Einstein Next Best Action (NBA) is the recommendation engine that surfaces "what should I do next" cards on a record — "offer this discount," "escalate to a manager," "share this article." Recommendations come from a Strategy that combines static rules, predictive Einstein models, and recommendation records. Setup is multi-step: Recommendation records → Strategy → Lightning component on the page.
- Confirm Einstein Next Best Action is licensed
NBA is part of Einstein for Service / Sales bundles. Confirm before enabling.
- Build Recommendation records
App Launcher → Recommendations → New. Each Recommendation is a card the user could see — Name, Description, Image, Action (URL / Flow / Quick Action).
- Open Setup → Next Best Action → Strategy Builder
Strategy Builder is the visual flow tool that decides which Recommendations to surface for which records.
- Build a Strategy
Drag Load Recommendations / Filter / Generate / Limit / Output elements. The Strategy reads context (the current record), filters Recommendations, applies Einstein-scored ranking, and outputs a final list.
- Activate the Strategy
Strategies need to be activated to fire. Inactive ones don't return recommendations.
- Add the Einstein Next Best Action Lightning component to the relevant Lightning Page
Lightning App Builder → drop the Einstein Next Best Action component on Account / Case / Opportunity record page → bind to the activated Strategy.
- Save the Lightning Page → Activate
Recommendations now appear inline on records using the Strategy's logic.
The cards users see. Manually curated or generated from external data.
Load / Filter / Generate / Limit / Branch / Predict / Sort / Output.
Optional — score recommendations using an Einstein Discovery model.
Per Lightning Page — different strategies on different records.
- Strategies don't fire until activated. A built-but-inactive Strategy returns nothing — common oversight after testing.
- Without Einstein Predictive ranking, NBA is rules-only — "if this and that, show recommendation X." Adding Einstein scoring requires an Einstein Discovery model bound to the Strategy.
- Recommendation Action types determine what happens on click. URL opens a link. Flow launches a flow. Quick Action invokes an action. Misconfigured action types fail silently when users click.