the question after the moments those studies point to: a points claim, a redemption or a status change. Then split the score by tier, by channel and by whether the member had a problem. Our guide to measuring customer loyalty and the entry on Net Promoter Score cover the other scores a program tracks.

How GRAVTY handles CSAT

A CSAT answer says how a member felt. The member record says what the member did next. GRAVTY records each earn, burn, tier change and offer response as a timestamped event. Each event ties to a member identity (retention economics). A score then has behavior to be checked against.

Agentic AI Compass pulls internal KPIs, partner data and customer sentiment into one place (Compass). A team can ask it why a metric moved.

Frequently asked questions

What does CSAT stand for?

CSAT stands for customer satisfaction score. The name covers both the one-question survey and the percentage it produces. IBM calls CSAT a real-time customer survey relating to a specific product or service (IBM).

Who invented the CSAT score?

No single person did. Richard Oliver gave the theory in 1980. In it, buyers weigh what they expected against what they got (Oliver). Sweden ran the first national survey (Fornell). The ACSI came in 1994 (ACSI).

Is CSAT better than NPS?

Neither wins in every study. In ACSI data, average and top-2-box satisfaction predicted business results, and recommendation-based metrics had little or no predictive value (Morgan and Rego, 2006). Bain, which created NPS, applies it to a whole company or to a single store or service team (Bain). Use CSAT for a moment and NPS for the relationship.

What is the difference between CSAT and the ACSI?

CSAT is a score a company collects from its own customers. The ACSI is an index built from national consumer surveys, scored 0 to 100 from a model of expectations, quality and value (ACSI). Its 2026 travel study rests on 14,910 completed surveys. A company cannot compare its own CSAT percentage with an ACSI score.

What scale should a CSAT survey use?

A five-point scale is the safest choice. The US federal standard prefers five points for its required satisfaction question (OMB Circular A-11). IBM lists 1 to 3, 1 to 5 and 1 to 10 as common options (IBM). Pick one and keep it, so trends stay comparable.