DiscoveryGoal Metadata Type in SFDX

Architectural Overview#

The DiscoveryGoal metadata type defines the strategic business objective for machine learning initiatives within Einstein Discovery and CRM Analytics. Before an algorithmic model can be trained or deployed, architects and data scientists establish a target business outcome—such as maximizing revenue per account, minimizing case handle duration, or predicting employee retention.

A DiscoveryGoal establishes the primary key performance indicator (KPI), the target outcome column on the source dataset, and the optimization direction (Maximize or Minimize). Serving as the parent container for multiple DiscoveryAIModel records, DiscoveryGoal provides an architectural framework for versioning, comparing champion vs. challenger models, and executing A/B test splits in production environments.

SFDX Directory and File Suffix#

Key Metadata Fields#

Declarative XML and Package.xml Examples#

DiscoveryGoal Metadata File#

<?xml version="1.0" encoding="UTF-8"?>
<DiscoveryGoal xmlns="http://soap.sforce.com/2006/04/metadata">
    <description>Strategic goal to maximize B2B enterprise opportunity close-win conversion rates.</description>
    <label>Maximize Opportunity Win Rate</label>
    <outcomeField>IsWon</outcomeField>
    <outcomeGoal>Maximize</outcomeGoal>
    <status>Active</status>
</DiscoveryGoal>

package.xml Manifest#

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

SFDX CLI Retrieval and Deployment Commands#

Retrieve and deploy discovery goals across environments:

# Retrieve all DiscoveryGoal metadata components
sf project retrieve start --metadata "DiscoveryGoal" --target-org analytics-dev

# Deploy the goal along with its deployed predictive models
sf project deploy start \
  --metadata "DiscoveryGoal:Maximize_Opportunity_Win_Rate,DiscoveryAIModel:Opportunity_Win_Predictor" \
  --target-org uat-org

# Query existing discovery goals via Tooling API
sf data query --use-tooling-api --query "SELECT Id, DeveloperName, OutcomeField, OutcomeGoal FROM DiscoveryGoal" --target-org uat-org

Senior Architect Best Practices and Gotchas#

• KPI Alignment with Business Stakeholders: Ensure the outcomeGoal directionality aligns with corporate OKRs (e.g. minimizing customer churn rate rather than maximizing churn probability).
• Security Scoping for Analytics Datasets: Restrict access to training datasets powering discovery goals using CRM Analytics sharing inheritance to prevent exposure of sensitive transaction history.

Frequently asked questions

What business purpose does a DiscoveryGoal serve?

It formalizes the business metric to optimize such as maximizing revenue or minimizing churn, establishing the target KPI for machine learning models.

Where are DiscoveryGoal files stored in an SFDX project?

They are stored in force-app/main/default/discoveryGoals with the file suffix .discoveryGoal-meta.xml.

Can multiple DiscoveryAIModel records be attached to one DiscoveryGoal?

Yes. Multiple model versions or competing algorithm architectures can attach to a single goal to facilitate champion-challenger AB testing.