Customer Satisfaction Score (CSAT): Formula, Benchmarks, and Limits

What is a customer satisfaction score (CSAT)?#

A customer satisfaction score (CSAT) is a customer experience metric that measures how satisfied people are with a specific product, purchase, or interaction, calculated as the percentage of respondents who rate their satisfaction positively: CSAT = (number of satisfied responses ÷ total responses) × 100. It is the most direct of the core CX metrics — a short-term, transactional read on a single touchpoint, usually collected right after the moment it measures.

CSAT is expressed as a percentage from 0 to 100, where a higher number means a larger share of customers were happy. It answers one narrow question well: were people satisfied with this? What it cannot answer on its own is the more valuable question — why? That gap is the theme this guide returns to after the mechanics, and it is the reason CSAT works best as the start of a conversation rather than the end of one.

How to calculate CSAT#

CSAT is calculated by dividing the number of satisfied responses by the total number of responses and multiplying by 100. The "satisfied" count is a top-box or top-two-box tally: on a 1–5 scale you count ratings of 4 and 5; on a 1–10 scale you typically count 7 through 10; on a 1–3 scale you count the top rating. Only positive ratings go in the numerator, so CSAT is a share-of-happy-customers figure, not an average.

The standard survey question is "How satisfied were you with [product / interaction / support experience]?" answered on a labeled "Very dissatisfied" to "Very satisfied" scale. Common scale variants are 1–3, 1–5, 1–7, and 1–10, and some teams collect a percentage rating directly. The 1–5 scale is the most widely used because it is fast to answer and maps cleanly onto the top-two-box method.

CSAT formula with a worked example#

The clearest way to see the CSAT formula is with real numbers. Suppose you send 100 post-support surveys and 80 customers respond. Of those 80, 50 rate their satisfaction a 4 or a 5. Your CSAT is (50 ÷ 80) × 100 = 62.5%. Note that the 20 people who never responded are excluded entirely — CSAT is calculated only on completed responses, which is the source of one of its biggest blind spots, covered below.

One alternative worth knowing: some teams report a composite or mean CSAT by averaging every numerical rating (for example, 4.2 out of 5) rather than applying a satisfied/not-satisfied threshold. The average method is easy to compute but harder to compare against industry benchmarks, most of which assume the top-two-box percentage. Pick one method and stay consistent, because switching between them mid-program makes your trend line meaningless.

CSAT benchmarks by industry#

A good CSAT score varies widely by industry, but as a rule of thumb most sectors treat 75–85% as a healthy range, with anything above 80% considered strong and sustained scores below 70% a signal to investigate. Because CSAT is collected differently by every team, the most reliable neutral benchmark is the American Customer Satisfaction Index (ACSI), a University of Michigan–founded national index that models satisfaction on a 0–100 scale across the U.S. economy.

The national ACSI slipped 0.3% to 76.7 in the first quarter of 2026 — a useful reference point for what "average" looks like across industries. The table below shows where major sectors landed:

Two cautions on benchmarks. First, the ACSI is a modeled 0–100 index, not a raw top-two-box percentage, so treat it as directional rather than a like-for-like comparison with your own CSAT. Second, a benchmark tells you where you stand, not what to change — chasing a competitor's number without understanding what drives it is how teams optimize the score and still lose customers. If you want the loyalty side of the picture too, pair CSAT benchmarks with 2026 NPS benchmarks by industry, since the two metrics move for different reasons.

CSAT vs NPS vs CES#

CSAT, NPS, and CES are the three core experience metrics, and they answer different questions: CSAT measures satisfaction with a moment, Net Promoter Score (NPS) measures long-term loyalty and likelihood to recommend, and Customer Effort Score (CES) measures how hard the customer had to work. Using them interchangeably is a common mistake — each is calibrated for a different decision.

CSAT's strengths are its immediacy and specificity: because you ask right after a defined event, the score attaches cleanly to that event. NPS trades that precision for a loyalty signal that predicts referral and retention over time. CES emerged from research published in Harvard Business Review arguing that reducing customer effort predicts repurchase and loyalty better than "delight" does — which is why support teams often lead with it.

Most mature programs run more than one. For the full field — CSAT, NPS, CES, customer lifetime value, retention, and the rest — the eight customer experience metrics that matter in 2026 maps when to reach for each, and the broader customer experience definition and the AI shift puts them in context. If your remit is operational, the 12 customer service metrics that matter covers where CSAT sits alongside first-contact resolution and handle time.

The limits of CSAT and how to fix them#

The central limit of CSAT is that it captures the number but not the cause — a 62.5% score tells you a third of respondents were unhappy without telling you what to fix. That single gap sits underneath most of the other well-documented problems with the metric, and it is the reason a score can look stable while the underlying experience quietly erodes.

