Einstein is the new generation of AI available directly within Salesforce CRM data. It learns from existing data and can provide predictions, insights, next steps, and recommendations. So, optimize your time by focussing on opportunities that are more likely to convert rather than calling the ones that are unlikely.
Features of Salesforce Einstein:
- Sales Cloud Einstein
- Service Cloud Einstein
- App Cloud Einstein
- Community Cloud Einstein
- IoT Cloud Einstein
- Marketing and Analytics Cloud Einstein
Einstein products which are formed by the incorporation of one or more smart assistant components. They are:
- Einstein Bots
- Einstein Voice
- Einstein Prediction Builder
- Einstein Next Best Action
- Einstein Discovery
- Einstein Vision and Language
What is Einstein Prediction Builder?
Einstein Prediction Builder is a declarative tool that helps fast-track predictions based on Salesforce fields. This helps a Salesforce admin to custom-build predictions on any object via few clicks on a visual interface and power workflows and apps using the newly-found AI. It learns from examples in the past to make predictions. Using projections it helps you work smarter by focusing your time on the right tasks. It predicts the answers to binary questions, i.e., Yes/No and numerical questions.
Understand How Einstein Prediction Builder Works in 3 Simple Steps:
Step 1 : Define your use case
In this step, we need to get a clear picture of how we plan to improve our business using a prediction builder. For Example, Consider an eye clinic where the people make appointments for checkup every month/year. The estimated number of people who did not turn up for an appointment cannot be brought to an exact number. This results in a loss of revenue and resources for that clinic. Next time, they can use historical data from appointments to arrive at an appropriate patient count.
Based on the use case, the prediction builder can calculate either of the below predictions:
- Binary Predictions (predicts a Yes/No)
- Numeric Predictions (predicts a number)
Step 2 : Identify the data used for prediction
The data is very important for prediction model. To help you understand whether the data being considered is sufficient or not, the Einstein team came up with a framework called the “Avocado Framework.”
Dataset: The complete set of data
For example, All the records present in the Appointment object
Segment set: This is the subset of your dataset; here, the subset is formed based on a filter condition specific to your prediction use case.
Example set: This is the data that is used to understand and build a prediction. The example set is also divided based on positive examples and Negative examples; consolidating them creates an example set. For Instance: Positive example set consists of people who have made their appointment and do not show up. The negative example set consists of people who have made their appointment but do show up. Prediction set: This set consists of records on which Einstein is going to perform the prediction. For instance, All the records other than future appointment.
Step 3 : Create prediction
Once we are ready with the use case and data, we can create a prediction using the prediction builder.
Below are the sequence of activities to create a prediction in Prediction Builder:
- Log in to Salesforce. Click on Setup. In the Quick Find box, search for Einstein Prediction Builder.
2. You’ll get to the Einstein Prediction Builder Home Page. To create a new prediction, you can click on a new prediction.
3. Name your prediction. We’ll call our prediction an No show up prediction. API Name here will be auto-populated. Click on Next.
4. Select the object to predict the records. As we are working on the appointment object, we select that. Before clicking on Next, validate the data, whether the minimum number of records are present or not by clicking the Check Data button. Once it validates, the data count is good; we can proceed to the next step by clicking on Next.
5. Einstein Prediction Builder answers two ways either as Yes/No or in a number. Based on the use case, one can choose what kind of answer will be useful to us. As per our use case, we are just predicting whether the patients have show or no show, so we can go for Yes/No. Then click on Next.
6. In this step, we decide whether we have a field based on which we are trying to predict data or not, and what it would be. As we have been using the No show up field based on which we predict the data, we can select the field in this step and click Next. We select the field no show ; for example, we are giving only Yes/No values, i.e., patients who have shown up and have not shown up. In this scenario, we are not considering future appointments, so excluding that, we give an example set to Einstein for prediction. Before moving to the next step, we should check data as an Example set of Yes condition and No condition should have minimum records. Then click on Next.
7. In this step, we can check and uncheck the fields based on which the prediction should occur. Then click on Next.
8. Give the name of Custom Field, which stores the Predictions:
9. In this step, we can find all the values and conditions given in the previous steps. So, we can have a glimpse of the details and confirm whether all information is provided correctly. In this step, we ensure that the data provided is accurate. Now, the only next step here is to click on Build Prediction, which will create a prediction based on the use case, data, and conditions that have been provided.
10. We have successfully made a prediction that will help the eye clinic with their future patient count. Einstein Prediction Builder takes from a few minutes to a few hours to build your prediction and gives a scorecard.
11. This is how the overview of the scorecard looks like. It gives the top five fields affecting the prediction. To get more details based on the predictors, one can click on View, All Predictors.
So we have learnt how easy it is to build custom Intelligence using Einstein Prediction Builder. One does not need to be a Machine learning expert to improve predictability, efficiency and customer experience.
Things to Keep in Mind:
- Prediction Builder requires at least 400 records to be able to make predictions. In the case of binary predictions, the number of records must be at least 100
- Make sure to take note of prediction quality. Having >95% could seem good but is likely due to hindsight bias. A prediction that is low is likely due to not having enough fields to make a great prediction
- Predictions can only be created in production and developer orgs
- Predictions work for all custom objects and support a subset of standard objects
- Einstein can’t build a prediction on an object with a deployment status of In Development. To build a prediction, the object needs to be deployed first
- So, to start building Predictions, look for Einstein Prediction Builder in setup.