The Topic Insights page sorts your data into topics and sub-topics, offering a comprehensive look at the range of topics the chatbot engages with your users. This structured organization allows in-depth analysis and easier decision-making based on robust data insights. You can find this feature within the Widget Insights tab.
Note: Topic Insights is only visible to customers with Discover enabled.
Why is this important?
When analyzing user queries and chat interactions, you can use Topic Insights to identify common themes and areas where users frequently seek information. This detailed categorization allows you to enhance the knowledge base articles and improve the overall workflow by addressing specific user needs and optimizing content delivery based on the identified topics.
How Discover topics are generated?
Forethought's Discover analyzes customer queries and organizes them into distinct topics and sub-topics based on thematic consistency. By thoroughly examining both the subject line and the content of each support ticket, Discover effectively identifies the key topics and themes within the queries.
Access Topic Insights
- Click Solve > Widget Insights > Topics.
Topic dashboard metrics
| Metric name | Definition |
| Chats with Topics | Shows the number of deflections and non-deflections during a period, along with potential and realized savings. |
| CSAT for chats with Topics | Displays the average Customer Satisfaction (CSAT) score of all chats, including the total number of responses and the breakdown of responses by rating. |
| Quick feedback | Displays the percentage of positive feedback from users, the total number of responses, and the number of users who did not provide feedback when articles surfaced. |
| Relevance for chats with Topics | Shows the number of relevant, somewhat relevant, and irrelevant chats with Topics. The relevance rating uses AI to determine how well the response directly addresses the user's query. Non-deflected chats and some historical chats won't have a relevance rating. |
| User engagement for chats with Topics | Shows the percentage of chats with Topics where users were engaged. It also displays the number of chats where users were engaged throughout the chat or dropped off midway. |
| Topics | Topics are categorized subjects that group similar inquiries and responses. To ensure accurate topic categorization, chats should contain sufficient dialogue length. The percentage of chats for which topics are created is calculated by dividing the sum of chats by the sum of non-deflections and deflections. |
| Deflections | The number of deflected chats within this topic. |
| Non-deflections | The number of chats within a topic that were not deflected. |
| Average CSAT Survey Score | The average CSAT scores of chats in which the chat topics were detected. |
| Top 3 surfaced workflows |
Most surfaced workflows found in the chats under a selected topic. |
| Top 3 surfaced articles |
Most surfaced articles found in the chats under a selected topic. |
| Realized saving | Dollarized savings assuming a $15 cost per deflection. |
| Potential saving | Maximum potential savings you could have realized for this topic with Discover automation. |
View Sub-topics and Topic details
- To view the sub-topics, click the dropdown arrow to expand and view the sub-topics.
- Click on any row for a more detailed view of the data.
- Upon clicking, a line graph displaying metrics related to the topic will appear.
- You can also export data by clicking the export icon.
Chats metrics
| Metric name | Definition |
| Detected workflows | Workflows or intents that have been detected in the selected chat. |
| Surfaced articles | List of articles surfaced in selected chat. |
| Quick feedback | This reflects the Quick Feedback from all articles used in this chat. |
| Relevance | The relevance rating uses AI to determine how well the response directly addresses the user's query. Non-deflected chats and some historical chats won't have a relevance rating. The relevance rating can be:
|
| User engagement |
It indicates whether the user stayed engaged throughout the chat or dropped off midway.
|
Knowledge articles
| Metric name | Definition |
| Deflections | Number of chats deflected where this article surfaced. |
| Surfaced & clicks |
The number of times the chatbot offered articles from the knowledge base in response to a user’s query. If multiple articles are used in a single chat session, each instance will be counted separately. |
| Avg. CSAT | The average Customer Satisfaction (CSAT) scores from chats where the article surfaced. |
| Quick feedback | Direct feedback received from users about the article’s usefulness when it surfaced. |
Export data
- Click the export icon.
- Receive the CSV file in your email.
- Click Download CSV.
Note: You have 48 hours to download your file. If you miss this window, click the export icon again in your Forethought dashboard to get a fresh link.