Live chat costs real money — software seats, agent hours, training time — so “it feels like customers like it” is not a measurement strategy. But most teams measure chat the way they measure everything else: they open the vendor dashboard, glance at a wall of metrics, and pick the ones that look good. That is how you end up optimizing response time while resolution rates quietly collapse.
This guide takes the opposite approach. Start from the business questions chat is supposed to answer, then pick the smallest set of metrics that answer them honestly — with the actual formulas, so you know exactly what each number means and where it can mislead you.
Start With the Business Question, Not the Dashboard
Every metric below exists to answer one of four questions: Are we available when customers need us? Are we fast enough? Are we actually solving problems? And is it worth what it costs? If a number does not serve one of those questions, it is decoration. Before adding any metric to your reports, write down which question it answers — if you cannot, leave it out.
The most common measurement mistake is tracking what the dashboard makes easy instead of what the business needs to know. Convenience is not a methodology.
Volume and Availability: Are We There When It Counts?
These metrics describe the shape of demand and whether your staffing meets it. They are the foundation — speed and quality numbers mean nothing if half your visitors chat into the void.
Total chats and chats per visitor
Formula: total conversations ÷ total site visitors (per period). This tells you what share of traffic engages with chat. It varies enormously by industry and placement, so its value is as a trend within your own site: a sudden drop usually means a widget or trigger broke, not that visitors stopped needing help.
Missed chats and coverage rate
Formula: coverage rate = answered chats ÷ (answered + missed chats). A missed chat is any conversation request that arrived while no agent was available to take it within your target wait. Track the raw count too — ten missed chats a month is a staffing tweak; two hundred is a coverage redesign. If your coverage rate sits below roughly nine in ten during staffed hours, fix staffing before optimizing anything else.
Offline messages
Formula: count of after-hours form submissions per week, plus your actual reply time to each. Offline messages are demand you are deferring, not serving. If the pile grows, it is evidence for extending hours — bring the numbers to the staffing conversation rather than anecdotes.
Speed: Are We Fast Enough?
Speed is where chat beats every other support channel, and it is the metric visitors feel most directly. But raw averages lie, so measure carefully.
First response time (FRT)
Formula: median time from the visitor’s first message to the agent’s first reply. Use the median, not the mean — a handful of chats where the agent was juggling five conversations will drag an average up and hide the typical experience. Also measure it separately for proactive chats (where the agent opened) versus inbound chats, since expectations differ.
Full resolution time
Formula: median time from first message to conversation close. This is the number that correlates with satisfaction, more than first response does. A fast first reply followed by a forty-minute meander is not good service. Segment by issue type if you can — billing questions and technical troubleshooting have legitimately different shapes.
Wait time before pickup
Formula: time from chat request to agent assignment. Keep this distinct from first response time: assignment without a greeting is a queue, not service. If wait times spike at specific hours, that is your staffing signal — see our guide to staffing a live chat team for how to turn these numbers into shift plans.
Quality: Are We Solving Problems?
Fast and available means nothing if the answers are wrong. Quality metrics are harder to game and harder to collect — which is exactly why they matter more.
Customer satisfaction (CSAT)
Formula: (4–5 star ratings ÷ total ratings) × 100, from the post-chat survey. Two honest caveats: response rates are usually low (often under a third of chats), and unhappy customers are more motivated to rate than content ones. Most platforms include built-in post-chat ratings — Zendesk’s support documentation, for example, explains how satisfaction surveys are enabled and interpreted. Treat CSAT as directional — a ten-point sustained drop is a fire alarm; a two-point wobble is noise. Never tie individual agent bonuses to CSAT alone; it teaches agents to beg for ratings instead of solving problems.
First-contact resolution rate
Formula: chats resolved without a follow-up ticket, callback, or second chat ÷ total chats. This is arguably the single most valuable quality metric: it measures whether the conversation actually ended the customer’s problem. Track it by tagging a simple “resolved / not resolved” disposition at close — one click for the agent, gold for your reporting.
