You are here:
Monitor Change Risk Prediction Card
Forecast the risk of a change request failing by using machine learning to generate a score from 0 through 100. Use these data-driven scores to replace subjective assessments with a consistent process that improves IT stability.
Required Editions
| Available in: Lightning Experience |
| Available in: Enterprise, and Unlimited Editions with Einstein for IT Services add-on, and AI Accelerator for IT Services add-on. |
- From the App Launcher, find and select Change Requests.
- From the Change Requests page, open a record.
-
In a change request record, navigate to the Change Risk Score card in the side panel.
The Change Risk Score card displays the probability that a change request will fail.
- Review the predicted score to determine the risk level. For example, a score of 33.35 indicates a moderate risk of change request failure.
-
In the Top Predictors list, identify the factors influencing the score. Common predictors
include:
- Subject and Description: Keywords like password or rejecting that correlate with historical patterns.
- Priority and Urgency: The impact of related incidents or current urgency levels on the final score.
- Historical Trends: Factors such as the percentage of past incidents breached by the current owner.
-
Use the feedback icons to help refine the machine learning model:
-
Select the Thumbs Up icon if the prediction is accurate.
- Select the Thumbs Down icon if the prediction is incorrect.
-
Example:
Sarah, a Change Manager at Cumulus Bank, oversees critical updates to banking applications. When a database schema migration for real-time transactions is scheduled, a change request is created with these details.
- Change Request Number: CHG-4055
- Subject: Database Schema Migration for Real-Time Transactions
- Category: Database
Sarah opens the record and sees the Change Risk Score card with a score of 72.45. This score indicates a high probability of failure for the planned migration.
Sarah reviews the risk factors identified by machine learning. Predictive AI identifies that the Subject and Description fields contain keywords such as Migration and Schema, which historically correlate with higher failure rates. Additionally, the historical trends predictor shows that an emergency change has higher likelihood of a failed change.
To maintain IT stability, Sarah postpones the migration by 24 hours for sandbox testing and assigns an expert reviewer. Using data analysis from the Change Risk Score, Sarah follows a consistent process that prevents a service outage.

