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

# Analytics

> Monitor your chatbot's performance — conversations, feedback, satisfaction, and token usage.

The Analytics tab gives you insight into how your chatbot is performing. Go to any chatbot → **Analytics** tab.

## Date Range Filter

All analytics data can be filtered by date range. Presets available:

* Today
* This week
* This month
* This year
* Last year
* **Custom** — pick specific from/to dates

## Overview Sub-Tab

The main dashboard with key metrics and trend charts.

### Metrics Cards

| Metric                            | Description                      |
| --------------------------------- | -------------------------------- |
| **Total conversations**           | Number of unique chat sessions   |
| **Total messages**                | Total messages sent (user + bot) |
| **Avg messages per conversation** | Average conversation length      |
| **Avg satisfaction rating**       | Average feedback score           |

### Charts

* **Conversations over time** — line chart showing daily conversation counts
* **Satisfaction trend** — line chart tracking average rating over time

## Feedback Sub-Tab

Detailed feedback analysis from the thumbs up/down buttons in the chat widget.

| Data                            | Description                                                         |
| ------------------------------- | ------------------------------------------------------------------- |
| **Thumbs up count**             | Total positive ratings                                              |
| **Thumbs down count**           | Total negative ratings                                              |
| **Feedback distribution chart** | Visual breakdown of positive vs negative                            |
| **Feedback list**               | Individual feedback entries with user's text comments (if provided) |

<Tip>
  Regularly review negative feedback to identify knowledge gaps. If visitors consistently rate certain topics poorly, add more content to those areas in your knowledge base.
</Tip>

## Conversations Sub-Tab

A table of all conversations with details:

| Column           | Description                                    |
| ---------------- | ---------------------------------------------- |
| **Date**         | When the conversation started                  |
| **Messages**     | User messages and bot responses in the session |
| **Duration**     | How long the conversation lasted               |
| **Satisfaction** | The rating the visitor gave (if any)           |

Use this to read actual conversations and understand what visitors are asking, how the bot responds, and where it falls short.

## LLM Sub-Tab

Token usage and cost analysis for the AI model.

| Metric                     | Description                                             |
| -------------------------- | ------------------------------------------------------- |
| **Input tokens**           | Total tokens sent to the model (questions + context)    |
| **Output tokens**          | Total tokens generated by the model (responses)         |
| **Token usage breakdown**  | Chart showing input vs output token distribution        |
| **Cost analysis**          | Estimated cost per conversation / per message           |
| **Model usage statistics** | If you've changed models over time, see usage per model |

## Using Analytics to Improve Your Chatbot

### Identify Knowledge Gaps

1. Go to **Conversations** tab
2. Filter for conversations with low satisfaction
3. Read the questions — are they about topics not in your knowledge base?
4. Add missing content to the Knowledge Base

### Optimize Costs

1. Go to **LLM** tab
2. Check average tokens per message
3. If token usage is high, consider:
   * Shortening the system prompt
   * Using a cheaper model tier
   * Adding more specific knowledge so the AI needs less reasoning

### Monitor Engagement

1. Go to **Overview** tab
2. Track conversations over time
3. Low engagement? Try:
   * Adding a proactive message to the widget
   * Improving welcome questions
   * Placing the widget on higher-traffic pages