Reopen and escalation rate
Formula: chats reopened or escalated to another channel ÷ total chats. A rising reopen rate with steady CSAT means agents are closing chats prematurely to protect their speed numbers — the classic sign that you are measuring speed too aggressively and quality not enough.
Money: Is It Worth What It Costs?
This is the section your finance team actually reads. Everything here connects chat activity to revenue or cost with explicit math.
Cost per conversation
Formula: (monthly software cost + monthly staffed chat hours × loaded hourly rate) ÷ monthly conversations. The loaded rate includes benefits and overhead, not just salary — using bare salary understates cost by a third or more. This number lets you compare chat honestly against phone, email, and ticket costs per contact. For the software-cost half of this equation, our live chat pricing guide breaks down how seat pricing and tiers actually work.
Chat-influenced revenue
Formula: revenue from sessions that included a chat ÷ total revenue (same period). “Influenced” is doing honest work in that name — it does not prove chat caused the purchase. Strengthen the claim by comparing conversion rates: visitors who chatted versus similar visitors who did not. If chatters convert at meaningfully higher rates on the same pages, you have a credible story; if they do not, chat may be serving people who would have bought anyway.
ROI, stated carefully
Formula: (chat-attributed value − total chat cost) ÷ total chat cost. Chat-attributed value combines influenced revenue (discounted for the attribution caveat above) and deflected ticket costs (chats that resolved issues which would otherwise have become tickets, valued at your cost per ticket). Present this as a range, not a point estimate — “between 2× and 4×” is honest; “347% ROI” is numerology. Anyone selling you a precise ROI number for chat is selling something.
Agent Productivity: Are We Staffed Sensibly?
Productivity metrics exist to right-size the team, not to rank agents on a leaderboard. The two that matter:
- Concurrent chats per agent: the average number of simultaneous conversations. Sustained concurrency above three degrades quality for most issue types; if your team lives at four-plus, you need more people, not faster people.
- Conversations per agent hour: total chats ÷ staffed hours. Useful for capacity planning — how many agent-hours does a hundred chats actually consume? — and for spotting when a new trigger or page launch changes the workload overnight.
Publish these numbers to the team as capacity data, never as individual performance scores without context. An agent with low chats-per-hour might be handling your hardest technical issues; an agent with sky-high concurrency might be rushing everyone. If you must use productivity data in performance reviews, pair it with quality metrics — resolution rate and CSAT — so the incentive points toward solving problems well rather than closing chats fast. The cheapest way to improve every productivity number at once is better self-service content: every question the help center answers is a chat your agents never have to take.
The One-Page Monthly Report
If you produce one artifact from all of this, make it a single page with these rows, this month versus last month, with a one-line note on anything that moved more than 10%:
| Metric | Answers | Direction you want |
|---|---|---|
| Coverage rate | Are we available? | Up (target: 90%+ in staffed hours) |
| Median first response time | Are we fast? | Down or stable |
| Median resolution time | Are we efficient? | Down or stable |
| CSAT | Are customers happy? | Stable or up |
| First-contact resolution rate | Are we solving it? | Up |
| Cost per conversation | What does it cost? | Stable or down |
| Chat-influenced revenue share | Does it pay? | Up |
Seven rows. Anything more belongs in an appendix for the one person who asks. Pair this with the setup it measures — if you are just getting started, our implementation checklist gets the operation running so these numbers have something honest to describe.
A Note on Benchmarks
You will find plenty of articles claiming the “average” first response time or CSAT for live chat. Treat them as entertainment. Benchmarks vary wildly by industry, ticket complexity, staffing model, and — crucially — by how each vendor defines the metric. A “response time” measured from chat request and one measured from first message are different numbers wearing the same name. The only benchmark that matters is your own trend line: are your numbers moving the right direction month over month? Everything else is someone else’s business.
Conclusion
Measure availability before speed, speed before quality theater, and quality before money — but make sure the money question gets answered every month, with explicit formulas and honest ranges. Seven metrics on one page, reviewed monthly, will tell you more about your chat operation than any vendor dashboard. And when the numbers point at a specific weakness — slow replies, missed chats, low resolution — you will know exactly which lever to pull, because you will know exactly what each number means.
