Analytics

Chats

A raw log of every conversation across every channel and the Playground: which channel it came from, when it started, how many messages it contains, and a preview of the last message. This is the ground truth the other two views are computed from — if a conversation doesn't show up here, it won't show up in Topics or Sentiment either.

Topics

Terms extracted automatically from real conversation text, ranked by how often they come up, shown as horizontal bars. The bar width uses a square-root scale rather than a straight linear ratio — at low volumes, a raw linear scale flattens everything next to one or two dominant terms into an unreadable sliver even when the underlying counts genuinely differ. Needs a handful of real conversations before anything meaningful shows up; a brand-new agent will show an empty state instead of fabricated data.

Sentiment

Every conversation is scored positive, neutral, or negative, with a summary count for each shown as three totals up top and a per-conversation list below (channel, date, label) beneath that. Like Topics, this is computed from actual message content — there's no manual tagging step, and nothing is scored until a conversation has actually happened.

How this relates to Activity

Analytics is about conversation content — what was discussed and how it went. For the operational side — working tickets, tracking who your contacts are — see Activity. The two draw from related but distinct data: a chat message and a Helpdesk ticket are different records, so a busy analytics chart doesn't necessarily mean a busy support inbox, and vice versa.