Model Builder
In Salesforce Einstein or Data Cloud, a tool for creating and managing machine learning models that can analyze data patterns and make predictions, configurable through a visual interface without deep data science expertise.
Definition
In Salesforce Einstein or Data Cloud, a tool for creating and managing machine learning models that can analyze data patterns and make predictions, configurable through a visual interface without deep data science expertise.
In plain English
“Model Builder is a tool in Salesforce Einstein and Data Cloud for creating machine learning models that analyze data and make predictions. You can build models through a visual interface without needing deep data science skills, using your Salesforce data as training input.”
Worked example
At Vanguard Solutions, the Customer Success team wants to predict which accounts are likely to churn so they can intervene before contract end. The CS Operations lead opens Model Builder in Data Cloud and selects the input signals - engagement score over the last 90 days, support Case volume, payment delays, login frequency, NPS scores. Model Builder trains a churn-prediction model on historical churned-vs-retained data and surfaces a daily risk score on every Account. CSMs see the score in their morning queue and prioritize at-risk accounts for outreach - replacing intuition-based prioritization with a measurable signal.
Why Model Builder matters
In Salesforce Einstein or Data Cloud, Model Builder is a tool for creating and managing machine learning models that can analyze data patterns and make predictions. It's configurable through a visual interface that doesn't require deep data science expertise: admins or business analysts can define what to predict, choose which fields the model should consider, and let Salesforce handle the underlying model training. The resulting models can be used to score records, drive automation, or inform recommendations.
Model Builder represents Salesforce's approach to making machine learning accessible to organizations without dedicated data science teams. While serious ML projects still benefit from data science expertise, Model Builder enables many practical use cases (churn prediction, lead conversion likelihood, opportunity scoring) without that expertise being a prerequisite. The Einstein product family also supports more sophisticated ML through Einstein Discovery and bring-your-own-model integrations for organizations that have data scientists.
How to set up Model Builder
Model Builder is the no-code ML model creation tool in Salesforce Data Cloud and Einstein — pick training data, an outcome to predict, the model auto-trains and deploys to score Salesforce records. Lower barrier than Einstein Discovery (which lives in CRM Analytics); Model Builder targets admin / business analyst users without data-science background.
- Confirm Data Cloud or Einstein Model Builder is licensed
Model Builder is part of paid AI bundles. Check Setup → Einstein Setup → Model Builder.
- Open Model Builder (App Launcher → Data Cloud or Einstein Studio)
Navigation depends on which licensed product. Most modern orgs find it under Data Cloud.
- Click New Model
Wizard walks through model creation.
- Pick the data source
Salesforce object / Data Cloud Data Lake Object / external connector. The dataset Model Builder trains on.
- Pick the outcome (target field)
What the model predicts — Yes/No / Numeric / Categorical. Drives the model algorithm.
- Select input features
Which fields the model considers. More isn't always better — irrelevant fields confuse training.
- Train the model
Salesforce trains in the background. Takes minutes to hours depending on data size.
- Review model performance
Accuracy / precision / recall / confusion matrix. If too low, iterate on feature selection.
- Deploy as a prediction field on records
Salesforce writes predictions to a field on the source object — auto-populated for new and updated records.
Binary / Multi-Class / Regression. Drives algorithm selection.
Which fields go in.
Auto-retrain frequency. Default weekly.
Custom field / Apex API / Flow Action.
- Models drift. Without retraining, accuracy degrades as data changes. Default weekly refresh is fine for stable datasets; faster-drifting data needs more frequent retraining.
- Class imbalance (10x more Yes than No) produces misleading models that just predict the majority. Salesforce flags this in performance metrics — adjust training data or use balanced-class models.
- Predictions write to read-only fields on records. Don't try to manually edit predicted values — the next refresh overwrites.
How organizations use Model Builder
Built churn prediction models in Einstein Model Builder for proactive retention outreach, without needing dedicated data scientists.
Uses Model Builder for lead conversion prediction, with the visual interface letting their sales ops team own the model.
Started with Model Builder for early ML wins, then graduated to Einstein Discovery as their needs became more sophisticated.
Test your knowledge
Q1. What is Model Builder?
Q2. Who can use Model Builder?
Q3. What should you do with model predictions?
Discussion
Loading discussion…