Define and Build a Prediction
Let Einstein Predict for You
Now that you have your special Developer Edition org, you’re ready to take the steps in this project to create a prediction. Let’s begin by defining a prediction.
The first step in crafting a prediction is to determine the question you want to answer. In this case, it’s “Is this person likely to want to join VeloStars?”
- Click and choose Setup from the drop-down menu.
- Click Get Started on the Einstein Prediction Builder tile. If it’s not visible, type Einstein Prediction Builder into the Quick Find box. If it’s your first time using Einstein Prediction Builder, click Get Started again on the Welcome screen. You may have to wait a short time while it sets up. Give it a minute or two and then refresh the page.
- Now you define the prediction you want to make. From the Einstein Prediction Builder Setup page, click New Prediction. This brings up the Einstein Prediction Builder wizard.
- Name your prediction VeloStars Membership Likelihood. Click Next.
- Select Contact from the list of objects to predict and ensure that No segment is selected. Click Next.
Choose Yes/No as the kind of question you want to ask. Click
Einstein Prediction Builder can answer questions in a numeric or yes/no format (Booleans for you mathy types). In this case, you want to ask a yes/no question: Is this person likely to become a VeloStars member? The response to that question comes in the form of a number, specifically a probability between 0 and 100. (A numeric question might be “How many family members is this customer likely to sign up?”)
For the field to predict, choose VeloStars Member. VeloStars Member
is a field on the Contact object.
Select and configure fields to tell Einstein which records to use as examples.
For Include Records That, choose Meet All Conditions. Then
- Field: First Purchase
- Operator: Less than or equal
- Type: Date
- Value: Jan 1,2018
We assume that anyone who’s been a long-time customer has either joined VeloStars by now or never will, so you limit your outreach to customers whose first purchase occurred on or after January 1, 2018.
- For Include Records That, choose Meet All Conditions. Then select:
- Click Check Data. It should return a valid record count. Click Next.
- You want to use all fields for the prediction, so leave all fields selected. Click Next.
- Create a custom field for the prediction called Membership Likelihood and click Next.
Now that we’ve created (that is, defined) a prediction, it’s time to let the system
build and assign it. Confirm that everything looks good, then click
Check Data to make sure you have enough records to work with. If all
looks good, click Build Prediction. And on the next screen, click
Normally, building a prediction can take anywhere from a couple of hours to a full day or more. In this case, we’ve created predictions for you so you don’t have to wait too long (generally 15-20 minutes should be enough).
When the prediction finishes, Einstein sends a notification, and the status of prediction changes to Ready For Review. (You may need to refresh the page for the Status field to update.)
Click the down arrow next to the status of your prediction and choose
Enable from the dropdown menu. The prediction’s Status changes to
Enabled, and you see a message saying that Einstein will begin scoring (that is, assigning
predictions to your Contacts).
If you’re so inclined, click the down arrow again and choose View Scorecard to see how your prediction rates.