Chat support needs its own scorecard. A WhatsApp or web chat team does not work like a phone queue: customers write at any hour, wait between messages, and an agent can keep several conversations open at once. If you measure chat with phone metrics such as AHT or service level in seconds, the numbers look strange and push the team in the wrong direction. This guide covers the chat support KPIs that matter, how to calculate them, sensible targets, and how to read them in a WhatsApp inbox.
Why chat needs different KPIs
On a phone call, the customer and the agent are both on the line until the call ends. On chat, the conversation is asynchronous. A customer may send a question at 9:05, reply to your answer at 9:40 and come back the next morning. Three things follow:
- Speed is about replies, not about answering. There is no ringing phone; the first reply is what the customer waits for.
- Handle time is hard to define. An open chat may last hours of clock time with only minutes of agent effort.
- Agents work in parallel. One agent can reasonably manage a few chats at once, which changes how you plan staffing.
For the phone side, see call centre KPIs and wallboards. For website chat specifically, see website live chat.
The core chat support KPIs
| KPI | What it measures | How to calculate |
|---|---|---|
| First response time | How long a customer waits for the first human or bot reply | Time of first reply minus time of first customer message, averaged |
| Average response time | How quickly every customer message gets a reply | Average gap between a customer message and the next reply |
| Resolution time | How long until the issue is closed | Time from first message to the point the issue is solved |
| Chats per agent per hour | Workload | Conversations handled ÷ agent hours |
| Concurrent chats | Parallel load per agent | Open conversations an agent is handling at one time |
| Bot containment | Share of chats the bot finishes | Chats closed by the bot without handover ÷ all bot chats |
| Sentiment trend | Mood of customer messages | Share of positive, neutral and negative messages over time |
First response time
This is the chat version of speed of answer, and the KPI customers feel most. During business hours, many teams aim for a first human reply within a few minutes. Outside hours, set the expectation clearly with an automatic message so the customer knows when a person will reply.
Response time percentiles, not just the average
An average hides the customers who waited a long time. If 90 chats got a reply in 1 minute and 10 waited 40 minutes, the average is about 5 minutes and looks fine. Look at the 90th percentile as well: the time within which 90% of replies were sent. That is where complaints come from.
Resolution time
Resolution time is the closest thing to handle time. Order status questions should close in one exchange; a refund or technical fault will take longer. A rising resolution time on a simple topic usually means agents are missing information, not effort.
Concurrent chats and chats per agent
Concurrency is the chat lever that phones do not have. Two or three simple chats at once is common; more than that and reply times stretch and mistakes creep in. Complex topics, such as complaints or payments, deserve one chat at a time. Use the hourly pattern of incoming messages to plan shifts, the same way shift scheduling uses call arrivals.
Quality KPIs for chat
- Repeat contact rate. The share of customers who come back about the same issue within a few days. High repeat contact means the first answer did not solve it.
- Handover rate from bot to agent. If the bot hands over most chats, its flows are not answering the real questions. If it hands over almost none, check that customers who need a person can reach one.
- Conversation review score. Review a sample of chats each week with a short scorecard: correct answer, clear language, right tone, proper close.
- Satisfaction. A one-question WhatsApp survey after the issue is solved. See CSAT and NPS for how to calculate and use the score.
Reading chat KPIs in a WhatsApp inbox
The BitVoice Social Media Platform answers WhatsApp, Instagram Direct and web chat in one shared inbox, and its admin dashboards cover most of the numbers above:
- Overview dashboard: total messages, active users, average response time, delivery rate, message volume, message types, sentiment and the hourly activity pattern, which you can use to plan agent shifts.
- Engagement dashboard: message volume by hour, a day and hour heatmap, messages per user, response-time percentiles and weekly retention.
- Sentiment dashboard: counts of positive, neutral and negative messages, the trend over time and the most common words on each side. A rising negative line is an early warning.
- Templates dashboard: sent, delivered, failed, pending and read for each template.
- Failed messages dashboard: every message WhatsApp refused, with the reason.
- Conversations report: every conversation with the last message and last activity, and a text download for audits and complaint reviews.
Each dashboard has a date range picker and export to PDF, and some to Excel or CSV. For the bot side, intent analytics show which intent rules fired, how often and with what confidence, which helps you see what customers ask and which rules send chats to a human.
Setting targets that make sense
Targets should follow what customers expect from each channel and topic:
- Start from your current numbers. Measure two to four weeks before setting targets.
- Set targets by hour type. Business hours and after-hours need different promises, and after-hours should be backed by an auto-reply that says when a person will respond.
- Pair a speed KPI with a quality KPI. First response time alone rewards quick, empty replies. Pair it with repeat contact rate or a review score.
Common mistakes
- Using phone AHT for chat, which punishes agents for customers who reply slowly.
- Measuring only averages and missing the long waits.
- Reporting chat and calls separately, so nobody sees that a slow chat queue is driving phone calls up. Look at both together, as in an omnichannel contact centre.
Want WhatsApp, Instagram and web chat in one inbox with dashboards your supervisors can use? Talk to BitVoice.
FAQs
What is a good first response time for WhatsApp support?
During business hours, many teams aim for a first human reply within a few minutes. Outside hours, an automatic reply that sets a clear time for a human answer matters more than speed.
How many chats can one agent handle at once?
For simple questions, two or three at a time is common. For complaints, payments or technical faults, one at a time gives better answers. Watch reply times: if they stretch, concurrency is too high.
Should chat support use average handle time?
Not as a main KPI. Chat conversations stay open while the customer is away, so clock time overstates effort. Use first response time, resolution time and chats per agent hour instead.
Why look at response time percentiles instead of the average?
The average hides long waits. The 90th percentile shows how long the slowest one in ten customers waited, which is where most complaints come from.
