What does an Einstein Prediction score represent?

An Einstein Prediction score represents the probability that a specific outcome will occur for a given record, calculated by analyzing historical patterns in your Salesforce data.

How do Einstein Predictions stay accurate over time?

Einstein Predictions automatically retrain as new data flows into Salesforce, ensuring the model stays current with changing business conditions and patterns.

What is the recommended minimum number of records needed to build an Einstein Prediction?

Einstein Prediction Builder recommends at least 400 records with the target outcome to provide enough data for the machine learning model to identify meaningful patterns.

Einstein Prediction

AI🟡 Intermediate

Definition

An Einstein Prediction is an AI-generated forecast that uses machine learning models trained on your Salesforce data to predict future outcomes. Built through tools like Einstein Prediction Builder, predictions analyze historical patterns in your records to score the likelihood of specific events — such as whether a lead will convert, an opportunity will close, or a customer will churn.

Real-World Example

A subscription-based SaaS company uses Einstein Prediction Builder to create a "Churn Risk" prediction on the Account object. The model analyzes historical data — login frequency, support ticket volume, contract renewal dates, and feature usage — to assign each account a churn probability score from 0 to 100. Accounts scoring above 75 are automatically flagged and routed to the customer success team for proactive outreach before their renewal date.

Why Einstein Prediction Matters

Einstein Predictions work by analyzing the historical values of fields on your Salesforce records to identify patterns that correlate with a specific outcome. When you build a prediction, you select a target field (the outcome you want to predict) and Einstein automatically identifies which other fields are the strongest predictors. The resulting model assigns a score to each record, indicating how likely that outcome is to occur.

Predictions are not static — they retrain automatically as new data enters the system, ensuring the model stays current with changing business conditions. Administrators can monitor model performance through scorecards that show prediction accuracy, and they can refine predictions by adjusting which fields are included or excluded. Predictions can also be used as criteria in flows, assignment rules, and reports, making them actionable across the platform.

How Organizations Use Einstein Prediction

•Pinnacle Corp — Built a custom Einstein Prediction to forecast which Opportunities would close within 30 days. The model identified that deals with executive sponsor engagement and more than two product demos had significantly higher close rates, helping reps prioritize their efforts on the most promising deals.
•Vandelay Industries — Created a churn prediction on the Account object that analyzes support case frequency, login patterns, and contract value. When the prediction score crosses a threshold, a Flow automatically creates a task for the customer success manager and sends a personalized retention email.
•Oceanic Corp — Uses Einstein Prediction to forecast which leads are most likely to convert based on industry, company size, lead source, and engagement history. The prediction score is displayed on the Lead record and used in assignment rules to route high-potential leads to senior reps.