The Einstein Prediction Builder can help you predict which opportunities will close and which goods prospects will buy. Predictions may be created based on the data provided without having to write any code, and they will continue to score as your dataset changes. There's no need to go through the trouble of ETLing the data. It's a simple technique that allows you to make bespoke predictions about any custom or standard item with only a few clicks. It makes predictions by learning from previous examples. It helps you work smarter by spending your time on the proper things when you use projections. It predicts binary (Yes/No) and numerical answers to binary questions.
Creating Einstein Predictions:
- Go to Setup and Enter Einstein Prediction Builder.
- On the Einstein Prediction Builder, click the New Prediction Button and Give a Name to your Prediction and Select Next.
- Search the object to predict and Define a Segment if you want to focus on records based on your criteria else No for all records and click Next.
- As we're addressing the question "Will this Opportunity be Won?" We'll use the "Yes/No" prediction type and click Next.
- Choose Field if the object on which your forecast is based already has a field that answers your prediction question. It's fine if you don't have a field that answers your prediction question. Choose "No Field," and then use filters to set up your prediction on the next screen and Click Next.
- Add conditions to show Einstein examples of Yes And No values and Click Next.
- Include any fields that are appropriate. We recommend filling out all fields because you could learn something surprising and Click Next.
- Give the name of the field where the predictions will be saved and click Next.
- Review the Prediction and click on build Prediction and then Click Done.
- It is recommended that you "View Scorecard" before Enabling after the Status changes to "Ready for Review."
You'll be able to check the Prediction Scorecard after you've completed the Setup, which you can use to verify prediction quality and big-picture metrics. The Scorecard contains a lot of information.
The Prediction Quality indicator on the Overview tab shows how accurate the prediction is likely to be. It gives the forecast a score between 1 and 100. It is advised that the prediction be enabled only if the score is at least 60.
The Details tab is another important feature of the Scorecard, as it displays values for the Impact, Correlation, and Importance of all fields and values utilised in the Prediction configuration. These scores will aid us in the next step of the automation process. Let's go over the terms used in these fields.
- The scaled weight or importance of a predictor is represented by the impact, which is a value between 0 and 1.
- The relationship between a predictor and the field being forecasted, whether positive or negative, is known as correlation.
- The importance and weight of a prediction are indicators of its importance. The importance or weight of the prediction is provided depending on the model type used to make it, but not both.