Set up Predictive Insights using EPB

Configure Predictive Insights

Ensure that user-level and object-level permissions are configured for Einstein Prediction Builder (EPB).

Predictive scores are generated for your data in three stages: labeling, training, and scoring.

Labeling

Labeling starts automatically as soon as Einstein for Nonprofits installs. The Train Models? window pops-up after labeling is complete.

1. Navigate to the Einstein for Nonprofits app.

2. Click Get Started from the landing page.

3. Click the Train Models button on the Train Models? window.

   • The Train Models? window might not be visible immediately. You can see it after Einstein for Nonprofits has completed preparing your data for model training.

   • You must click the Train Models button to train the models for your org. Otherwise you will be prompted to do so repeatedly and won’t be able to generate predictions.

   • This action schedules a new Apex job which needs to be completed prior to the models displaying on EPB. The jobs are queued and might take some time to complete depending on which other processes are running in your org at that time.

Training

Training adds information to the models at the org level to help predict one-time donors, recurring donors and top donors.

1. Click the Einstein Platform tile to check if the models in EPB are in Pending state.

2. Wait for the models to transition to the Ready for Review state.

   • If the models failed because there is insufficient data with less than 400 rows, then use Backup Models to generate predictive insights instead.​
   • ​The models continue to display a ​Model Failed​ status until the batch jobs are successfully run again, even if additional data is added. By default these jobs remain in a failed state for 1 month.

Scoring

Scoring adds percentile probability scores to each individual Contact record.

1. Select Enable from the drop-down menu for each model in EPB to turn on scoring. It might take up to 24 hours for scores to be displayed on individual Contact cards. You can’t generate predictions until models are trained.

2. Learn about the required edition and permissions and considerations for setting up EPB to add intelligence to your apps.

You can edit EPB predictions after they are live. However, editing a prediction isn't recommended because the edits are overwritten at the time of the next release.

Display Predictive Insights

Einstein for Nonprofits generates predictions about the likelihood to become a first time donor and the likelihood to become a recurring donor for all constituents. However, we only generate predictions about the likelihood to become a top donor for those constituents who have donated in the last three years.

You can view predictive insights on the EPB dashboard or on the constituent’s Contact cards:

Note If a score is generated and then predictions are disabled, the prediction still persists on the Contact card. The likelihood of the insights being stale is high.

• View the results on the Einstein Prediction Builder scorecard after your prediction finishes building.

• View predictive insight percent probability scores on each constituent’s Contact card. Modify the page layout for the Contact object to show these scores by using Custom Lightning Web Components.