Introduction: Analyze overview provides insight into the performance of your chatbot. In one overview, there are graphs showing the number of sessions, conversations, top intents, and top URLs. With these graphs, you can easily identify metrics such as the number of sessions, conversations, and top intents.
Table of Contents:
Definitions
- Session: widget-chatbot is open or chatbot initiates a conversation.
- Conversation: there is at least 1 interaction from the end user, the chatbot received a message to process.
Analyze > Overview
The various graphs in the analysis overview are organized based on the selected date range and can be sorted by hour, day, week, month, or year. Exporting to CSV provides a CSV file with the number of sessions for the selected date range. Each of the graphs has a tooltip (the icon below) providing an explanation of what is shown in the graph.
Export
You can export each graph to CSV. This export provides a CSV file with, among other things, the number of sessions for the selected date range.
Explanation per graph
-
Sessions / Conversations: This chart shows the number of sessions and conversations over time that your chatbot may have handled. Important to note: a session represents an open chat. When someone opens a session and actually starts a conversation, it becomes a conversation.
-
Drill-down function: Clicking on the Conversations allows you to navigate directly to Analyze > Conversations to analyze the Conversations.
- Conversation per device type: This graph shows the number of conversations over time, broken down by device type. This provides details on how many conversations your chatbot handles on different devices.
- Top URL's for start conversation: This chart shows the top 20 URLs for conversations. At the beginning of a conversation, a URL is captured. URLs displayed in the report have a maximum page path of level 3. Example: www.example.com/pagepath1/pagepath2/pagepath3. Download the CSV export to view all available page path levels.
-
Messages - (Overview of the intent handled by the bot): This chart shows the number of messages over time.
- Handled: number of messages where the bot sent at least one response.
- Unhandled: number of messages not recognized by the bot. The bot is not programmed to process the message.
- Unanswered: Number of messages not understood by the bot. These messages need to be reviewed, and the bot needs to be trained to understand these new intents. Detailed analysis is possible in Analyze > Messages.
- Handed over: number of messages handled by an agent (transferred to Engage).
Note! The status 'handed over' does not apply to a WhatsApp bot. The WhatsApp bot functions as a Spotler Engage agent, and the conversation is immediately visible in Spotler Engage.
The status 'handed over' is only applicable when the conversation is genuinely managed by the Aigent (not WhatsApp) and subsequently forwarded to Spotler Engage, where a human agent takes over. However, in the case of WhatsApp, the Aigent operates as a Spotler Engage agent.
- Top intents - (triggered by open question): This graph shows the top 20 intents recognized by your chatbot. This graph only shows all intents for messages that are a response after an open question from the chatbot.
- Top Intents - (triggered by user node or closed question): This graph shows the top 20 intents recognized by your chatbot. Only intents recognized after a closed question (i.e., user turn) are shown.
- # handovers initiated: This graph shows the number of initiated handovers. Started handovers occur when chat users indicate they want to speak with a live agent via live chat. Note that multiple handovers can be initiated from one session.
Analyze > Conversations
It is possible to filter the Conversations by Open and Archived.