What Is CSAT? Guide to Customer Satisfaction Score

What is CSAT?

CSAT, short for Customer Satisfaction Score, measures how satisfied a customer was with one specific experience: a support case, an onboarding call, a delivery, a project milestone. It is a snapshot of a moment, not a verdict on the relationship. Hold on to that, because most CSAT mistakes come from asking a snapshot to answer a question about loyalty.

The standard question is "How satisfied were you with [experience]?", answered on a scale. Three formats are in use:

Use the 5-point scale unless you have a specific reason not to. It is also the scale most benchmarks assume, which matters later in this article. For how this kind of transactional metric differs from a relationship metric, see transactional vs. relational NPS.

How do you calculate CSAT?

CSAT is the percentage of respondents who gave one of the two highest answers (CustomerGauge, 2026).

CSAT = (Number of satisfied responses / Total responses) × 100

On a 5-point scale, answers of 4 and 5 count. Answers of 1 to 3 do not. That makes the score simple to calculate and exactly as easy to flatter.

Say you send a survey after support interactions to 300 customers and 150 respond:

Satisfied (4 and 5): 120. CSAT = 120 / 150 × 100 = 80%.

Now read the same table from the bottom. Twenty answers, 13%, were neutral. Ten answers, 7%, were a 1 or a 2. The 80% tells the board that things are fine. The ten are the ones who need a phone call.

Why should you manage the bottom box, not the average?

Most CSAT reporting stops at the top-box percentage, and that is the least useful number in the survey. Two programmes can both report 80% and be in very different shape. One has a few lukewarm customers. The other has a cluster of 1s concentrated in the accounts that pay the bills.

Take Nordika A/S, a fictional B2B supplier with 150 survey responses and a CSAT of 80%, just like the example above. Seven of its ten 1s and 2s come from its three largest customers. The average says nothing is wrong. The revenue at risk says something is.

So report three numbers, not one: the top-box percentage, the count of 1s and 2s, and which accounts they came from. Then treat each low score as a task with an owner and a 48-hour deadline, not as a data point. That is closing the loop, and it is where CSAT earns its keep. Reading the results by account value follows the same logic as account-based CX.

The counterintuitive part: a falling average with a stable bottom box is less alarming than a stable average with a growing one. You can lose three points of CSAT and be fine. You cannot ignore a 1 from your largest customer. If you have to choose what to watch, watch the 1s and 2s.

Who is missing from your CSAT?

CSAT is calculated on the people who answered, and in B2B that is a small group. CustomerGauge puts average B2B survey response rates between 2 and 20 percent (CustomerGauge, June 2026). Its own warning in the same release is that the largest and most strategically important accounts are often the ones that never respond, which leaves them effectively invisible to the CX programme. A CSAT of 85% on a 10% response rate is a statement about the 10%.

Four things help:

• Track coverage, not just score. Measure the share of your top accounts that answered in the last two quarters. Silence from a key account is a signal in itself, and it belongs in a customer health score next to usage and support volume.
• Reach silent accounts another way. Use an account-owner conversation, operational signals, or an AI agent that reads the e-mails and tickets already flowing through your business, instead of waiting for a survey reply.
• Raise response where it is cheap. Send within hours of the experience, ask one rating and one open question, and use a sender the customer recognises. More in how to improve your survey response rate.
• Read small samples as patterns. With a handful of contacts per account, a percentage means little. Look at the trend across milestones instead, as described in B2B survey sampling with a small n.

Where does CSAT work, and where does it mislead?

CSAT is a touchpoint metric. It answers "was this specific experience acceptable?" It does not answer "will this customer stay?"

Where it works. Customer support, after resolution, broken down by agent, channel and issue type. Onboarding, at day 7, 30 and 90, where a low score is an early warning for first-year churn (see the B2B onboarding survey). Delivery and implementation milestones. Feature releases, measured in the product while usage data is still ambiguous. Events and webinars, while the experience is fresh.

Where it misleads. CustomerGauge states plainly that CSAT does not predict renewal or repurchase (CustomerGauge, 2026). A customer can score a 4 today and leave next quarter because a competitor offered a better price. CSAT also ignores effort: a customer can be satisfied with the outcome of a case that took three transfers and two days. For loyalty and friction you need NPS and CES.

CSAT in B2B: measuring a project, not a ticket

CustomerGauge's benchmark research finds that 49% of NPS users also measure at least one additional metric, and over 33% add CSAT, typically to cover touchpoints where NPS is too blunt (CustomerGauge, 2026). But most CSAT guidance is written for transactional B2C moments: a support ticket, a delivery, a single purchase. B2B project work rarely fits that model.

