How to Calculate CSAT in a Call Center (2026)
CSAT in a call center is calculated with a simple formula: CSAT (%) = (Number of satisfied customers ÷ Total number of responses) × 100, where "satisfied" is usually a 4 or 5 on a 1 to 5 post-interaction survey. Most teams land between 75% and 85%, which is considered strong, and the real work is less about the math than about picking the right scale, calculating cleanly across channels, and segmenting results so they drive action. Balto , the AI Workforce for the contact center, automates CSAT scoring across every interaction and connects each score to the agent, queue, and topic patterns driving it, so teams stop computing in spreadsheets.
The CSAT Formula: The 10-Second Answer
The CSAT formula is one of the simplest metrics in the contact center. You take the number of satisfied customers from your post-interaction surveys, divide by the total number of survey responses, and multiply by 100 to get a percentage. "Satisfied" is almost always defined as customers who rated the interaction a 4 or a 5 on a 1 to 5 scale.
Here is the formula and a worked example:
CSAT (%) = (Number of satisfied customers ÷ Total number of responses) × 100
Worked example. If 100 customers responded to your post-call survey and 82 of them rated the interaction a 4 or a 5, your CSAT is (82 ÷ 100) × 100, which equals 82%. That lands in the "good" range for most contact centers.
Two things usually trip teams up at this stage. First, you count only the top ratings (4s and 5s on a 5-point scale) as "satisfied," not everyone above 2 or 3. Second, "total responses" is the number of customers who actually filled out the survey, not the number of customers you sent it to. Confusing those two numbers is the single most common CSAT calculation mistake.
6 Steps to Calculate CSAT Accurately
The formula itself takes 10 seconds. Running a CSAT program that gives you numbers you can trust takes a repeatable process. The six steps below are the exact sequence most well-run contact centers follow, in order.
Step 1: Collect post-interaction responses
Trigger the CSAT survey immediately after each interaction closes. For voice calls, the standard is a post-call IVR prompt or an SMS link sent within 5 minutes of the call ending. For chat, use an in-window prompt the moment the chat resolves. For email, send a short follow-up with a single question.
Step 2: Filter valid responses
Not every response counts. Exclude test accounts, internal transfers, and interactions where the customer hung up before reaching an agent. Many teams also exclude interactions under 30 seconds because those tend to be misroutes rather than real conversations. Set the filter once, document it, and apply it consistently.
Step 3: Define what counts as "satisfied"
On a 1 to 5 scale, the industry standard is to count 4s and 5s as satisfied. On a 1 to 3 scale, count only 3s. On a 1 to 10 scale, teams typically count 9s and 10s, following the Net Promoter Score convention. The critical rule is to pick one definition and stick with it across every reporting period.
Step 4: Run the formula
Divide satisfied responses by total responses, multiply by 100, and round to one decimal place. Report CSAT as a percentage with a visible denominator. "CSAT 82% (n=100)" is more useful than "CSAT 82%" alone because it shows the sample size behind the number.
Step 5: Segment the result
A single company-wide CSAT number hides most of the insight. The same 82% average can mean "every queue around 82%" or "one queue at 95% and another at 68%." Break the score down by agent, by queue, by channel, by shift, and by issue type. The segmentation is where the coaching opportunities live.
Step 6: Report with trend context
A single-week CSAT number is noise. Report it alongside a rolling 4-week average and a year-over-year comparison so you can tell signal from random variation. Quality programs that only show the current week overreact to normal swings and miss the real patterns.
CSAT-3 vs. CSAT-5 vs. 10-Point Scales: Which Should You Pick?
The scale you pick determines everything downstream: response rate, "satisfied" definition, interpretability, and how comparable your scores are to industry benchmarks. Three scales dominate contact centers in 2026, and each one has clear strengths and tradeoffs.
| Scale | Satisfied Definition | Response Rate | Interpretability | Best For |
|---|---|---|---|---|
| CSAT-3 (1 to 3) | Count only 3s | Full | Light | IVR-only programs where speed matters more than nuance |
| CSAT-5 (1 to 5) | Count 4s and 5s | Partial | Full | Standard voice and omnichannel programs (industry default) |
| 10-point (1 to 10) | Count 9s and 10s | Light | Full | Enterprise teams that want NPS-comparable granularity |
For most call centers, CSAT-5 is the right default. It matches the ACSI benchmark methodology, maps cleanly to industry averages, and holds up well across voice and digital channels. CSAT-3 wins on response rate but loses nuance, which hurts coaching signal. The 10-point scale gives you the most granularity but completion rates drop meaningfully, which creates sample-size problems for smaller queues. For a deeper comparison across CX metrics, see our guide to CSAT vs NPS vs CES .
