Live Chat Metrics: 10 KPIs Every Support Team Should Track

Live chat metrics tell you whether your chat is helping customers or just sitting in the corner of the page. The trouble is that most dashboards show more numbers than anyone can act on. This guide narrows it down to ten KPIs that cover volume, speed, coverage, quality, team workload and the AI assistant — what each one tells you, how to read them together, and a weekly review you can run in fifteen minutes.

Why chat needs its own metrics

Email and phone metrics don't translate directly to chat. Chat customers expect a reply in minutes, not hours, and they often leave if they don't get one. Conversations are shorter, operators handle several at once, and with an AI assistant in front, some chats never reach a human at all. So the numbers that matter are about speed, coverage and how well the bot and the team work together.

Volume metrics: chats and messages per day

1. Conversations per day

The baseline. How many new chats start each day, and how that changes by weekday and hour. Volume drives everything else: staffing, response times, how much the bot is worth.

2. Messages per day

Messages per day, alongside conversations, tells you how long chats run. If messages grow faster than conversations, chats are getting longer — maybe because of a confusing product change, or because the bot is going back and forth without resolving anything.

Use volume to plan, not to judge. A drop in chats can mean fewer problems, or a broken widget.

Speed metrics: first response time, wait time after handoff

3. First response time

How long a visitor waits for the first reply. It's the metric customers feel most. Track the median as well as the average: a few chats that waited an hour can drag the average up while most customers got fast replies. Decide upfront whether bot replies count and whether you measure only within working hours. Our first response time guide covers the calculation choices.

4. Wait time after bot handoff

If an AI assistant answers first, the first response is almost instant — which makes first response time look great even if humans are slow. Wait time after handoff measures the gap between the bot transferring the chat and an operator replying. It's often where the real delay hides. See chatbot to human handoff for how to design that transition.

Coverage: response rate and missed chats

5. Response rate

The share of conversations that got a reply at all. A chat that nobody answered is worse than no chat: the visitor tried, and you ignored them. As a rule of thumb, anything below nearly all chats answered within working hours deserves a look at schedules and notifications.

6. Missed chats by hour

Break unanswered chats down by hour of day and day of week. Patterns jump out quickly: lunch breaks, shift changes, Monday mornings, weekends. These are the hours to cover with a schedule tweak, an on-call person, or an AI first line.

Quality: CSAT and resolution

7. CSAT

Customer satisfaction after the conversation, usually on a 1–5 scale. CSAT is commonly reported as the share of 4s and 5s. Watch the trend rather than a single week, and read the comments behind low scores — they tell you what to fix. The CSAT score guide covers calculation and sample size.

8. Resolution

Whether the chat actually ended with the problem solved. In practice, teams track this through conversation status (resolved vs still open or snoozed) and through repeat contacts: a customer who comes back with the same question a day later wasn't really helped.

Team metrics: replies and response time per agent, presence

9. Per-agent replies and response time

How many replies each operator sends and how fast. Use this to balance load and spot who needs help, not to rank people. An operator with slower replies may be taking the hardest chats. Pair the numbers with transcript reviews before drawing conclusions.

Presence is the context for these numbers: when each person was online or away. A timeline of team presence shows coverage gaps you won't see in averages — for example, everyone going away at the same time. The live chat support team guide covers scheduling around these gaps.

Bot metrics: share handled by AI, handoff rate

10. Bot share and handoff rate

If you run an AI assistant, track two related numbers:

• Share of conversations handled by the bot alone — chats that ended without a human.
• Handoff rate — the share of bot chats transferred to a person.

There's no universal "right" level. A rising bot share with stable CSAT is a good sign. A rising bot share with falling CSAT means the bot is answering things it shouldn't. A very high handoff rate usually means the bot lacks facts.

How to read metrics together, not in isolation

Single metrics mislead. Read them in pairs:

• First response time + wait time after handoff. Fast first response with a long post-handoff wait means the bot is masking slow humans.
• First response time + CSAT. Faster replies with dropping CSAT can mean agents are rushing.
• Volume + response rate. Response rate falling as volume rises points to staffing, not effort.
• Bot share + CSAT for bot chats. Automation is only a win if customers are still satisfied.
• Per-agent load + presence. Uneven load is often a schedule problem, not a performance problem.

Always compare with the previous period of the same length, and watch for one-off events — a sale, an outage, a holiday — that distort a week.

A simple weekly metrics review

Fifteen minutes, once a week, is enough for most small teams:

1. Check volume against last week. Any unexpected spike or drop? Find the cause.
2. Check first response time and wait time after handoff — median, not just average.
3. Check response rate and missed chats by hour. Fix the worst hour.
4. Check CSAT and read every low rating's transcript.
5. Check bot share and handoff rate. Read five handed-off chats; add any missing facts to the bot's instruction.
6. Check per-agent load and presence. Rebalance if one person carries too much.
7. Pick one thing to change this week. Just one, so you can see whether it worked.

In ChatDeck most of this sits on one analytics screen: messages per day, response rate, average first response time, wait time after bot handoff, CSAT trend, per-operator replies and response times, comparison with the previous period, and a team presence timeline. If you prefer to dig into conversations with an AI assistant like Claude or Cursor, the read-only MCP server lets you query analytics and transcripts with a personal token. The Read your analytics tutorial walks through each chart, and for what to look for inside the conversations themselves, see live chat transcript analysis.

Frequently asked questions

What are the most important live chat KPIs?

For most teams: first response time, response rate, CSAT and conversation volume. If you run an AI assistant, add wait time after handoff and handoff rate. Those six cover speed, coverage, quality and automation.

How often should I review chat metrics?

Weekly for most teams, with a quick daily glance at unanswered chats. Monthly reviews are useful for trends and staffing decisions, but weekly is frequent enough to catch problems before customers notice.

What is chat response rate?

The share of chat conversations that received a reply from your team or bot. It shows coverage: whether the people who tried to reach you got an answer. Look at it within working hours separately from after-hours.

How do I compare one period with another fairly?

Compare periods of the same length and the same weekdays, use medians for time-based metrics, and note one-off events like sales, outages or holidays. Comparing a promotion week with a quiet week will mislead you.

Want these numbers without spreadsheets? Create a free ChatDeck workspace — the analytics are included with no per-operator fee — and start with our guide to first response time.