📘

Part of AI Quality Monitoring. This page describes the Conversations tab, screen by screen.

Conversations is about behaviour, not quality. It answers who your users are, how they interact, when they're active, and where they come from. There are no AI metrics on this report — it's traffic and engagement, and it covers your full history (no 2026 cutoff). None of these sections drill down to individual conversations.

All three global filters — Languages, Channels and the date range — apply, with a Daily / Weekly / Monthly control on each over-time chart.

Engagement

How do users interact with the bot?

Two KPI tiles — Messages / conversation and Time to resolution — plus two charts: Conversation depth (a stacked area of 1 msg / 2 msgs / 3+ msgs, with a 100% view) and Interaction types (a donut with per-type counts). Depth is your best quick read on whether conversations are getting longer — and longer often means more effort, which drags CX down.

Users

Who are the users?

Four tiles — Total users, New users, Returning users, Convs / user — and a Users over time stacked bar splitting Returning vs New. The new/returning mix tells you whether you're serving first-timers or a base that keeps coming back.

Activity

When are users most active?

A heatmap of conversations by hour and day of week. Use it to line up staffing and handover coverage with the hours your users actually show up.

Sources

Where do users come from?

Starting URLs (where the widget was opened, with volume bars), two mini pie charts — Languages and Channels — and a Top user channels list with trend badges. This is the report to check before reading any segment: it tells you which languages and channels your traffic is really made of, which is exactly what you'll filter on elsewhere.