Triage Analytics Dashboard

What is it?

Triage Analytics is a tab in your Forethought dashboard that is accessible to you after Forethought begins triaging your organization's support tickets.  This dashboard will allow you to see into the performance of each AI classifier that is automating your ticket tagging process.  In this dashboard you see the following, broken down by predicted value:

• Number of support tickets we have made predictions on
• Coverage (percentage of tickets we are making predictions on)
• Accuracy of those predictions

Why is this important?

We know that support agents are focused on addressing and resolving customer issues as quickly as possible. Unfortunately, they’re stuck doing mundane tasks like manually prioritizing, categorizing and routing cases.

With Triage, we are able to automate these tasks.  Through the dashboard, Forethought can provide analytics as to how many tickets are automatically and accurately classified.

Using Forethought's Triage Analytics Dashboard

How to get there

On Forethought's dashboard, under Triage, you can click on Dashboard to view Triage analytics. If you do not see this, please contact Forethought support. 

Analytics Breakdown

You have multiple options to view Triage analytics:

• Select a date range
  
  • Last 7 days
  • Last 14 days
  • Last 30 days (Default)
  • This month
  • Previous month
  • Custom
• Select table view or chart view to visualize the classifier metrics
• Select the classifier you want to view the metrics for. 

In addition, when the table view is selected, you can download the metrics in a csv file by clicking on the share button. 

Table View

Once the date range and the AI classifier are selected, in the table view you see 5 columns. Each row shows the unique classifier outputs corresponding to the classifier selected.

• Model name/values (column 1)
  • This column will have the name of each value the selected triage classifier is predicting.
    • For example, when using the spam classifier the table rows will consist of 3 rows
      1. Total and Averages
      2. Spam
      3. Not Spam

• Actual
  • For each row, this is the total number of tickets Forethought received from the help desk which should receive the value from column 1.
    • For example, when using the spam classifier the Actual column of the Spam row will contain the number of tickets that are actually spam.

• Predicted
  • For each row, this is the total number of tickets that Forethought classified as the value from column 1. 
    • For example, when using the spam classifier the Predicted column of the Spam row will contain the number of tickets that Forethought labeled as spam.
    • Note: Sometimes the predicted count can be smaller than the actual count. This can be because 
      • EITHER: The classifier predictions were not very confident
      • OR: Some custom filters have been applied which prevent Forethought from classifying all tickets. To learn more about your organization's customizations, please contact your CSM.
• Coverage

  • For each row, this is the percentage of tickets that Forethought classified as the value from column 1 out of the actual number of tickets that ended up being categorized this way.

  • Coverage = (# of predicted tickets / # of actual tickets) * 100

• Accuracy
  • For each row, this is the percentage of tickets that Forethought accurately classified as the value from column 1, out of all tickets that Forethought predicted to have this value. In other words, it's the percentage of tickets in the Predicted column where the prediction was correct.

Chart View

Once the date range and the AI classifier are selected, in the chart view you see a graph. A dropdown on the left side of your screen will allow you to see charts for the different metrics described above in the Table View section.

How are these number calculated?

There are 2 different scenarios based on the Triage implementation. To learn more about your specific Triage implementation, please contact your CSM.

• If Forethought is writing our predictions to a clone field and not your production field, analytics are calculated by comparing the value of the clone field against the value of the production field. For example, if you have a Product field with business logic attached, but Forethought is configured to write predictions to a clone field called FT_Product, analytics are calculated by comparing the values of these two fields.  If FT_Product does not match Product, then the prediction is considered to be incorrect.
  
  
• If Forethought is writing to a production field or a field that can't be directly compared with another field, analytics are calculated when an agent changes the prediction that Forethought wrote to that field.  For example, consider a configuration where you have a Product field with business logic attached, and Forethought is configured to write predictions to this same field.  If Forethought writes Product_A to this field but an agent notices that the ticket is really related to Product B, the agent should change the field value to Product_B to ensure business logic executes the expected actions. After the ticket has closed, the current field value is compared against the original prediction value, and since these do not match, the prediction is considered to be incorrect.

Please reach out to support@forethought.ai if you have any questions!