Triage Insights enhances how customers analyze their triage models. It consolidates all model and prediction data from both legacy and new Triage models onto a single page. This integration allows users to gain valuable insights into model performance and manage tickets more effectively. This feature is available to all customers with Triage who have admin access to the Forethought dashboard. To access Triage Insights, simply navigate to Triage > Insights.
Use Case
A customer has recently launched a new Triage model. After conducting initial tests, they aim to determine whether the model's performance meets their expectations or if adjustments are needed. To achieve this, they used Triage Insights. With Triage Insights, the customer was able to easily analyze metrics such as the total number of tickets processed, prediction accuracy, and any data discrepancies. This capability empowers customers to make informed decisions about necessary adjustments to the model based on their real-time data.
Key Features of Triage Insights
-
- Unified Analytics Dashboard: Triage Insights brings together both legacy Triage models and new Triage LLM models into a single interface. This integration simplifies the analysis process by allowing users to view all relevant data in one place.
- Funnel Data Visualization: Triage Insights includes a detailed funnel view that tracks ticket metrics over selected time periods. Aside from funnels, users can also visualize data through a time series chart, making it easier to understand trends and performance fluctuations.
-
Comprehensive Metrics: Triage Insights contains the following metrics:
-
- Total tickets
- Eligible tickets
- Predicted tickets
- Finalized tickets
- Correct predictions
-
- Detailed Ticket Breakdown: Users can access a data table that breaks down the triage funnel by labels. This allows for in-depth analysis of how different categories are performing.
- Interactive Ticket Details: By clicking on specific labels, users can delve into detailed ticket information. The Ticket tab features a robust filtering system that enables users to sort tickets by various criteria, including ticket status, model name, predicted labels, ground truth labels, correct predictions, reasons for no prediction, and whether they were written to a helpdesk.
Metrics (Models Tab)
To view the metrics in the Models tab, go to Triage > Model > Labels/Tickets.
| Location | Metric name | Definition |
| Labels | Total tickets | Total number of tickets that Forethought has access to. |
| Labels | Eligible tickets |
Tickets that bypass the filters configured on the Help Desk, preventing Triage from assigning labels. Common filters includes the following: i) Tickets coming from a non-supported channel, such as voice ii) Tickets from a specific customer that are to be excluded from triage iii) Internally created tickets that do not need a prediction |
| Labels, Tickets | Predicted tickets | Tickets that Triage successfully predicted and labeled. |
| Labels | Pending | Ticket with predictions but haven’t marked as closed or resolved yet. |
| Labels | Finalized tickets | Predicted tickets which are close or resolved. |
| Labels, Tickets | Correct labels | Total number of correct, finalized labels. If agents did not override a label and the corresponding ticket is closed or resolved, that label’s Triage prediction is considered accurate. |
| Labels | Accuracy | This indicates the accuracy of Triage's label predictions. The calculation formula is Correct Predictions/Finalized tickets x 100 |
| Tickets | Accuracy | The percentage of correct predictions out of all finalized tickets. Calculated as: (Correct Predictions / Finalized Tickets) X 100 |
Metrics (Insights Tab)
To view the metrics in the Insights tab, go to Triage > Insights > Models/Tickets.
| Location | Metric name | Definition |
| Models | Total tickets | Total number of tickets that Forethought has access to. |
| Models | Eligible tickets |
Tickets that bypass the filters configured on the Help Desk, preventing Triage from assigning labels. Common filters includes the following: i) Tickets coming from a non-supported channel, such as voice ii) Tickets from a specific customer that are to be excluded from triage iii) Internally created tickets that do not need a prediction |
| Models | Predicted tickets | Tickets that Triage successfully predicted and labeled. This number may exceed the metrics below if Triage cannot write to the Help Desk due to misconfigurations or access issues. |
| Models | Finalized tickets | Predicted tickets which are close or resolved. |
| Models | Correct predictions | Total number of correct, finalized labels. If agents did not override a label and the corresponding ticket is closed or resolved, that label’s Triage prediction is considered accurate. |
| Tickets | Eligible tickets | This metric represents the number of customer support tickets Triage is authorized to view based on the model’s active date, your chosen time window, and the Triage filters. |
| Tickets | Finalized labels | The total number of labels finalized and recorded in your helpdesk. If a single ticket is assigned multiple labels (for example, by different models), each label is counted separately. |
| Tickets | Correct labels | Total number of correct, finalized labels. If agents did not override a label and the corresponding ticket is closed or resolved, that label’s Triage prediction is considered accurate. |
| Tickets | Accuracy | This indicates the accuracy of Triage's label predictions. The calculation formula is Correct Predictions/Finalized tickets x 100. |