TapCub is live — analytics, insights and live chat, free on one platform
Glossary

What is chat analytics? Metrics, formulas and how to read them

Chat analytics measures live chat conversations: volume, first response, resolution, satisfaction and demand by hour. The five numbers to read together, a two-week worked example, and the mistakes that skew them.

Xu An Reviewed by Sophia Bennett Also called live chat analytics Last updated Nov 17, 2026 Glossary

Quick answer

Chat analytics is the measurement of live chat conversations: how many start, how fast someone replies, how many end resolved, how visitors rate them, and when they begin. Read the metrics together, because any one of them can look healthy while another fails.

Formula: Resolution rate = resolved conversations ÷ all conversations. First response time = average seconds from conversation start to the first reply.

What is chat analytics?

Chat analytics is the measurement of live chat conversations: how many start, how fast someone answers, how many end resolved, how visitors rate them, and when and where they begin. It turns a stream of individual chats into numbers a team can compare week to week and act on.

The phrase covers two different jobs, and the search results mix them. The first is reporting on chat operations: volume, response time, resolution, satisfaction, workload per agent. The second is text analytics on chat transcripts: grouping conversations by topic, intent or sentiment, which is sometimes called conversation analytics. This page is about the first job, because it is the one every chat team needs on day one and the one you can verify with arithmetic. Topic and sentiment mining sits on top of it.

The core chat metrics

Five numbers answer most operating questions. Read them together: any one of them can look healthy while another is failing.

MetricQuestion it answersFormulaWhat to watch
ConversationsHow much demand is there?Conversations started in the periodCount chats, not visits; a chat is not an analytics session
First response timeHow long does a visitor wait?Seconds from conversation start to the first reply, averagedAverages hide long waits; read the median next to it
Resolution rateDo chats end with a solved problem?Resolved conversations ÷ all conversationsChats that began late in the period may still be open
Satisfaction (CSAT)Did the visitor feel helped?Average of the 1–5 scores givenFew visitors rate; check how many did
Hourly demandWhen do chats start, and is anyone there?Conversations started, by weekday and hourCompare against office hours, not against a flat average

Two supporting measures matter once you add automation. Self-service rate is the share of questions answered from a knowledge base before a person is needed (see ticket deflection for the general idea). Handoff reasons tell you why a bot gave way to a person: no matching answer, the visitor asked for a human, a turn limit, or an agent taking over.

Definitions are not standard across tools. One vendor documents first response time as the time after an agent is assigned, and starts the clock at bot handover when no agent is assigned (Freshworks, linked in Sources). A tool that starts the clock at the visitor's first message would produce a different number from the same chats. Before comparing numbers across tools or against a benchmark, find out where the clock starts, whether bot replies count as a response, and what counts as resolved. The related glossary pages cover first response and resolution rate in more detail, and live chat defines the channel itself.

A worked example: two weeks of chat

Sample data: a store compares two consecutive weeks of live chat.

MetricWeek 1Week 2Change
Conversations400520+30.0%
Resolved312364+16.7%
Resolution rate312 ÷ 400 = 78.0%364 ÷ 520 = 70.0%−8.0 points
Average first response52 s41 s−11 s
Median first response30 s25 s−5 s
Satisfaction ratings received6149
Rated share of resolved chats61 ÷ 312 = 19.6%49 ÷ 364 = 13.5%−6.1 points
Average satisfaction4.44.1−0.3
Started outside office hours96 ÷ 400 = 24.0%162 ÷ 520 = 31.2%+7.2 points

Read alone, "first response got faster" says the week went well. Read together, the picture is different. Volume rose 30%, answers came faster, but the share of chats that ended resolved fell eight points, and the extra chats are arriving when nobody is on shift: 31.2% of week-2 chats began outside office hours against 24.0% a week earlier. The satisfaction average also rests on fewer ratings, so treat the 0.3-point drop as a signal to look, not as proof.

This is a hypothesis, not a diagnosis. The decision it supports is where to look next: the hour-by-weekday view for the uncovered hours, the pages where chats start, and the conversations that closed unresolved.

Why the median sits next to the mean

Take nine first-response times in seconds: 12, 15, 18, 20, 25, 30, 40, 60 and 480. They sum to 700, so the mean is 700 ÷ 9 = 77.8 seconds. The median, the fifth value in order, is 25 seconds. One visitor who waited eight minutes pulls the mean to three times the typical wait. Report both, and when they diverge, go and read the slow conversation.

