Skip to main content

Session Overview

The Session Page in the Peki Chatbot Platform provides detailed information about each chatbot interaction. It is designed to help teams analyze conversations, measure chatbot performance, and understand user behavior. This guide explains the key elements of a session and how they can be used for insight and improvement.

What is a Session?​

A session refers to a single conversation between a user and the chatbot. It begins when the user initiates a message and ends when the interaction is completed or inactive for a set period. Each session is tracked to gather data about user behavior, chatbot performance, conversation topics, and outcomes.

Sessions are crucial for analyzing chatbot performance, as they provide key indicators such as user sentiment, conversation topics, and token usage.

Key Session Metrics and Details​

1. Topic and Summary​

  • Topic: Automatically detected theme or category of the conversation (e.g., "Greeting", "Product Inquiry").
  • Summary: A brief, auto-generated description of the session's key points.

These details help you quickly understand the purpose and outcome of the conversation without reading the full message log.

2. Sentiment Analysis​

  • Sentiment: The platform detects the emotional tone of the conversation (e.g., positive, neutral, or negative).
  • Behavior: Behavioral insights indicate whether the user was calm, urgent, or confused based on their message patterns.

Sentiment analysis can help you assess customer satisfaction and detect areas where the chatbot might need improvement.

3. Status and Progress​

  • Status: Indicates whether the session is active, completed, or unresolved.
  • Result: Tracks the progress of the session, such as whether the issue was resolved or still pending.

4. Token Usage​

Tokens refer to small units of text, including words and punctuation. Every message in the conversation consumes tokens, which helps track the chatbot's resource usage.

For example:

  • The message "Hello, how can I help you?" may count as several tokens.

Monitoring token usage ensures your chatbot stays within the limits of your subscription plan and helps optimize prompt design for efficient responses.

5. Platform and Timestamps​

  • Source: Identifies where the conversation took place (e.g., Web, Mobile).
  • Session Created: The date and time when the session began.
  • Last Updated: The most recent timestamp indicating when the session was updated.

These timestamps help track user activity and engagement patterns.

How to Use Session Data for Analysis​

  1. Identify Common Topics
    Review session topics to understand what users frequently ask, which can guide chatbot training.

  2. Monitor Sentiment Trends
    Detect recurring negative sentiment to identify areas where the chatbot may need better responses or improved tone.

  3. Optimize Token Usage
    Analyze sessions with high token counts to ensure that prompts and responses are concise and efficient.

  4. Measure Completion Rates
    Track session statuses to understand how often the chatbot resolves issues on its own versus needing human intervention.

FAQs​

  1. How can I view session data?
    Navigate to the Messages Page, where sessions are listed along with their topics, summaries, and key metrics.

  2. What should I do if token usage is too high?
    Review the chatbot's prompts and responses to make them more concise and efficient.

  3. How does sentiment analysis work?
    The platform evaluates the tone of user messages to determine if they are positive, neutral, or negative. Sentiment data can help identify user satisfaction levels.