Channel-Specific CSAT Calculation
The formula does not change by channel, but everything around it does: when you send the survey, what the response rate looks like, and how you weight scores when you report an omnichannel total. Getting these per-channel mechanics right is the difference between a CSAT program that reflects reality and one that reflects whoever happens to answer.
Voice. Send the survey within 5 minutes of the call ending, through IVR or SMS. Expect an 8 to 12% response rate on average. Voice responses skew toward the extremes (very happy or very frustrated), so calibrate by including comment-box open text for context on scores at either end.
Chat. Trigger the survey in the chat window the moment the conversation closes. Expect 15 to 25% response rates, the highest of any channel. Chat CSAT tends to run slightly lower than voice because the medium itself is less personal, so compare chat scores to other chat scores, not to voice.
Email. Send a short follow-up email within an hour of resolving the ticket. Response rates sit at 3 to 7%, which makes smaller queues especially sensitive to sample size. Report email CSAT with weekly rolling averages rather than daily numbers.
Omnichannel. When you report a single CSAT number across channels, use a weighted average, not a simple mean. The weight is the number of valid responses per channel. This keeps a high-response, low-score channel from being masked by a low-response, high-score channel. For deeper multichannel analysis, see our guide to analyzing Voice of the Customer in a contact center .
Segmenting CSAT by Agent, Queue, and Time
A company-wide CSAT number is a report, not a lever. The useful score is the one broken down by the dimension you can actually coach against. In practice, four segmentation cuts drive almost all CSAT-based coaching.
By agent. Compute per-agent CSAT weekly, with a minimum sample size threshold (usually 15 responses) to avoid overreacting to tiny samples. Rank agents in bands (top 25%, middle 50%, bottom 25%) rather than individually, which keeps coaching conversations constructive.
By queue or skill group. Different queues drive different baseline expectations. A refunds queue and a new-sales queue will never post the same CSAT, so compare each queue against its own trend rather than against the company average. First Call Resolution is a strong leading indicator per queue, so pair CSAT segmentation with FCR data per our guide to measuring and improving First Contact Resolution .
By time-of-day and shift. CSAT typically drops 3 to 7 points late in the shift as agent fatigue sets in. If your evening shift scores noticeably lower, the fix is usually coaching plus break policy, not agent replacement.
By issue type or interaction reason. Tag each interaction with the reason customers called (billing, technical, cancellation, etc.) and compute CSAT per reason. The pattern often shows one or two reasons dragging the overall score down, which gives you a targeted process fix.
CSAT Calculator
Use the calculator below to compute CSAT instantly. Enter the number of satisfied customers (4s and 5s on a CSAT-5 scale) and the total number of valid responses, and the calculator returns the CSAT percentage plus a quick interpretation band.
5 Common CSAT Calculation Mistakes (and How to Avoid Them)
Most CSAT programs get the formula right. What breaks is everything around it. Here are the five mistakes that cause the most trouble in real contact centers.
1. Counting sent surveys as the denominator instead of completed responses. The formula is satisfied divided by RESPONSES, not satisfied divided by surveys sent. Using surveys sent artificially shrinks your CSAT because your response rate is baked into the denominator. Always divide by valid completed responses only.
2. Changing the "satisfied" definition mid-year. Some teams start with 4s and 5s, then switch to "4s, 5s, and 3s" to boost the number. Your trend line is now broken because the comparison between weeks is apples to oranges. Pick one definition and document it. For teams that need to improve scores, our guide to proven ways to improve customer satisfaction covers the coaching plays that actually move the number.
