Customer satisfaction score, or CSAT, asks customers to rate a specific interaction or experience. In ecommerce support, it can reveal whether a resolution felt clear, fair, and easy. This ecommerce CSAT guide explains how to use the metric without overreading a small or biased sample.
CSAT is usually calculated as positive responses divided by all valid responses, multiplied by 100. Define which ratings count as positive and keep that definition stable.
Survey at a meaningful moment
Send the survey after the customer has enough information to judge the support outcome. A delivery claim may not be resolved when the first agent reply is sent. Surveying too early measures message tone rather than resolution.
Avoid repeated surveys across one issue and set a sensible expiry. Make the question clearly about the support interaction if that is what the score represents.
Segment before drawing conclusions
| Segment | Why it matters |
|---|---|
| Ticket intent | Refund disputes and product questions have different baselines |
| Channel | Chat and email create different timing expectations |
| Outcome | Approved, rejected, escalated, and unresolved cases differ |
| First-contact resolution | Extra effort often affects satisfaction |
| Market and language | Translation and policy variation can change experience |
| Automation involvement | Compare assisted, automated, and manual paths fairly |
An overall score may rise because the queue contains more easy questions, not because service improved.
Read comments with the score
Customer comments explain whether dissatisfaction came from speed, clarity, policy, tone, effort, or the underlying product and delivery outcome. Use customer service intent classification and a small set of experience themes to analyze comments at scale.
Read positive comments too. They show which behaviors and workflows should be repeated, not merely which problems should be fixed.
Separate support quality from policy dislike
A customer can dislike a return decision even when the agent applied policy accurately and communicated well. That feedback still matters: the policy or expectation may need review. But it should not automatically be treated as an agent failure.
Combine CSAT with customer service quality assurance , repeat contacts, resolution time, and escalation outcomes to understand the full case.
Check response bias
Report survey response rate and sample size. People with unusually good or bad experiences may be more likely to respond. Compare the surveyed group with the full ticket population by channel, intent, market, and outcome.
Do not rank individual agents on tiny samples. Use comments and quality reviews for coaching, with enough context to be fair.
Turn findings into action
- Identify a repeated driver in comments and scores.
- Validate it against tickets and operational data.
- Assign the issue to the team that can change it.
- Improve the policy, workflow, content, or system.
- Monitor the relevant segment after the change.
For example, low satisfaction on delayed shipments may reflect missed follow-up promises. The fix could be customer service SLA automation , not warmer apology language.
AI can categorize comments and surface themes, but humans should review representative examples and sensitive cases. CSAT is most useful as a listening system connected to action, not a decorative number on a dashboard.