---
title: "Analytics"
description: "See how many conversations your agents are having, how many finish, how long they run, and how deep each one goes, workspace-wide and for a single agent."
---

> Documentation Index
> Fetch the complete documentation index at: https://docs.speakai.co/llms.txt
> Use this file to discover all available pages before exploring further.

# Analytics

Analytics turns your agents' raw call history into numbers you can act on: whether conversations
are actually finishing, how long people stay on the line, and whether an agent is having real
back-and-forth exchanges or getting cut short. You'll find the same kind of numbers in two places:
rolled up across every agent on **Home**, and broken out for one agent on its **Analytics** tab.


## Why look at this

A conversation existing isn't the same as a conversation going well. An agent can pick up every
call and still lose most people in the first exchange, or run long without ever reaching what the
caller wanted. These numbers separate those two situations: totals and completion rate tell you
whether people are finishing, and the engagement funnel tells you whether the conversations that
do happen are shallow or substantial.

## The four numbers at the top

Both **Home** and an agent's **Analytics** tab open with the same four stat cards:

- **Total conversations**: every conversation counted, with a caption showing the change versus
  last week.
- **Completion rate**: the share of conversations that reached a completed status, shown as a
  percentage, captioned "Conversations completed."
- **Avg duration**: the average length of a conversation, captioned "Per conversation."
- **Avg satisfaction**: the average post-call rating, out of the ratings actually submitted. Until
  anyone rates a call it reads "No ratings yet"; once ratings come in, the caption shows how many
  ratings the average is based on.

On **Home**, these four numbers cover your whole workspace. Open a specific agent and go to its
**Analytics** tab, and the same four cards recalculate for that agent alone.

## Conversations over time

Below the stat cards, a chart plots conversations per day over the last 14 days, labeled
"Conversations - last 14 days." It's a quick way to see whether volume is climbing, flat, or
dropping, without opening a single conversation. Before an agent has taken any calls in that
window, the chart reads "No conversations in this period yet" instead of an empty grid.

## The engagement funnel: how deep conversations go

Completion rate tells you whether a call finished; it doesn't tell you whether anything happened
during it. The engagement funnel answers that by counting how many back-and-forth turns each
conversation had and sorting every conversation into one of four buckets:

- **Bounce** (0–1 turns): the caller left almost immediately, before a real exchange started.
- **Brief** (2–5 turns): a short exchange that didn't go far.
- **Engaged** (6–15 turns): a real back-and-forth conversation.
- **Deep** (16+ turns): an extended conversation with many exchanges.

A high bounce share usually means people are hanging up before the agent gets anywhere, worth
checking against how it opens the call. A healthy engaged-or-deep share means the conversations
that happen are substantial ones.

### Workspace-wide: the Agent engagement table

On **Home**, the **Agent engagement** section lists every agent as a row, with its Bounce, Brief,
Engaged, and Deep rates as percentages side by side, so you can compare agents at a glance.
Selecting a row opens that agent. An agent with no conversations yet shows "(no data)" next to its
name instead of rates. Until any agent in your workspace has had a conversation, the whole section
reads "Engagement breakdown appears once your agents have conversations."

### Per-agent: the full funnel breakdown

An agent's **Analytics** tab goes further, under the **Engagement funnel** heading. Two controls
sit above it:

- A **Date range** selector: **Last 30 days** or **All time**.
- A view toggle between **Bar view** and **Column chart view**, showing the same four buckets as
  horizontal bars or as vertical columns.

For each bucket, both views show the percentage of conversations, the count, and the average
duration of conversations in that bucket, so you can see not just how many calls were brief, but
how long a "brief" call actually runs for this agent. This section only appears once the agent has
had at least one conversation in the selected range.

## Related pages

- [Read Speak AI voice agent engagement metrics](/api/voice-agents/analytics/): the API reference
  for the overview and engagement funnel endpoints this page reads from.

Source: https://docs.speakai.co/help/voice-agents/analytics/index.mdx