3. Mixing scales across channels. Running CSAT-5 on voice and CSAT-3 on chat and then reporting a combined number produces a meaningless average. Keep scales consistent within a reporting period, or report per-channel only.
4. Ignoring sample size. A per-agent CSAT of 100% based on 3 responses is not a coaching signal, it is noise. Set a minimum response threshold (usually 15 for agent-level, 30 for queue-level) and label anything below the threshold as "insufficient sample" rather than reporting a misleading percentage.
5. Reporting the single-week number without trend context. Weekly CSAT bounces by 2 to 4 points on normal operational variation. If you react to every weekly swing, you coach from noise. Report alongside a 4-week rolling average and compare to the same week last year.
How to Automate CSAT Calculation with AI
Running CSAT manually in a spreadsheet works at 500 responses per week. It breaks around 2,000 and becomes impossible around 10,000. At scale, there are three places AI meaningfully changes how CSAT calculation works in a contact center.
Automated response coding. Open-text comments attached to CSAT responses contain the real "why." AI sentiment analysis reads every comment, codes it by driver (wait time, agent knowledge, resolution quality, etc.), and surfaces the top three or four drivers each week. This is usually the first automation step. For a deeper look, see our guide to replacing customer service surveys with sentiment analysis .
Automated QA scoring at 100% coverage. Manual QA scores 1 to 3% of calls. AI-powered QA scores every call against the same standards and flags the interactions most likely to drive low CSAT before the survey even goes out. Our guide to automated quality assurance software covers how this layer connects to CSAT.
Real-time agent coaching on the standards that drive CSAT. When agent assist runs the same playbook QA scores against, frontline agents fix in-the-moment what they would otherwise lose a CSAT point on. This is the closed-loop that lifts QA scores 10 to 20 points in the first months of deployment, which is a strong leading indicator for CSAT lift of similar magnitude.
What Counts as a Good CSAT Score After You Calculate It?
A good CSAT score typically lands between 75% and 85%. Top performers maintain 90% or higher consistently. For a complete definition of CSAT and how it fits with other CX metrics, see our full guide to CSAT in customer service . For the full 2026 benchmarks by industry, channel, and scale, see our dedicated guide to what is a good CSAT score .
How Balto Automates CSAT Calculation and Lift
Balto automates both halves of the CSAT workflow: measuring the score accurately and lifting the underlying drivers. Insights tracks CSAT across every interaction without a separate spreadsheet pipeline, segmented automatically by agent, queue, channel, and topic. QA scores 100% of conversations against the same playbook that powers real-time agent assist, which closes the gap between what gets measured and what gets coached.
The result is what Balto customers describe as a closed-loop: every call makes the next call better because the same standards guide the agent in real time, score the interaction, and reinforce the behavior in coaching. In the first months of deployment, customers typically see QA scores lift 10 to 20 percentage points and CSAT follow within a quarter.
Turning CSAT Numbers Into Action
Calculating CSAT in a call center is simple on the math and hard on the mechanics. The formula is one line. What separates teams that use CSAT to drive real improvement from teams that use it as a weekly reporting ritual is the discipline around scale choice, channel-specific mechanics, segmentation, sample-size hygiene, and trend context. Pair that discipline with automation that connects the score to the underlying drivers and you have a CSAT program that actually moves the number.
For the definitional backdrop, see our full guide to CSAT in customer service .
FAQs
CSAT in a call center is calculated with a simple formula: CSAT (%) = (Number of satisfied customers ÷ Total number of responses) × 100. On a 1 to 5 survey scale, "satisfied" is the customers who rated the interaction a 4 or 5. For example, if 100 customers responded and 82 rated the interaction a 4 or 5, CSAT is (82 ÷ 100) × 100, which equals 82%.
The denominator is completed survey responses, not the number of surveys sent. That distinction catches a lot of teams at the start of a CSAT program, so always divide by valid responses only.
Customer Satisfaction Score (CSAT) is calculated the same way across industries: divide the number of satisfied respondents by the total number of respondents, then multiply by 100. The result is expressed as a percentage.
