RecommendationStrategy in SFDX

Architectural Overview#

Front-line customer service agents and sales representatives are frequently overwhelmed with competing priorities, policy manuals, and disconnected customer data. Einstein Next Best Action (NBA) addresses this challenge by delivering context-aware, prioritized recommendations directly inside Lightning record pages and Experience Cloud portals.

The RecommendationStrategy metadata type represents the declarative decision engine of Next Best Action. Strategies load candidate recommendations, evaluate business rules, incorporate machine learning predictions from Einstein Prediction Builder, filter out ineligible offers, and sort recommendations by business value.

By maintaining recommendation strategies in source control, enterprise architects guarantee that customer engagement logic, compliance guardrails, and cross-sell rules remain synchronized across all delivery pipelines.

Directory and File Conventions#

In SFDX source format, RecommendationStrategy metadata files reside in the recommendationStrategies/ directory:

Key Metadata Fields#

The XML schema of RecommendationStrategy models the decision graph and strategy nodes:

Declarative XML and Package Manifest Examples#

Below is a declarative XML definition of an enterprise customer retention strategy:

<?xml version="1.0" encoding="UTF-8"?>
<RecommendationStrategy xmlns="http://soap.sforce.com/2006/04/metadata">
    <description>Evaluates churn propensity scores to surface proactive retention incentives.</description>
    <masterLabel>Retention Offer Strategy</masterLabel>
</RecommendationStrategy>

Package Manifest (package.xml)#

<?xml version="1.0" encoding="UTF-8"?>
<Package xmlns="http://soap.sforce.com/2006/04/metadata">
    <types>
        <members>Retention_Offer_Strategy</members>
        <name>RecommendationStrategy</name>
    </types>
    <version>68.0</version>
</Package>

CLI Retrieval and Deployment Commands#

Manage RecommendationStrategy metadata using the Salesforce CLI:

# Retrieve all recommendation strategies
sf project retrieve start --metadata RecommendationStrategy

# Deploy a specific recommendation strategy
sf project deploy start --metadata "RecommendationStrategy:Retention_Offer_Strategy"

# Validate deployment into sandbox
sf project deploy validate --metadata RecommendationStrategy

Architect Best Practices and Gotchas#

• Flow Dependency Management: If a recommendation node triggers a screen Flow, that flow must be deployed and active in the target org; otherwise, clicking the recommendation in the UI will throw a runtime execution exception.
• Lightning Page Performance: Strategies execute in real-time during Lightning page loading. Strategies with excessive external callouts or unindexed SOQL filters will introduce page load latency. Use Limit elements early in the strategy graph.
• Fallback Offers: Always include a fallback recommendation branch so front-line agents are never presented with an empty or broken component when advanced criteria are not met.

Frequently asked questions

What is the primary role of RecommendationStrategy?

It orchestrates decision logic in Einstein Next Best Action, filtering, ranking, and surfacing contextual recommendations and offers to users.

Where are RecommendationStrategy metadata files stored in SFDX?

They reside in force-app/main/default/recommendationStrategies/ with the .recommendationStrategy-meta.xml suffix.

What happens when a user accepts a recommendation?

Accepting a recommendation launches the configured screen Flow or Quick Action bound to that recommendation record.

Can RecommendationStrategy incorporate external AI predictions?

Yes, strategies can evaluate predictions from Einstein Prediction Builder, CRM Analytics, or external machine learning models via Apex callouts.