Enable the feature, verify the data threshold, expose the score on the Opportunity layout, and tune excluded fields. Initial scores appear within 24-48 hours of enable.
- Check the data threshold
Setup, Einstein Opportunity Scoring shows a Data Readiness check. Confirm 200+ closed Opportunities in the past 6 months with at least 20% won and 20% lost. Below threshold, the feature stays unavailable.
- Enable the feature
Setup, Einstein Opportunity Scoring, Enable Einstein Opportunity Scoring. The platform begins training on existing closed Opportunities. First training takes 24-48 hours; scores then appear on open Opportunities.
- Surface the score on layouts
Object Manager, Opportunity, Page Layouts. Add Einstein Score and Einstein Score Tier to the Opportunity Information section. Drop the Einstein Opportunity Scoring Insights component on the Lightning record page for the Top Factors panel.
- Configure excluded fields
Setup, Einstein Opportunity Scoring, Excluded Fields. Exclude PII, free-text comments, and any fields the model could overfit on. Audit the predictor list and remove fields that are not meaningful in your org''s context.
- Build dashboards and adjust pipeline reviews
Standard reports include Einstein Score Tier as a group field. Build a dashboard showing Forecast Category by Einstein Score Tier to surface mismatches. Use the dashboard in weekly forecast calls to identify questionable Commit deals.
Opportunity fields to leave out of the model. Always exclude PII, free-text, and fields that would let the model trivially overfit (such as the explicit Stage value when predicting close).
Hot, Warm, Cold cutoffs (defaults 80/30). Customize to your org''s definition of high-priority pipeline.
Automatic every 10 days. Not user-configurable. Re-enabling the feature triggers an immediate retrain.
Field-level security on Einstein Score controls which users see the score. Hide from junior reps if scoring should drive central triage only.
- Data threshold: 200 closed Opportunities in 6 months with both wins and losses represented. Below this, the feature is unavailable.
- The model retrains every 10 days. Rapid market shifts lag the model by up to two weeks.
- Stage is often excluded from the model to prevent overfitting; reps think Stage is the strongest predictor, but in the model''s view it leaks the answer.
- Einstein cannot see signal in unstructured data (email content, calls). For orgs where most context lives in conversations, pair with Einstein Conversation Insights.
- The score updates daily, not real-time. Field changes today reflect in tomorrow''s score.