Einstein Discovery enhances business intelligence by leveraging statistical modeling and machine learning to uncover, highlight, and present insights derived from your business data. This article is the second part of Overview of Einstein Discovery Part 1 – Mortgage Banking Technology | Origination Strategy Consulting (takefiveconsulting.com) which covers the steps of deciding the target outcome, preparing data and creating a model.
4. Evaluate Model
The model can be evaluated in the next stage. In the Model page one can evaluate for model performance, Review model accuracy, Set a threshold value to classify the prediction as either true or false in case of binary classification models. Einstein Discovery performs a model quality check and surfaces detected issues in the Assess Deployment Readiness panel of the model page. The ‘Training Data and the Model panel’ displays the distribution of the observed values. Ie. How many TRUE and FALSE values are in the training data. The ‘Top predictors’ shows the predictor variables with the highest correlation to the outcome.
5. Explore Insights
The model summary shows the model goal, the total rows analyzed, the average outcome, and any difference from the previous version. The Explore a Feature button toggles the variables panel.
The variables panel shows the list of explanatory variables in your model and their correlation to the model outcome, starting with the highest correlation. The higher the correlation, expressed as a percentage, the stronger the statistical relationship.
The insight summary panels show the top positive and negative impacts on the outcome variable.
6. Deploy Model
We can deploy the model from the Model performance Overview tab. The Einstein Discovery deployment wizard walks us through the steps to deploy the model. In Einstein Discovery, every deployed model belongs to a container object called a prediction definition. This screen prompts you to create or update models and prediction definitions, as well as to change the model name. A prediction definition can contain multiple models in which each model produces predictions for a different segment (subset) of your data.
7. Predict and Improve
- Prediction: It’s a forecasted value generated by the model, representing a potential future outcome based on the analyzed data.
- Predictors: These are the variables within the dataset that influence the predicted outcome. Top predictors are those that contribute most significantly to the outcome.
- Improvement: An improvement suggests actionable steps that users can take to enhance the predicted outcome. These are linked to variables that users can potentially control, like adjusting the shipping method or altering a subscriber’s membership level.
Once a model is deployed, it can be utilized to retrieve and exhibit predictions and improvements for both standard and custom objects on Lightning Experience record pages. Additionally, after deployment, Einstein Discovery models offer predicted outcomes and suggested improvements through simple clicks, eliminating the need for coding. This functionality can be accessed in various ways:
- Lightning record pages
- Experience Cloud sites pages
- CRM Analytics datasets using Data Prep recipes (predictions and improvements) and dataflows (predictions only)
- PREDICT function used in process automation formulas
- Salesforce flows (with Flow Builder)
- Tableau flows, dashboards, and calculated fields
This is the final part of Overview of Einstein discovery. To learn more, click Introduction to Einstein Discovery (salesforce.com).
Take Five Consulting is a technology company, based in Virginia U.S., that specializes in the Mortgage Banking vertical especially LOS implementation and application development. Take Five Consulting creates and implement mortgage technology and software specifically for Mortgage Industry.
