Analytics
Measure how your agent performs: conversation volume, resolution rate, CSAT and thumbs ratings, channel and topic breakdowns, sentiment trends, action success, and credit usage — all windowed by date.
View as MarkdownThe Analytics tab tells you how well your agent is doing its job — how much it's handling, how often it resolves issues on its own, how customers feel about it, and what it's being asked about. Use it to spot problems early and to prove the agent's value.
All metrics are windowed by date. Pick a range (the default is the last 7 days) and every chart and total recalculates for that window.
Headline metrics
| Metric | What it tells you |
|---|---|
| Conversations | Total conversations in the window — your support volume. |
| Resolution rate | The share of conversations the agent handled without escalating to a human. Your core efficiency number. |
| CSAT | Customer satisfaction from end-of-conversation 1–5 scores: an average rating and a satisfaction % (4–5 star share). |
| Ratings | Thumbs-up and thumbs-down counts on individual answers, from the widget and the playground. |
| Credits | Credits this agent consumed in the window — usage attributable to this agent. |
| Unanswered | Replies with no citations — a proxy for answers the agent couldn't ground in your data. |
A high resolution rate means the agent is deflecting routine tickets so your team handles only what needs a person. If it dips, your chat logs and the "unanswered" count usually point straight at a knowledge gap.
Volume over time
A per-day conversation volume chart across your selected window shows trends and spikes — a product launch, an outage, a promotion. Pair it with topic and sentiment breakdowns to understand what drove a spike, not just that one happened.
Channels and status
See conversations broken down by channel (website widget, email, WhatsApp, and so on) and by status:
- Bot — handled entirely by the agent.
- Escalated — handed off to a human in the Help Desk.
- Resolved — closed out.
The channel breakdown shows where your volume actually comes from, which helps you decide where to invest — for example, whether to prioritize a WhatsApp deployment.
Topics
Conversations are classified into topics so you can see what customers ask about most. Topics are the fastest way to find where to improve your data: if "returns" is a top topic and shows low confidence or thumbs-down ratings, that's exactly where to add a Q&A pair.
Sentiment
Each conversation is scored as positive, neutral, negative, or unspecified, with a sentiment-over-time view. Rising negative sentiment is an early warning — often a sign of a policy change customers dislike, or an answer the agent is getting wrong at volume.
Action success
If your agent uses actions — collecting leads, escalating, calling your APIs — Analytics shows how many action runs succeeded versus failed. A falling success rate flags a broken integration before customers complain.
Turning analytics into action
- 1Set a weekly reviewPick a window (last 7 days) and scan resolution rate, top topics, and sentiment.
- 2Follow a weak signal to the logsA low-resolution topic or rising negative sentiment → open chat logs filtered to it and read the real conversations.
- 3Fix the dataAdd a source or Q&A pair for the gap you found, then retrain.
- 4Confirm the trendWatch the metric the following week to verify the fix moved it.
Detailed analytics are available on Standard and higher plans. See Plans & billing for what each plan includes.