The Performance tab of Insights shows how your support is doing: how many conversations come in, how many the AI handles on its own, how often a chat is handed to a person, how quickly people get an answer and how long it takes to resolve things — overall, per channel and per operator.
Performance is available on every plan, including Free, and to every member of the project. It covers all channels — web chat, Telegram, WhatsApp, Viber, email and the helpdesk — except Telegram group chats.
Opening it
Click Insights in the sidebar, pick the project and switch to the Performance tab. In the toolbar choose a date range and granularity, and optionally narrow everything to one channel or one operator.
A web chat or other channel’s settings page also has a small Performance card with that channel’s last seven days and a View full dashboard link.
The numbers
- Conversations — conversations that started in the period.
- Escalation rate — the share that were handed to a person, by the AI, by your escalation rules or by an operator. See Escalation and Operator Handoff.
- Touchless rate — the share where no operator ever replied: the AI handled the whole conversation. Conversations younger than 30 minutes aren’t counted yet, so a chat that just started doesn’t count as touchless before anyone had the chance to reply.
- AI first response — how long the customer waited for the AI’s first answer.
- Operator response after handoff — from the moment a conversation was handed to a person to the first operator reply. Conversations that were never handed over are not part of it, because nobody was waiting for a human.
- Resolution time — from the start of the conversation until it was resolved or closed, including automatic closing.
- Customer satisfaction — the average 1–5 rating and how many people gave one. See Customer satisfaction ratings.
The time tiles show the average with the median underneath; the median is the better guide when a few very slow conversations pull the average up. Click a tile (all except Customer satisfaction) to plot it on the graph. The period still in progress is drawn as a dotted line.
Every number is grouped by the day the conversation started: a chat opened on Monday and resolved on Wednesday counts towards Monday.
By channel and by operator
- By channel — conversation volume per channel.
- Resolution time by resolver — average resolution time for each operator who resolved conversations, plus an AI row for conversations resolved without an operator. The small number is how many conversations each row covers.
- Response after handoff by operator — how quickly each operator answered conversations handed to them.
Click a row to filter the whole tab to that channel or operator. The operator filter keeps the conversations that operator resolved; choose AI to see the ones resolved without a person.
Reading it well
- A rising touchless rate with a steady customer satisfaction score means the AI is taking real work off your team. If satisfaction drops as touchless rises, check what the AI is answering — How AI Responses Work.
- A high escalation rate on one channel often points to a gap in your knowledge base for the questions that channel brings.
- Long operator response after handoff outside working hours is expected if nobody is on shift; set up schedules so conversations route to whoever is on duty — Managing Your Team.