Four limits show up in nearly every CSAT program:

• Response and self-selection bias. CSAT is calculated only on people who respond, and responders skew toward the extremes — the delighted and the furious answer, the indifferent middle ignores the survey. A high score can simply mean your quietly disappointed customers stopped replying.
• The ceiling / loyalty gap. Satisfaction is a weak predictor of loyalty. Research popularized in Harvard Business Review's work on the loyalty question found that large shares of customers who defect had reported being "satisfied" beforehand. Satisfied is not the same as retained.
• Scale and cultural inconsistency. A 4/5 from one respondent and a 4/5 from another can mean very different things, and satisfaction scores vary systematically by region and culture — which makes cross-market comparisons shaky.
• The recency trap. CSAT measures a single moment, so it swings with whatever just happened and rarely explains the pattern behind the swing.

You can shore up the first three with better survey hygiene — consistent scales, larger and more representative samples, transactional timing tied to the event. But the fourth problem, the missing why, is not a sampling issue you can fix with more of the same survey. A rating scale has no room for a reason. The standard patch is a single open-ended follow-up ("What's the main reason for your score?"), and it helps — but an open text box is a lonely, one-shot prompt. Most customers leave it blank, and the ones who fill it in write a fragment with no chance for anyone to ask "why?" back.

This is where the measurement model itself is shifting. Instead of a score plus a dead-end comment box, a conversational follow-up asks the customer to explain in their own words and then probes — the same way a good researcher would. Perspective AI runs that follow-up as an AI-moderated interview at the scale of a survey: when a customer gives a low CSAT, the interviewer asks what happened, why it mattered, and what would have made it better, turning a 62.5% into a ranked list of root causes. That is the difference between knowing your CSAT dropped and knowing it dropped because onboarding stalled at step three.

The broader move from dashboards to explanations is the subject of customer experience analytics: from dashboards to the why behind the numbers and the model-level argument in survey-based CX measurement vs conversational VoC. If your specific problem is a stuck CSAT number, the tactical playbook lives in AI CSAT analysis: turning satisfaction scores into root causes and how to use conversational AI to improve CSAT. For the underlying methodology debate, why conversations beat surveys for real customer research is the deeper cut, and CX teams standardizing on this approach can start from the resources built for CX teams.

Frequently Asked Questions#

What is a good CSAT score?#

A good CSAT score for most industries falls in the 75–85% range, with scores above 80% considered strong and sustained scores below 70% a signal to dig into what's going wrong. "Good," though, is relative to your sector: the American Customer Satisfaction Index put the 2026 U.S. average near 76.7 on its 0–100 scale, and full-service restaurants (82) routinely outscore internet service providers (73). Benchmark against your own trend and your industry, not a universal target.

How do you calculate a CSAT score?#

You calculate CSAT by dividing the number of satisfied responses by the total number of responses and multiplying by 100. "Satisfied" is typically the top two ratings on your scale — 4 and 5 on a 1–5 scale, or 7 through 10 on a 1–10 scale. For example, 50 satisfied responses out of 80 total gives (50 ÷ 80) × 100 = 62.5%. Only completed responses count, so non-responders are excluded from the math.

What is the difference between CSAT and NPS?#

CSAT measures satisfaction with a specific interaction or purchase in the short term, while NPS measures long-term loyalty and how likely a customer is to recommend you. CSAT uses a satisfaction scale scored as a top-two-box percentage (0–100%); NPS uses a 0–10 recommendation scale scored from −100 to +100. Use CSAT to evaluate a moment and NPS to gauge the overall relationship — most mature programs track both, as detailed in the CX metrics guide.

What scale should a CSAT survey use?#

A 1–5 satisfaction scale is the most common and usually the best default because it is quick to answer and maps cleanly onto the top-two-box calculation. Some teams use 1–3, 1–7, or 1–10 scales, and all are valid as long as you apply the "satisfied" threshold consistently. The one rule that matters most is not switching scales mid-program, since that breaks your ability to compare scores over time.

Why is CSAT not enough on its own?#

CSAT is not enough on its own because it reports what customers felt without explaining why they felt it, and it is skewed by who chooses to respond. A score can hold steady while the underlying experience decays, and a single open-ended comment box rarely recovers the reasoning. Pairing the score with a conversational follow-up that probes for the cause is how teams turn a satisfaction number into something they can act on.

Conclusion#

A customer satisfaction score is one of the fastest, clearest reads in the CX toolkit: an easy formula, a familiar percentage, and an immediate signal after any interaction. Learn the CSAT formula, benchmark it honestly against your industry rather than a vanity target, and pair it with NPS and CES so you are not reading a moment as if it were a relationship. Used that way, CSAT is a genuinely useful instrument.

But an instrument that reports the number without the cause can only take you so far. The teams pulling ahead in 2026 treat every CSAT rating as the opening line of a conversation, not the last word — and let an AI interviewer ask the follow-up a rating scale never can. If you're ready to hear the why behind your satisfaction scores, start a conversational study with Perspective AI and turn your next round of CSAT responses into root causes you can fix.