Help Center
Insights & Reports

Reading your analytics

What each metric means and what action it should trigger — conversations, resolution, CSAT, peak hours, channels.

Dashboard → Analytics (Starter plan and above). Pick a date range at the top; everything below follows it. Test conversations from the dashboard are excluded, so the numbers reflect real visitors only. Times are shown in your own timezone (set in Settings).

The core numbers

  • Conversations & messages — volume over time. Watch the trend, not the absolute number: the charts compare against the previous period.
  • Resolution / automation rate — the share of conversations the bot handled without a human. This is the number that translates to saved hours (shown as an estimate).
  • CSAT / feedback — thumbs up vs down from visitors. A falling CSAT usually means a knowledge gap opened up — check Unanswered next.
  • Lead conversion — how many conversations produced a captured lead.
  • Assisted revenue (Shopify stores) — orders placed within 7 days of a conversation on the bot, matched by the customer's email, split into with your team (a human greeted or joined the chat) and AI-only. It is deliberately labelled assisted, not generated: anonymous chats and post-purchase order lookups are never counted, so the number under-reports rather than over-claims. The same figure appears as a card on your dashboard home and one line in the Monday digest email.

Diagnostic views

  • Top unanswered questions — the fastest route to improvement; each is a Q&A away from fixed.
  • Peak hours — when your visitors actually write. Useful for staffing live chat and setting business hours.
  • Channels — web widget vs WhatsApp split, so you can see where your volume lives.

Customer satisfaction (CSAT)

On Growth and above, the chat widget asks visitors for a 1-5 rating (with an optional comment) when they close a conversation, and again when a human handoff is resolved — at most once per conversation, and only after they actually chatted. The Analytics page shows the average score, the rating distribution and recent comments for the selected period, with a trend against the previous period; your weekly digest includes a CSAT line whenever anyone rated. Declined or expired handoff requests are never surveyed, and dashboard test chats never collect ratings.

Exporting

CSV export takes any view to a spreadsheet for reporting.

A weekly routine that works

  1. Check the automation rate — did it move? Why?
  2. Open top unanswered → convert the top 3 into Q&As.
  3. Scan thumbs-down conversations in Training.

Fifteen minutes a week keeps the bot measurably improving.