AI Model Wiki

    Skip to main content
    14 min
    Lesson 3 of 4

    Advanced Analytics & Best Practices

    From the course Tableau AI & Pulse

    Take your Tableau AI usage to the next level with advanced analytics patterns and organizational best practices.

    Einstein Discovery Deep Dive

    Building a Prediction Model

    1. Select the dataset: Choose a dataset with historical outcomes
    2. Define the outcome: What you want to predict (numeric or boolean)
    3. Einstein analyzes: Identifies correlations and builds the model
    4. Review insights: See key drivers and their impact
    5. Deploy: Add predictions to dashboards or Salesforce

    Model Quality

    • Einstein provides model quality metrics (R², accuracy)
    • Review feature importance to validate the model makes sense
    • Check for bias in predictions
    • Retrain models periodically with new data

    What-If Analysis

    Einstein Discovery supports what-if scenarios:

    • "What if we increase marketing spend by 20%?"
    • "What if we reduce response time to under 1 hour?"
    • See predicted impact on your outcome metric

    Organizational Best Practices

    Data Governance

    • Certify data sources used for AI features
    • Document data lineage and transformations
    • Set up permissions for who can create predictions
    • Regularly audit AI-generated insights for accuracy

    Adoption Strategy

    • Start with Pulse for executives (low effort, high visibility)
    • Roll out Ask Data to business users with training
    • Reserve Einstein Discovery for analysts who can validate models
    • Share success stories to build momentum

    Common Pitfalls

    • Using AI predictions without understanding the underlying data
    • Over-relying on automated insights without domain knowledge
    • Not retraining models as business conditions change
    • Ignoring data quality issues that affect AI accuracy

    Integration with Salesforce

    If you use Salesforce CRM:

    • Einstein Discovery models deploy directly to Salesforce
    • Predictions appear on record pages
    • Triggers and flows can use prediction scores
    • Single AI platform across analytics and CRM

    Key Takeaways

    • •Einstein Discovery provides model quality metrics, feature importance, and what-if analysis
    • •Start AI adoption with Pulse for executives before rolling out to broader teams
    • •Certify data sources and regularly audit AI insights for accuracy
    • •Einstein Discovery models can deploy directly to Salesforce records and flows
    0 of 4 lessons marked complete

    Progress is kept in this browser only.