Customer satisfaction measures whether a system helps users accomplish their goals. In production, it is a user-centered measure of success: strong uptime, fast responses, and controlled costs do not compensate for answers that users find unhelpful. section 10.4 section 10.3.1
Assess satisfaction through both direct feedback and user behavior:
- Collect explicit feedback, such as ratings and complaints about responses. section 10.4
- Look for behavioral signals: whether users get answers, ask follow-up questions, abandon conversations, or escalate to human support. section 10.3.1
- Interpret these signals alongside technical performance; operational health alone cannot show whether users are being helped. section 10.3.1 section 10.4
Fast and available, but unsatisfying
A customer-service bot maintained perfect uptime, subsecond responses, and controlled costs, yet satisfaction fell to 2.1 out of 5. Users described its answers as technically correct but unhelpful, too generic, or off point—a failure traditional system metrics missed. section 10.4