Common mistakes when reading chat data

  • Adding chat conversations to analytics sessions. A conversation is one thread with one visitor. A session is a visit cut by an idle timeout. They are different units and should never be summed.
  • Counting a bot reply as a first response. If an automated greeting counts, first response time looks excellent and tells you nothing about people. Decide the rule once and keep it.
  • Treating offline messages as resolved chats. A message left after hours has not been answered yet. Count it on its own line.
  • Judging satisfaction from a tiny base. A 4.8 average from 6 ratings is weaker evidence than a 4.2 from 90. Always show the count of ratings next to the score.
  • Reading resolution rate over a window that ends today. Chats started in the last hours of the window may still be open, so the latest days understate the rate. Compare complete periods.
  • Comparing periods with different staffing. A holiday week with half the agents will show slower first response and lower resolution without any change in quality.
  • Treating every handoff as a failure. A handoff to a person on a question the knowledge base should not answer, such as a complaint or a custom quote, is the system working.

Chat analytics, ticket reports and conversation analytics

TermUnitTypical question
Chat analytics (this page)A live chat conversationAre we answering quickly and solving the problem?
Conversation analyticsThe text of chats, calls, emailsWhat are customers talking about and how do they feel?
Self-service reportingA question answered from a help sourceWhat share of questions never needed a person?

They answer different questions and share inputs. Start with operating numbers; add topic analysis when volume is high enough that nobody can read every chat.

Reading chat analytics in TapCub

Take the two-week table above. Here is how to rebuild it, and where to look next.

  1. Open the overview. Go to Live chat › Reports. The page shows conversations, resolution rate (resolved ÷ conversations), average first response with a median alongside, and satisfaction (a 1–5 score the visitor can give within 24 hours after the chat is resolved, one score per conversation). Each metric shows its change against the previous period, and date ranges follow the site time zone, not your browser's.
  2. Check what the first-response number includes. TapCub counts only a person's reply. An AI reply is not a first response, and a conversation answered only by the AI is excluded from that average. If the AI alone answered and nobody on the team ever replied, any score is credited to the AI rather than to an agent name.
  3. Find the uncovered hours. The heatmap on the same page counts chat starts by weekday and hour. Set it beside your hours in Live chat › Chat settings › Office hours to see whether demand lands outside them. The hot-pages list shows the top 10 pages where visitors started a chat, without the query string.
  4. Read workload by person. Agent workload lists, per person, conversations handled and resolved, replies, average first response, satisfaction, and transfers out and in. Members with the Agent-only role see only their own handling by default; an owner can widen that under Live chat › Agents.
  5. Check how much the knowledge base carries. Knowledge base hit rate is the share of questions in the period answered by a direct Q&A match or a retrieved answer. In the inbox, each handoff is labeled with its reason.

Two boundaries to keep in mind. Offline messages are counted on their own and are never part of "resolved". And TapCub does not have ticket numbers or service-level timers: the reports measure conversations, and a handoff-rate percentage is not one of the metrics on the page. If you need one, count the labeled handoffs yourself.

Because chat and analytics share one visitor ID, the Conversion attribution view on the same page compares goal conversion for visitors who chatted with those who did not. That answers a different question, whether chat is worth having, and sits outside the five operating numbers above.

Sources

  • Contentsquare, What is chat analytics?, glossary definition and metric list, checked 2026-11-17.
  • Freshworks, Chat analytics module definitions, first response time and resolution time definitions, checked 2026-11-17.
  • TapCub documentation, Chat reports and conversion (Live chat › Reports), checked 2026-11-17.

Was this explanation clear?

We use your answers to fix and expand this page.

Xu An

Customer Experience Consultant

Writes about live chat, visitor conversations, and self-service, sharing practical ways to make customer communication clearer and more effective.

FAQ

Questions people ask

Short answers to the follow-up questions that come up most.

Still have a question?

Chat with the team behind TapCub — we usually reply within a few hours on business days.

What is chat analytics?

Chat analytics measures live chat conversations: how many start, how quickly a person replies, how many end resolved, how visitors rate them and when chats begin. It can also mean text analysis of transcripts for topics and sentiment, but the operating metrics come first.

Which chat metrics matter most?

Conversations, first response time, resolution rate, satisfaction and demand by hour. Read them together: faster replies mean little if the resolution rate falls, and a high satisfaction average from a handful of ratings is weak evidence.

Why show the median first response time?

One very slow chat pulls the average up. With nine chats of 12, 15, 18, 20, 25, 30, 40, 60 and 480 seconds, the mean is 77.8 seconds but the median is 25. Reporting both shows the typical wait and the outlier.

Do offline messages count as resolved chats?

No. A message left after hours has not been answered yet, so count it on its own line. In TapCub, offline messages are counted separately and are not part of resolved conversations.

See this number on your own site

Add one script tag for web and apps, and the first reports fill in within minutes.

See your data clearly. Find your growth.

Every click, backed by data. Install one line of code and see your first numbers in a minute.

No credit card · Free plans for analytics and chat · Cookieless analytics