How do you calculate CSAT?
Short answer
· 5 min read
How to calculate CSAT: what does a worked example look like?
A worked example below applies the formula step by step, using real response counts from a single rating question.
Collect responses to one rating question, such as “How satisfied were you with this interaction?” on a 5-point scale.
Count how many responses landed at each score, from 1 (least satisfied) to 5 (most satisfied).
Define which scores count as satisfied. The standard CSAT formula uses top-2-box: scores of 4 and 5.
Divide the satisfied count by the total number of responses collected, including neutral and negative ones.
Multiply by 100 to get a percentage.
Here is a 400-response month applying that definition:
Score | Responses | Counts as satisfied? |
|---|---|---|
5 | 210 | Yes |
4 | 110 | Yes |
3 | 45 | No |
2 | 20 | No |
1 | 15 | No |
Satisfied responses: 320. Total responses: 400. CSAT: 320 ÷ 400 × 100 = 80.0%.
The neutral score of 3 stays in the denominator under the standard method. It counts as a response but not as a satisfied one. Drop those 45 neutral responses from the denominator instead, and the same data produces 320 ÷ 355 × 100, or about 90%. That ten-point gap, created by one silent choice about what belongs in the denominator, is a common reason a live dashboard and a monthly report disagree on the same underlying data.
The fix is to write the definition down once: which scores count as satisfied, whether neutral responses stay in the denominator, and whether partial responses or reopened tickets are included.
How do the scale and the definition of satisfied change the score?
Understanding how to calculate CSAT consistently starts with picking one scale and one threshold and never mixing them mid-program. Top-2-box counts only the top two ratings on a scale as satisfied and divides that count by total responses. Mean score averages every rating, including neutrals and negatives, then expresses that average as a percentage of the maximum possible score. Applied to the 400-response dataset above, each method produces a different number from identical data.
Method | What counts as satisfied | Result from the same responses | When to use it |
|---|---|---|---|
Top-2-box (5-point) | Scores 4 and 5 | 80.0% | Standard scale for support teams; comparable month to month |
Top-box only (5-point) | Score 5 only | 52.5% | Stricter read for teams chasing excellence, not just adequacy |
Mean score (5-point, as % of max) | All scores, averaged | Varies; hides distribution | Internal trend tracking only, not external benchmarking |
10-point, 9-10 threshold | Scores 9 and 10 | Not comparable to 5-point data | New programs choosing a scale, not switching mid-program |
3-point emoji scale | Smiling face only | Not comparable to either | Low-effort, high-volume channels like chat pop-ups |
Mean score hides distribution: a mix of 5s and 1s can average out to the same score as a block of neutral 3s, which is why most support teams report top-2-box instead. Scores from different scales are not comparable, and switching scales resets the baseline, so trend lines break the month a team changes format. Teams with no process yet need only one question, one 5-point scale, and one written definition of satisfied.
How do you know a change in CSAT is real and not noise?
A CSAT move is real only when it is larger than the margin of error for the sample size behind it, because CSAT is a proportion and every proportion carries statistical noise. In plain words, that margin at 95% confidence is roughly two times the square root of (satisfied-rate times dissatisfied-rate, divided by the number of responses). The practical anchors: a sample size around 100 responses carries a margin of up to roughly plus or minus 10 points, around 400 responses narrows that to roughly plus or minus 5 points, and around 1,000 responses gets to roughly plus or minus 3 points. A score moving from 80% to 83% on 150 responses sits inside the noise band and is not a change worth acting on.
Response bias compounds this. Only a minority of contacts respond to any satisfaction survey, and unhappy customers and delighted customers tend to respond at different rates than the indifferent middle, so any filter that changes who gets surveyed shifts the score independently of actual experience. The fix is segmentation: calculate CSAT separately by contact reason, by channel, and for AI-handled versus human-handled conversations, since one blended number usually hides a single failing reason code.
How does OnClarity help with this?
OnClarity (www.onclarity.com) does not send surveys and is not a survey tool. OnClarity Voice of Customer analyzes the survey results, calls, chats, tickets, and reviews a team already collects, and ties score movements to the topics and root causes behind them. So when CSAT drops from 80% to 74%, the analysis can trace that drop to the specific contact reasons driving it, including differences in CSAT by channel, rather than leaving teams to guess at a single blended number.
