Replace Deployed Models Easily

To expedite model improvements, a streamlined workflow makes updating deployed models simpler. If you create a version of a story that is associated with a deployed model, you must redeploy the model based on the updated settings. Previously, you deployed the new model segment, disabled the old deployed model segment, and enabled the new one. You can now simply replace the existing model with the latest settings. The model accuracy monitoring graph indicates when model updates occurred.

Where: This change applies to Einstein Analytics in Lightning Experience and Salesforce Classic. Einstein Analytics is available in Developer Edition and for an extra cost in Enterprise, Performance, and Unlimited editions.

How: Open the story version associated with the model you want to update. Change settings to improve it (such as field filters), and create a story version. Verify your changes by comparing story versions, and then compare the new and deployed models in Model Metrics to confirm model improvements.

Next, deploy the new model. In the deployment wizard, select Add or replace a model in an existing prediction, select the new model from the list, select Replace model, and select the deployed model you want to replace.

Model Versioning

Finish deploying the new model, then open it in Model Manager to view its accuracy graph. A vertical bar indicates when the model was updated.

Model Versioning

The bar appears as more statistics are collected. To trigger the display of this line, you can run a bulk score on the model.

Note

Note

To revert to a previous model, you can deploy the model from its earlier story version and replace the current model in production.