A mid-size ERP or software implementation runs for months, involves a buying committee instead of a single respondent, and delivers value in milestones (kickoff, go-live, first quarter of use) rather than one-off tickets. IT-Effect, a Danish IT services company in the Microsoft Dynamics ERP world, shows the project-first principle in practice: instead of one blended company-wide score, satisfaction is measured after every project and support interaction, tied to the responsible project manager, and read project by project. A dip right after kickoff means something different from a dip after go-live. Blending the two into a single average would hide exactly the signal a project manager needs to act on.

Three adjustments make CSAT work for project-based B2B:

• Survey milestones, not every interaction. Tie the CSAT question to project phases (scoping, kickoff, go-live, 90 days post-launch). With a handful of stakeholders per account, fewer and better-timed surveys beat frequent, generic ones. The 90-day milestone doubles as an early churn signal when you time it alongside a B2B onboarding survey.
• Weight by project or account value. A CSAT built equally from a 50,000-kroner project and a 2-million-kroner implementation misrepresents your portfolio.
• Feed it into a health score, not just a dashboard. A single low CSAT at go-live is a data point. A low CSAT combined with slipping usage and a support ticket spike is a warning.

To turn any satisfaction data, CSAT included, into a ranked list of what to fix first, see key driver analysis.

How does CSAT compare with NPS and CES?

CSAT is the thermometer, NPS is the forecast and CES is the friction detector. None of them is complete on its own. In CustomerGauge's 2021 benchmark of 776 B2B companies, 41% named NPS their most trusted metric, 26% named CSAT and 11% named CES (CustomerGauge, 2021). A layered set-up uses NPS for relationship health, CSAT for touchpoint quality and CES for friction in processes. The full comparison is in NPS vs. CSAT vs. CES, and the two companion metrics are explained in What Is NPS? and What Is CES?. For where CSAT sits in the wider discipline, see what customer satisfaction is and how to measure customer satisfaction.

What is a good CSAT score?

There is no universal answer, and that is the honest one. Retently's 2026 benchmark, published in April 2026, finds that a score between 65 and 80 is the dominant range across industries. The spread between sectors is wide:

Source: Retently, 2026 CSAT benchmarks by industry.

Digital Marketing Agency made the biggest jump in the data, up from 72 the year before. Healthcare went the other way, from 81 to 57. Retently's article does not state the sample size or how the score is calculated, which is part of the point.

Read the table as a compass, not a scorecard. As CustomerGauge puts it: "Unlike NPS, CSAT and CES lack standardized cross-industry benchmarks." The scale (1-5, 1-7, thumbs, stars), the calculation method and what counts as a top score all differ between companies and between benchmark providers. Track your CSAT against your own history, and against competitors that publish on the same scale. A cross-industry average built on a scale you cannot verify tells you less than it appears to.

What we see in practice

Among the companies we work with, three CSAT patterns come up again and again.

Aggregate scores hide everything. At Nordika A/S, our fictional example, a company-wide CSAT of 82% tells leadership that things are fine. Break it down by team and you might find one team at 92% and another at 64%. The aggregate hid a serious problem.

Score inflation is the default. Agents who ask customers to "rate me a 5", or who choose which interactions get surveyed, produce data that measures their skill at gaming the system, not customer satisfaction. Use CSAT for coaching, not for bonuses.

The open question is where the value lives. The number tells you something is wrong. The text tells you what. Companies that skip text analysis leave the most actionable part of CSAT unused. An AI agent that reads and triages those comments as they arrive turns this from a quarterly research exercise into an operational one. See how that works in AI agents for customer satisfaction.

How do you run CSAT without fooling yourself?

1. Trigger surveys automatically and immediately. Integrate with your support, CRM or project tool so the survey fires after the interaction or milestone. The closer to the experience, the better the data.
2. Ask one rating and one open question. That is the ideal transactional CSAT survey. The follow-up, "What is the main reason for your score?", is not optional.
3. Segment by what matters. Break results down by account value, project, agent, channel and customer tenure. The insight lives in the segments, not in the average.
4. Act on every 1 and 2 within 48 hours. Route low scores to the owner automatically, as described in close the loop. Customers who give feedback and see nothing change become less satisfied than customers who were never asked.
5. Keep the scale fixed. Switching from a 5-point to a 3-point scale mid-programme destroys your trend data. Pick one and commit.
6. Give every touchpoint an owner before launch. A CSAT score nobody owns is a dashboard decoration.

What should you do first?

Pick two or three touchpoints that matter most to your business and run CSAT there. Report the bottom box next to the top box. Check which of your largest accounts actually answered. Act on the low scores. At SurveyGauge we build milestone-level CSAT into the same programme as NPS, read it by account value and name an owner for every low score. That is the difference between a dashboard number and a trigger for action.