Customer satisfaction score (CSAT) is the percentage of surveyed customers who say they were satisfied with a specific interaction, product, or service, usually those who answer 4 or 5 on a five-point scale. In customer service it is asked right after a conversation, so it measures how that contact went rather than loyalty to the brand.
CSAT is the most common quality metric in support because it is simple to ask and easy to explain to a board. Its weakness is the same simplicity: one question, answered by a minority of customers, shortly after the contact.
Behind the number sits a long-standing model of satisfaction. Customers compare what happened with what they expected, and satisfaction depends on that gap as much as on the outcome itself.3 A fast refund pleases a customer who expected a fight and disappoints one who expected it automatically.
How CSAT is calculated
The survey asks one question, such as "How satisfied were you with the help you received today?", on a scale from 1 (very dissatisfied) to 5 (very satisfied). The score is the share of responses in the top two boxes. U.S. federal agencies use the same top-two-box method for their five-point customer experience questions: the number of 4 and 5 responses divided by the total.2
- Conversations closed this week
- 4,000
- Survey responses received
- 600 (15% response rate)
- Responses rated 5
- 330
- Responses rated 4
- 162
- Responses rated 1 to 3
- 108
- CSAT = (330 + 162) ÷ 600 × 100
- 82%
Illustrative numbers. Report the response rate next to the score: 82% from 15% of customers says nothing about the other 85%.
Some teams use a thumbs up or down instead of five points, which gives the share of positive answers directly.2 Others average the 1 to 5 ratings, which produces a number like 4.3. Both are called CSAT, so check the method before comparing scores across teams or vendors.
CSAT, NPS, and CES
The three metrics answer different questions, and a support team usually needs more than one.
| Metric | Question | Scale and score | Measures | Best used for |
|---|---|---|---|---|
| CSAT | How satisfied were you with this interaction? | 1 to 5; share of 4s and 5s | One contact or experience | Quality of support conversations, by team, topic, or agent |
| Net Promoter Score (NPS) | How likely is it that you would recommend us to a friend or colleague?4 | 0 to 10; % promoters (9 to 10) minus % detractors (0 to 6)5 | Relationship and loyalty | Company or product level, tracked over time |
| Customer Effort Score (CES) | How much effort did you personally have to put forth to handle your request?6 | 1 (very low) to 5 (very high effort)6 | Friction in getting help | Service design: transfers, repeat contacts, channel switching |
- Responses
- 600
- Promoters (9 or 10)
- 300, or 50%
- Passives (7 or 8)
- 180, or 30%
- Detractors (0 to 6)
- 120, or 20%
- NPS = 50 minus 20
- 30
Illustrative numbers. NPS is reported as a number from -100 to 100, not a percentage.
The CES research found effort a better predictor of loyalty in service interactions than satisfaction or NPS, while NPS performed better at the company level.6 In practice, CSAT tells you whether a conversation went well, CES tells you why customers struggle, and NPS tells you how the relationship is trending.
What CSAT does not tell you
Non-response bias
ControlThe customers who answer may not be typical of everyone you served. Track the response rate and compare respondents with the full population by topic and channel.
Missing the customers who left
ControlSomeone who abandoned a chat or hung up on an IVR is never surveyed. Pair CSAT with abandonment and repeat-contact rates.
Mixing the product with the service
ControlA customer rates a late parcel low even when the support agent did everything right. Tag the reason for the score and separate service issues from product, policy, or logistics issues.
Survey design changes the number
ControlWording, scale, timing, and channel all shift scores, so a change in the survey can look like a change in quality. Fix the method and change one thing at a time.
Gaming
ControlWhen agents are paid on CSAT, some ask for good ratings or avoid sending surveys after hard cases. Send surveys automatically and audit low-scoring conversations.
CSAT for conversations handled by AI
When an AI agent handles part of the volume, CSAT becomes one of the main checks that automation is not hurting customers. It needs a few adjustments to stay fair.
- Compare like with like. AI agents usually take the simpler requests first. Compare AI and human CSAT for the same request type (order status with order status), not overall.
- Attribute handovers carefully. If the agent hands over and a person resolves the case, decide in advance whether the score counts for the AI, the person, or both.
- Read the low scores. Sample every conversation rated 1 or 2 and tag the cause: wrong answer, no action possible, slow handover, or a policy the customer disliked.
- Pair it with outcomes. A pleasant conversation that did not solve the problem still scores well sometimes. Track repeat contacts within a few days and the containment rate alongside CSAT.
How teams use CSAT
Peak-season quality check
During peak season, an online retailer tracks CSAT daily by topic to see whether delivery questions are handled as well as in a normal week.
survey → by topic → staff or fixFinding broken processes
Low CSAT clusters on one topic, such as contract cancellations, point to a policy or process problem rather than an agent problem.
low scores → tag cause → change processRolling out an AI agent
A lender launches an AI agent on instalment questions and compares its CSAT with human agents on the same topic before adding more topics.
pilot topic → compare CSAT → expandDisruption handling
An airline measures CSAT separately for disruption contacts, where expectations are low and speed matters most.
segment → measure → set targetsIllustrative examples, not client results.
Common questions
What is a good CSAT score?
It depends on the industry, the survey method, and what customers contact you about, so the useful comparison is your own score over time and by topic. Published benchmarks rarely use the same scale and question, which makes them hard to compare.
How do you calculate CSAT?
Divide the number of satisfied responses (4 or 5 on a five-point scale) by the total number of responses and multiply by 100. For example, 492 satisfied out of 600 responses gives a CSAT of 82%.
What is the difference between CSAT and NPS?
CSAT asks how satisfied a customer was with a specific interaction. NPS asks how likely they are to recommend the company, on a 0 to 10 scale, and subtracts the share of detractors from the share of promoters. CSAT measures a moment; NPS measures the relationship.
When should a CSAT survey be sent?
Right after the interaction closes, while the customer remembers it, in the same channel where possible. For a case that spans several contacts, send it when the case is resolved, not after every message.
References
- ISO. ISO 10004:2018 Quality management: Customer satisfaction: Guidelines for monitoring and measuring. International Organization for Standardization, 2018 (catalog entry, Estonian national standards body).
- Performance.gov. Trust in Major Government Service Providers dashboard: handbook. U.S. federal customer experience program (OMB Circular A-11, Section 280).
- Oliver, R. L. A Cognitive Model of the Antecedents and Consequences of Satisfaction Decisions. Journal of Marketing Research 17, no. 4 (1980): 460-469.
- Reichheld, F. F. The One Number You Need to Grow. Harvard Business Review, December 2003.
- Bain & Company. Measuring Your Net Promoter Score.
- Dixon, M., Freeman, K. and Toman, N. Stop Trying to Delight Your Customers. Harvard Business Review, July 2010.