How to Track Response Time and Resolution Rate

Key Metrics to Track

First Response Time

The time between when a customer sends their first message and when they get a response. For AI chatbot conversations, this is essentially instant (under 5 seconds). For human-handled conversations, this depends on agent availability and queue length. Industry benchmarks for live chat are under 1 minute, for email under 4 hours, and for SMS under 10 minutes. Track this separately by channel since customer expectations differ. See How to Handle Support Across Multiple Channels.

Average Response Time

The average time between any customer message and the next agent reply throughout the entire conversation, not just the first message. A support interaction might have a great first response time but take 30 minutes between follow-up messages. Track both to catch agents who are fast to pick up conversations but slow to follow through.

Resolution Rate

The percentage of conversations where the customer's issue was resolved. This is the hardest metric to measure accurately because "resolved" is subjective. Common approaches: the agent marks the conversation as resolved, the customer confirms resolution through a satisfaction survey, or the conversation ends without the customer reopening it within a set period (24-48 hours). Use at least two of these signals together for accuracy.

First Contact Resolution

The percentage of issues resolved in a single conversation without the customer needing to reach out again. This is the gold standard metric because it means the customer got their answer without multiple back-and-forth interactions. High first contact resolution correlates directly with customer satisfaction. AI chatbots with good knowledge bases typically achieve 60-80% first contact resolution on FAQ-type questions.

AI Deflection Rate

The percentage of conversations handled entirely by the AI without human involvement. This tells you how much work your AI is actually doing. A 70% deflection rate means 7 out of 10 conversations never reach a human agent. Track this alongside resolution rate because deflection without resolution (the AI answered but the customer came back to ask a human) is not real deflection. See How to Calculate Customer Support ROI.

How to Set Up Tracking

Common Mistakes in Support Metrics

Optimizing Speed Over Quality

If agents are measured only on response time, they send faster but less helpful replies. "Please try restarting your device" is fast to type but rarely solves the problem. Always pair speed metrics with quality metrics (resolution rate, customer satisfaction) so agents are incentivized to be both fast and thorough.

Counting Bot Messages as Resolutions

If the AI sends a response and the customer leaves without replying, that is not necessarily a resolution. The customer might have given up, found the answer elsewhere, or decided to call instead. Use a follow-up signal (no return contact within 24 hours, positive satisfaction rating) to validate AI resolutions rather than counting every unanswered bot message as a success.

Ignoring Channel Differences

A 2-minute response time is great for live chat but meaningless for email. A 90% resolution rate for password resets is easy, but 50% for complex billing disputes might be excellent. Break metrics down by channel and issue category to get actionable insights rather than blended averages that hide problems.