CSAT (Customer Satisfaction) is a survey metric that measures how satisfied a customer is with a specific interaction, product, or experience, expressed as the percentage of respondents who rated it favorably. You ask one question — “How satisfied were you with X?” — on a 1-to-5 scale, and CSAT is the share of people who picked the top two boxes (4 or 5).
CSAT is not a relationship metric and it is not a loyalty metric. It measures satisfaction with a moment — a support ticket, an onboarding call, a feature — not whether the customer will renew or recommend you. That is what NPS and retention metrics like NRR are for. Treating a high CSAT as proof the account is healthy is the most common way teams misread it: a customer can rate every ticket 5/5 and still churn because the product no longer fits.
The formula
CSAT = (Number of 4 and 5 ratings / Total responses) × 100On a 5-point scale, 4 (“satisfied”) and 5 (“very satisfied”) are the “top-two-box” responses. If 80 of 100 respondents rate you 4 or 5, your CSAT is 80%.
Some teams report CSAT as the raw average score (e.g. 4.2 / 5) instead of a top-two-box percentage. Both are valid, but they are not comparable — pick one method and keep it consistent, because a 4.2 average and an 84% top-two-box describe the same data with very different-looking numbers.
When to send it
CSAT is transactional. Trigger it right after the interaction you want to measure, while the experience is fresh:
- Post-ticket — fires when a support conversation is closed (the canonical CSAT use case in Zendesk, Intercom, Freshdesk).
- Post-onboarding — after a kickoff or go-live milestone.
- Post-feature — in-product, after a customer uses a new capability.
Keep the survey to one question plus an optional free-text comment. Response rates collapse past two questions.
Benchmarks
CSAT varies enormously by industry and by what you are measuring, so treat cross-industry numbers as loose. For B2B SaaS support interactions:
| Context | Weak | Solid | World-class |
|---|---|---|---|
| Post-ticket support CSAT | under 80% | 85-90% | 92%+ |
| Onboarding / CSM-led | under 75% | 80-88% | 90%+ |
The absolute number matters less than the trend and the segmentation. A flat 88% that drops to 78% for one product area is a clearer signal than the headline figure. Always read CSAT by segment (product area, agent, plan tier), not just in aggregate.
CSAT vs NPS vs CES
These three are not interchangeable — they answer different questions:
- CSAT — “Were you satisfied with this interaction?” Transactional, top-two-box of a 1-5 scale. Best for measuring a specific touchpoint.
- NPS (Net Promoter Score) — “How likely are you to recommend us?” Relationship-level, on a 0-10 scale, reported as % promoters minus % detractors. Best for tracking overall loyalty over time.
- CES (Customer Effort Score) — “How easy was it to get what you needed?” Effort-focused, and the strongest predictor of repeat behavior and reduced churn for support and self-service journeys.
The practical rule: use CSAT for touchpoint quality, CES when the journey is about reducing friction (support, onboarding flows), and NPS for the relationship trendline you report to the board. Most mature CS orgs run CSAT and CES transactionally and NPS on a relationship cadence.
Common pitfalls
- Reading CSAT as a health score. Guard: feed CSAT into a health score as one input alongside product usage and support volume — never let it stand alone.
- Survivorship bias. Only happy or only furious customers answer. Guard: track response rate alongside the score; a 90% CSAT at a 4% response rate is noise.
- Scale drift. Mixing 5-point and 7-point scales, or switching top-two-box to top-three-box, breaks the trendline. Guard: lock the scale and the calculation method, and annotate any change in your dashboard.
- Aggregating away the signal. A healthy average hides the failing segment. Guard: always cut CSAT by product area, agent, and plan tier.
Related
- NRR vs GRR — the retention metrics CSAT is often (wrongly) used as a proxy for