# Build Your First Agent with Rasa Copilot

Rasa Studio ships a built-in Analytics dashboard which gives you a clear view of how your assistant performs over time. It helps you answer questions like:

- Are users getting the help they need?
- Is the assistant faster or slower after a release?
- Which flows are working well, and which may need improvement?

This page explains what you’ll see on Analytics, how to interpret the numbers, and how to use filters and insights to take action.

## Metrics at a Glance

**Conversation Volume:** The total number of conversation sessions between your assistant and users within the selected time period.

Each session is evaluated against your success criteria. These are the conditions you define to determine whether a conversation was successful. For example, you might count a session as successful if the user completes a booking flow without being handed off to a live agent.

You can configure these criteria in Settings > Success Criteria so they reflect what success means for your use case.

> **Note:** By default, any session that is resolved without reaching a human agent is counted as successful. Changes only apply to new sessions going forward and do not retroactively affect historical data.

**Containment:** A contained conversation is one where the user's request was handled end-to-end by the assistant, no escalation to a human.

To measure containment, your assistant must have `action_human_handoff` configured. This is the action Rasa uses to detect when a conversation is escalated to a human agent.

**Latency:** The average time it takes for your assistant to respond to a user message.

**CSAT Customer Satisfaction (CSAT):** The percentage of conversations where users provided positive satisfaction feedback. CSAT collection is handled automatically: `pattern_customer_satisfaction` triggers after `pattern_completed` once the user indicates they are done, prompting them to rate their experience.

CSAT mapping:

If a CSAT scale from **1 to 5** is used, ratings are grouped as follows:

- **Satisfied:** ratings **3–5**
- **Dissatisfied:** ratings **1–2**

Conversations where no rating is submitted are categorized as **No response**.

## Analytics Overview

The Analytics dashboard gives you a single view of your assistant's health across four key metrics: conversation volume, escalation, latency, and CSAT. Each metric is visible at a glance at the top of the dashboard, with charts and insights below that help you dig deeper.

## Conversation Volume

The conversation volume chart displays all conversations within your selected time period, divided into successful and unsuccessful conversations.

### Viewing Conversations from the Chart

You can go directly from the Conversation Volume chart to the individual conversations behind the data. Click on any bar in the chart to open a filtered view of the conversations that make up that data point.

This lets you:

- Review successful conversations to understand what's working well.
- Investigate unsuccessful conversations to identify where users are dropping off or getting stuck.
- Spot patterns across conversations for a specific day or time period without manually setting up filters.

### How to Define Success:

To configure what counts as a successful conversation:

**Open the configuration panel:** Click the ⚙️ settings icon on the Conversation Volume chart.

**Add filters to define success:** Use the available filters to specify what success looks like for your team.

## Escalation Insights

Escalation Insights surfaces a ranked list of the flows with the highest escalation rates — the ones most often resulting in a handoff to a human agent.

## Latency Insights

Latency Insights breaks down your assistant's response times across the selected period into three percentile trend lines.

## Filters

Applying filters to the Analytics dashboard helps you compare how different assistant versions perform, isolate channel-specific issues, or zoom into a specific week to investigate a spike.

| Filter Name | Filter Description |
| --- | --- |
| Time period | Select the time window for your analytics data. Choose from **Last 7 days**, **Last 30 days**, or **Last 90 days**. |
| Assistant Version | Filter by the version of your assistant. Compare performance across deployments to measure the impact of changes between releases. |
| Channel | Filter by the channel where the conversation happened. Isolate metrics for chat or IVR. |

## Direct Data Access

You can access the Analytics database and pipe it into your own dashboards. Or connect directly to keep all your reporting in one place.