Chat Analytics
The Analytics component of Converse provides information about the system usage. It allows you to view data concerning metrics revolving around engagement, user performance, user satisfaction & user statistics. These data can be filtered by several parameters, such as the time period, bot, or delivery channel.
Overview
The overview page shows the chatbot active users trend and chatbot performance. To navigate the Analytics page, select the 
button on the navigation bar.
Filter Panel
The analytics page shows data for ALL bots by default. To access the filter, select the 
button on the top right hand corner of the page. The filter panel will slide in from the right.
Filter by Bot: To filter the analytics data by bots, select the particular bot by expanding the dropdown panel. The dropdown panel will display all the bots you have permission for.
Filter by Delivery Channel: Filter the analytics data by Delivery Channel by selecting the relevant checkboxes.
Chatbot Usage Trend
This graph shows the data trend lines for:
New Users (Red Trend Line) - Unique first time users of the chatbot.
Returning Users (Blue Trend Line) - A returning user.
Total Users (Green Trend Line) - Total number of new users and returning users.
You can specify the date range to filter the data and the frequency for day, week, or month. You can also filter the bot and delivery channel used.
Chatbot Performance
The Chatbot Performance displays an overview of the top-level metrics that indicate the performance of the bot over the specified period in the Date Range Filter.
Changes in metrics are expressed in either percentages or the absolute value, with the change being obtained by comparing this period vs the last period. For example:
If the Date Range filter is set from 7 to 7 May (1 day), the period to compare with is 6 - 6 May (1 day)
If the Date Range filter is set from 8 to 14 May (7 days), the period to compare with is 1 - 7 May (7 days).
If the Date Range filter is set from 1 to 31 May (31 days), the period to compare with is 31 March - 30 April (31 days).
The following metrics are covered and shown as “cards” on the chatbot performance page:
Name
Interpretation
Description
Total Users
Higher is better
The quantity of users who talked with the chatbot (does not need to be unique users)
Chatbot Rating
Higher is better
The average chatbot rating received
Deflection Rate
Lower is better
The percentage of conversations where the user:
obtained a useful answer or;
had significant activity,
AND
the Escalate intent was not triggered
Total Fallback Count
Lower is better
The total number of instances where the system was not able to understand the user
Avg Session Duration
Higher is better
The average duration from Session Start to Session End
Overall Confusion
Lower is better
The percentage of interactions where the system was not able to understand the user
Engagement
The engagement component of analytics consists of data regarding how the user interacted with the bot. To access this page, select the 
tab on the left hand panel of the analytics page.
User Activities
This metric shows the overall usage in terms of total number of sessions (conversations) and messages sent by the user.
No of Sessions
The total number of sessions occurs in a given day. It is the sum of all conversations that was conducted between 0000hrs and 2359hrs for a given day.
Average Duration
The average duration is calculated as the Total Session Duration / No. of Sessions
Messages Sent
The total number of messages sent by users and the bot.
Deflection Rate
The deflection rate represents the percentage of conversations where the user obtained a useful answer or had significant activity, and the Escalate intent was not triggered. In essence, it signifies the rate at which a user question / query was fulfilled without needing the help of a live agent.
The deflection rate is calculated by comparing the conversations that had occurred in the current period against the conversations from the previous period. Hence, the deflection rate is dependent on the date range selected.
Metric
Description
Deflection Rate
Deflection describes the scenario where the user was able obtain to useful answer(s) from the chatbot, which may have resulted in him/her not asking the physical customer support helpdesk for help (Escalate was reached)
Deflection Rate % = ( No. of Helpful Conversations + No. of Conversations with significant activity ) / Total Conversations
No Activity
No. of Conversations with No Activity is the Total Conversations:
WITHOUT No. of Conversations that Escalated
WITHOUT No. of Helpful Conversations
WITHOUT No. of Conversations with significant activity
Escalation Enquiries Reached
Sum of total conversations in which there was at least 1 message that was escalated
Marked an Answer as helpful
No. of Helpful Conversations within a period is:
Sum of conversations in which there was at least 1 system message rated to be useful by users
Conversations
This component captures the conversation with users, alongside with metadata such as conversation ID, start, end, duration of conversation. Conversations can be filtered by clicking on the 
button. You can also mass export all conversation data as an excel spreadsheet by clicking on the 
button.
Viewing Individual Conversations
You can view the individual conversation history by clicking on the 
button under the action column for each conversation. A conversation history pop up will then appear, alongside its relevant metadata and export functionality.
Content Performance
The following metric cards are shown in the content performance page:
Name
Interpretation
Description
Total Fallback Count
Lower is better
The total number of instances where the system was not able to understand the user
Overall Confusion
Lower is better
The percentage of interactions where the system was not able to understand the user
No. of Message Sent by User
Higher is better
No. of Message Sent by User
Total No. of Answer Rated Helpful
Higher is better
Total No. of Answer Rated Helpful
Flow Diagram
This page shows the detailed content performance for a specific Intent in the form of a flow diagram. To access the flow diagram, click on the “View Flow Diagram” option.
The following metric cards are shown in the each individual flow diagram page:
Name
Interpretation
Description
Times Asked
Higher is better
The number of times this Intent was triggered
Fall Completion Rate
Higher is better
The percentage of instances where the user reached the final state of the Intent's dialog flow
Fallback Count
Lower is better
The number of occurrences where the user said something unexpected while progressing within this Intent's dialog flow
Confusion Rate
Lower is better
The percentage of interactions where the system was not able to understand the user, while handling this Intent
Positive Rating
Higher is better
The number of instances where a user marked an answer within this flow as correct by clicking on the "tick" icon
Interpreting the Flow Diagram:
Name
Image
Description
State Cards
Shows the number of times a particular subsequent state was triggered.
Exit Reason Popover
Shows the number of drop offs. Drop Offs are instances where the user did not continue using the chatbot (and allowed the session to expire) immediately after this State.
State Card Popover
Income: The number of users who reached the state.
Outcome: The number of users who completed the state.
Drop offs: The number of users who dropped off from the state.
Flow Lines
Shows the number and percentage of users moving from one state to the others.
User Satisfaction
Name
Interpretation
Description
Average Rating
Higher is better
The average rating score based on the feedback given via the Feedback “stars” rating (1-6)
Last updated
Was this helpful?