Most organizations count only the top ratings as "satisfied" (4s and 5s on a 1 to 5 scale, or 9s and 10s on a 1 to 10 scale). The precise cutoff depends on the scale you pick, and the key operational rule is to stay consistent so your trend data stays comparable period over period.
The CSAT calculation is: (Number of satisfied customers ÷ Total number of responses) × 100. Count 4s and 5s on a 5-point scale as "satisfied," divide by the total completed responses, and multiply by 100 for the percentage.
The simplest way to remember it is "top-rating respondents divided by all respondents, times 100." Everything else in a CSAT program is about the operational hygiene around that formula: picking the right scale, filtering valid responses, segmenting by agent and queue, and reporting with trend context.
To compute the CSAT percentage, take the count of customers who rated the interaction a 4 or 5 (on the standard 5-point scale), divide by the total number of customers who responded to the survey, and multiply by 100. The result is a percentage between 0 and 100.
If 82 of 100 respondents rated the interaction a 4 or 5, the CSAT percentage is 82%. Always round to one decimal place for reporting, and show the response count alongside the percentage so readers understand the sample size.
The CSAT formula is: CSAT (%) = (Number of satisfied customers ÷ Total number of responses) × 100. It is one of the simplest metrics in the contact center and is the same formula used across industries from retail to healthcare.
The variation across programs is in how "satisfied" is defined and which scale the survey uses. On a 1 to 5 scale, "satisfied" is 4s and 5s. On a 1 to 10 scale, it is 9s and 10s. Pick the definition, document it, and apply it the same way across every reporting period.
CSAT-3 uses a 1 to 3 rating scale (unsatisfied, neutral, satisfied), and "satisfied" is counted as ratings of 3. CSAT-5 uses a 1 to 5 scale, and "satisfied" is counted as 4s and 5s. CSAT-5 is the industry default because it balances response rate with interpretive nuance.
CSAT-3 wins on completion rate (customers fill out shorter surveys more often) but loses coaching signal because it compresses sentiment into three buckets. CSAT-5 gives you a clearer view of which agents or queues are consistently delivering 5-star versus merely acceptable experiences, which is where the coaching value lives.
The formula is the same across channels, but the response rates and survey mechanics differ. Voice uses post-call IVR or SMS surveys (8 to 12% response rate). Chat uses an in-window prompt (15 to 25%). Email uses a follow-up message (3 to 7%).
For a single omnichannel CSAT number, use a weighted average rather than a simple mean. The weight is the number of valid responses per channel, which keeps a high-response, low-score channel from being masked by a low-response, high-score one. Report per-channel CSAT alongside the blended number so leaders can see both the aggregate and the channel-specific picture.
Segment CSAT by agent and queue by computing the formula at that level rather than company-wide. For per-agent CSAT, set a minimum sample threshold (usually 15 responses per week) to avoid overreacting to tiny samples. Report per-queue CSAT weekly and compare each queue against its own historical trend rather than the company average.
The segmentation itself is where the coaching value lives. A single company-wide CSAT of 82% can hide a top queue at 95% and a struggling queue at 68%. The per-agent and per-queue breakdown tells you where the coaching should go.
For transactional CSAT tied to specific interactions, calculate after every interaction or a representative sample. For reporting, weekly is the standard cadence for most contact centers, paired with a rolling 4-week average and a year-over-year comparison to tell signal from normal variation.
Reporting daily is usually noise unless your interaction volume is very high (a bounce of 2 to 4 points on daily numbers is operational variance, not a trend). Reporting monthly can be too slow to catch coaching opportunities. Weekly with trend context hits the useful middle.
Exclude partial responses (customers who opened the survey but did not answer the rating question) from both the numerator and the denominator. They are neither "satisfied" nor "not satisfied," so they should not count either way.
For missing responses (customers who never opened the survey), the formula naturally handles them because they are not in your response count. The operational concern is response-rate decay over time: if your response rate drops from 12% to 6% over a quarter, your CSAT trend is probably misleading because you have lost representativeness. Monitor response rate alongside the score itself, and investigate any decay larger than a few percentage points.
Liked what you read? See Balto in action.
Balto helps leading contact centers turn insights into outcomes, in real time. Book a live demo to discover how our AI powers better conversations, coaching, and conversions.