Got a question about your future business?
Enter that question as a prediction in Einstein Prediction Builder and it'll give you an answer. To answer this question, the application uses Machine Learning and historical data to predict the future.
Einstein Prediction Builder is a ready-to-use AI application for administrators. With just a few clicks, any administrator can create personalized predictions on any object in Salesforce.
Step-by-step explanation of how Einstein Prediction Builder works:
Step 1. Define the use case.
Choose a use case with high added value for your company, an item for which you are experiencing difficulties in terms of management or results, for example.
Then make sure that the item chosen for your prediction has enough information to give rise to a prediction. The more records the chosen item has, the more accurate the predictions will be.
Select KPIs that will be useful for measuring whether the prediction has had a positive impact.
Once your use case is clear, you need to transpose it into the language of Einstein Prediction Builder.
Note that the prediction you obtain will appear as a :
Binary (No/Yes)
Numerical
Step 2. Identify the data related to your use case.
To make a prediction, you need to choose an object in your Salesforce environment. This can be a standard object or a customized one.
Let's take a simple example: a company wants to know if an invoice will be paid late. To be able to give an answer to this question, Einstein will need to have examples of invoices paid late and examples of invoices paid on time. Thus, the object used for prediction will be the "Invoice" object.
Please note that the data used for prediction are all the records present in the Invoice object. If some records are not useful for your prediction, you can segment the data.
Step 3. Create your prediction.
Once your use case is clear, and the data required for it to run smoothly are known and selected, you can switch to Einstein Prediction Builder and let yourself be guided by the wizard screens. Creating a prediction will take no more than a few minutes.
Select the desired object (Invoice object in our case).
Select an answer type (Yes / No, here, to answer the question "Will an invoice be paid late?)
Choose between Field and No Field (Field = If a checkbox directly related to your prediction question exists on your object. And No Field if it doesn't).
Set up logical conditions so that the answer to the prediction is "Yes" or "No".
Choose the fields to include or exclude from your prediction.
Choose the field name in which the result of your prediction will appear.
Step 4. Review, update and activate your prediction.
With "View Scorecard" you can view the different results of your prediction. In particular, check the "Quality of your prediction" and the Top predictors.
Depending on the results obtained in View Scorcard, you can improve your prediction by :
Adding or removing data from the prediction.
Focusing on a smaller segment.
Updating your filters.
Adding relevant data from other objects via formula or summary fields.
Once you're satisfied, you can activate your prediction!
Thereafter, all new and updated registrations in this set will be re-evaluated hourly. The model will also be automatically readjusted every month, so you won't have to worry about obsolescence over time!
Step 5. Check your prediction.
Activating your model doesn't necessarily mean it's ready to be integrated into your business processes! In fact, we recommend that you let it run for a while behind the scenes, until you can evaluate its performance on new data.
Once your model has been running for a while on new data, you'll be able to tell whether it's really working by comparing your predictions with what's actually happened within your company. Using reports is an easy way to perform this analysis.
Step 6: Deploy and use your prediction
There are many ways to use a Prediction Builder prediction in Salesforce, here are a few examples:
Add the prediction field to a list and sort it by score. With the example we've chosen, the invoices most likely to be paid late will be listed at the top.
Add the Lightning Component Einstein Predictions to the Invoice page. This will give you useful information when you contact each customer.
Run automated workflows based on predictions, using Process Builder. For example, send a task to account managers a little in advance for invoices where payment is expected to be late.
We hope you've found this article about Einstein Prediction Builder useful, so don't hesitate to follow Cloud Ambiance's LinkedIn page so you don't miss out on our next publications. We'll be back soon with more Salesforce content to share!