Searj

Conversation Reports

Conversation reports let you review all conversations your bot has had with visitors. Use them to understand customer needs and improve bot performance.

Viewing Conversations

To access conversations for a specific bot, follow these steps:

  1. 1Open the dashboard and click on the desired bot
  2. 2Navigate to the "Conversations" tab
  3. 3You'll see a list of all conversations sorted by date

Conversation List

The conversation list displays the following information for each conversation:

ColumnDescription
TitleConversation title automatically derived from the first message
Message CountTotal number of messages in this conversation
DateDate and time when the conversation started
LanguageThe detected language of the conversation (Arabic or English)

Conversation Detail

Click on any conversation in the list to view the full message history. You'll be able to see:

  • All user messages and bot responses in chronological order
  • Timestamps for each message
  • Clear distinction between user messages and bot responses

Deleting Conversations

You can delete individual conversations that you no longer need:

  1. 1Open the conversation you want to delete
  2. 2Click the delete button (trash icon)
  3. 3Confirm the deletion in the confirmation dialog

⚠️ Warning: Deleting a conversation is permanent and cannot be undone. All associated messages will be deleted.

CSV Export

You can export analytics data to a CSV file for external analysis or sharing with your team:

  1. 1Navigate to the bot's analytics page
  2. 2Click the "Export CSV" button
  3. 3The file will be automatically downloaded to your device

Understanding Conversation Flow

It's important to understand how conversations work in Searj to interpret the data correctly:

  • New Conversation Per Visitor Each new visitor starts a separate conversation with the bot
  • Persistence Across Refreshes Conversations persist even if the visitor refreshes the page, via the session ID
  • Response Generation Bot responses are generated using RAG (Retrieval Augmented Generation) which searches training content to find the best answer
  • Returning from Human Handoff If a conversation waits for a human support agent without a reply, Searj automatically returns it to the AI agent after 30 minutes. The customer can also return immediately by typing "back to AI".

Identifying Improvement Areas

Use conversation reports to identify patterns in questions the bot struggles to answer:

Improvement Steps:

  • Review recent conversations regularly
  • Identify recurring questions the bot doesn't answer well
  • Add new training content covering these topics
  • Verify response improvement after adding new content