Measuring success of Predictive Intelligence
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
‎07-28-2024 10:19 PM
Hi, We have just implemented Predictive Intelligence to predict assignment groups and would like to measure the solution's effectiveness. Other than manually engaging agents and reviewing reassignment data from cases, are there any reports that can be leveraged?
I note that there is a Prediction Results dashboard, but I cannot seem to locate instructions on how to add/change the solution element.
- Labels:
-
Predictive Intelligence

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
‎07-29-2024 04:17 AM - edited ‎07-29-2024 04:18 AM
These are the best practices to gauge the effectiveness of Predictive Intelligence:
1. Build a baseline report, noting the re-assignment counts BEFORE Predictive Intelligence (PI) is put into production.
2. Build a report against the ml_predictor_results table. The table uses the following logic:
- The "predicted correctly field" which is set to true if the "predict output value" and the "final output value" of these two fields are the same.
- For example, IF PI predicts assignment group = database (predicted output value field) and the saved assignment group field on the incident table = database (final output value") THEN "predicted correctly field is set to true = true" which is where we assume the agent is ok with the prediction and that it is correct.
- IF PI predicted assignment group = database and the "final value field = software" THEN "predicted correct field" = false, thus the prediction is incorrect and the agent changed the prediction.
- You can build a report which tallies the predict correct fields in true or false categories to assess the effectiveness of PI.
3. Leverage the Predictive Analytics Dashboard. This is a PA dashboard and you can select the ml_solution and also modify the dashboard using Platform Analytics. This dashboard monitors the effectiveness of the PI solution and can indicate when you need retraining.
HTH -lener
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
‎08-07-2024 01:31 AM
Hi Lener, specific to bullet 3, is the predictive analytics dashboard an OOB item? Do we need to install anything to make it visible? I can only see the Prediction Results dashboard but am facing challenges when attempting to add solutions to the dropdown.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
‎08-22-2025 01:59 PM
Hello,
Did you get to know how to add solutions to the dropdown?
Thanks
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
‎08-12-2024 11:39 PM
Hi @Lener Pacania1 another question. For s/n, I assume there is an assumption in there where we set Top N to 1? I realized that if we set Top N to 3, there will be three values in "Predicted Output Value". That will thus create the situation where the data will not match for the flag to flip to True.