OnClarity is SOC 2 Type II certified, GDPR and HIPAA Ready, and aligned to Saudi PDPL.
Learn more on the Voice of Customer page, or see the related CSAT glossary entry and customer satisfaction index glossary entry. To see this on your own data, book a demo.
Frequently asked questions
Do neutral responses count in CSAT?
Yes, under the standard method neutral responses (a score of 3 on a 5-point scale) stay in the denominator but are never counted as satisfied. Removing them inflates the score, as shown earlier when dropping 45 neutral responses pushed the data from 80% to roughly 90%. Keep every rated response in the denominator, satisfied or not.
What is a good CSAT score?
No independent source publishes one good CSAT score, since scale, threshold, and survey population all change the number. Track your own trailing baseline by contact reason and channel, and judge movement against it rather than an external figure. The American Customer Satisfaction Index (2025) reports national satisfaction near 77, not comparable to a top-2-box score.
How often should you report CSAT?
Report CSAT monthly for a blended view, but only report a segment once it clears roughly 100 responses, since smaller segments carry margins of error too wide to separate real change from noise. High-volume channels and common contact reasons can support weekly reporting once the sample size per segment stays above that threshold; low-volume segments should roll up quarterly instead.
Key terms
CSAT
CSAT, or customer satisfaction score, is a post-contact rating averaged across responses and usually collected with a single question immediately after an interaction closes. Because only a small fraction of customers respond, CSAT reflects the experience of people motivated to answer rather than the whole contact population.
Voice of customer
Voice of customer is the discipline of turning reviews, support tickets, surveys, and conversations into operational decisions. The distinguishing feature of a working programme is the closed loop: findings reach the team that can fix the cause, and the resulting change is measured against contact volume.
Customer satisfaction index
A customer satisfaction index is a composite score that combines multiple satisfaction measures into a single number, usually scaled from 0 to 100. The term covers two distinct uses. The American Customer Satisfaction Index, founded at the University of Michigan and now run by ACSI LLC, applies a standardized econometric model across roughly 400 companies to benchmark satisfaction by industry and over time. A company-internal customer satisfaction index instead blends its own CSAT, effort, and loyalty responses into one weighted figure. Unlike a single-question CSAT percentage, which reflects one interaction, a CSI aggregates several attributes. Its weighting method has to stay documented and unchanged for the score to be comparable period over period. See OnClarity’s CSAT glossary entry for how the single-question version works.
Root cause analysis
Root cause analysis is the practice of tracing a symptom, such as a spike in complaints, back to the underlying process, policy, or product defect that produced it. The alternative, resolving each ticket in isolation, never stops the symptom from recurring. Established methods include 5 Whys (repeated questioning to reach the originating cause), fishbone (Ishikawa) diagrams (mapping possible causes by category), and Pareto analysis (isolating the small number of causes driving most of the volume). In CX, this starts with clustering tickets by theme, then tracing the largest theme to its source. OnClarity’s Voice of Customer automates this by surfacing root causes directly from feedback. Grubhub uses it to turn feedback into insights that shape its roadmap. The real test of success is recurrence: a fix that addresses the root cause should stop the issue from reappearing.
More answers on CX metrics
How do you improve CSAT?
How to improve CSAT comes down to fixing the causes behind low scores, not the survey itself. Read the comments and conversations tied to low ratings, fix the top two or three drivers, then re-measure CSAT by contact reason to confirm which fix actually moved the number.
What is a good NPS score?
Above 0 is a reasonable starting point, since it means promoters outnumber detractors. Scores above +50 are commonly read as strong within the NPS score range of -100 to +100. But there is no single good NPS score across industries. Judge your result against your own baseline over time, not an outside number.
What is a good CSAT score?
A good CSAT score typically falls between 75 and 85 percent for transactional support, with anything above 85 percent considered strong, though that range is a convention rather than a fixed standard. What is a good CSAT score for your team depends on industry, channel, and contact reason, so your own trend matters more than any published benchmark.
