Batch Personalization

Run personalization at scale outside of real-time decisioning. Batch personalization lets you generate personalized recommendations and content for large audiences on a schedule — ideal for email campaigns, push notifications, CRM-driven workflows, and other offline or asynchronous channels.

1. What Is Batch Personalization?

Real-time personalization evaluates decisions at the moment a user interacts with your website or mobile app. Batch personalization, by contrast, generates personalized decisions for a set of individuals ahead of time — either on a recurring schedule or triggered by a specific event — so the results can be consumed by downstream systems.

Key differences from real-time personalization:

2. Use Cases

Batch personalization unlocks a wide range of scenarios where decisions need to be made ahead of time or at scale:

Personalized Email Content

Generate individualized product recommendations or content for each recipient in an email campaign. For example:

• "You may also like" product recommendations in a weekly newsletter
• Abandoned cart follow-up emails with the specific products the customer left behind
• Post-purchase cross-sell emails recommending complementary items

Pre-Computed Recommendations for Push Notifications

Mobile push notifications can include personalized product or content recommendations that were pre-computed via batch personalization. This avoids the latency of making a real-time API call at send time.

CRM-Driven Next-Best-Action

Use batch personalization to feed personalized recommendations into CRM workflows — for example, surfacing the next-best-offer for a sales representative to discuss during a customer call. By pre-computing decisions, the recommendations are available instantly when the rep opens the account record.

Agentforce Integration

Salesforce Personalization extends decisioning services into the Agentforce ecosystem through agent actions. Available actions include:

• Get Recommendations: Allows an agent to request a personalized decision configured in Salesforce Personalization and inject individualized, ML-powered relevance into the agentic conversation.
• Get Context: Personalization's context curation service ensures that your agent is always armed with the relevant data needed to address the customer's question.

3. Prerequisites

Before configuring batch personalization, ensure the following are in place:

4. Personalization Points Across Channels

A key concept in Salesforce Personalization is that personalization points are reusable across channels. Any channel that can be configured to call the Personalization decisioning API can make a request against a personalization point. This means:

• The same personalization point you created for your website homepage recommendations can also be used in an email campaign, a mobile push notification, or an Agentforce conversation.
• Channel-specific logic (when to make the request, where to render, how to display) is handled by the consuming application — not by the personalization point itself.
• Each decision request requires an individual ID and one or more personalization point IDs. Context data (such as an anchor item ID) can be included for recommendations filtering.

How the Decisioning Pipeline Works in Batch

The decisioning pipeline follows the same stages whether the request is real-time or batch:

1. Personalization Request — The request includes an individual ID and personalization point ID(s). Context data can optionally be provided.
2. Augmenting Phase — Personalization retrieves the individual's profile data graph from Data Cloud. For batch requests, this is read from the Lakehouse.
3. Qualifying Phase — The pipeline evaluates targeting rules on each decision to determine which decision the individual qualifies for. Experiments are treated as the highest-priority object on a personalization point.
4. Personalizing Phase — For manual content, the configured text/image values are returned. For recommendations, the profile is passed to the recommendations service to generate 1:1 personalized results.
5. Decision Response — The response is returned to the requesting application and logged to the personalization log in Data Cloud.

5. Invocable Actions — Get Personalization Decision

The primary mechanism for triggering batch personalization is the "Get Personalization Decision" invocable action. This action can be used in Salesforce Flow and Agentforce to request a personalization decision programmatically for a specific individual.

What Is an Invocable Action?

An invocable action (IA) is a reusable unit of logic that can be called from Flows, Process Builder, Agentforce agents, Apex, and the REST API. The Get Personalization Decision IA calls the Personalization decisioning pipeline and returns the decision response — just like a real-time request from the Web SDK, but triggered from server-side logic.

When to Use the Invocable Action

Setting Up the Get Personalization Decision Invocable Action in Flow

Follow these steps to use the invocable action in a Salesforce Flow:

1. Navigate to Setup → Flows and create a new Flow (Auto-Launched or Scheduled, depending on your use case).
2. Add an Action element to your Flow canvas.
3. Search for and select the "Get Personalization Decision" invocable action.
4. Configure the input parameters:

5. Map the output of the action to Flow variables to capture the decision response.
6. Add downstream logic (e.g., send an email, update a record, call another service) that consumes the personalized decision data.
7. Activate the Flow.

jsonCopy// Example: Decision response structure returned by the invocable action
{
  "personalizations": [
    {
      "personalizationPointId": "home_page_recs",
      "personalizationId": "0ppxx000000ABCDEF",
      "contentId": "decision_001",
      "data": {
        "recommendations": [
          {
            "itemId": "prod_12345",
            "name": "Running Shoes Pro",
            "imageUrl": "https://example.com/images/shoes-pro.jpg",
            "price": 129.99,
            "url": "/products/running-shoes-pro"
          },
          {
            "itemId": "prod_67890",
            "name": "Trail Running Jacket",
            "imageUrl": "https://example.com/images/trail-jacket.jpg",
            "price": 89.99,
            "url": "/products/trail-running-jacket"
          }
        ]
      }
    }
  ]
}

6. Scheduled Batch Personalization with Flow

One of the most common batch personalization patterns is a scheduled Flow that runs on a recurring basis and generates personalized decisions for a target segment of customers.

Example: Nightly Product Recommendations for Email

Goal: Every night at 8 PM, generate personalized product recommendations for all customers in the "Active Shoppers" segment, then store the results for the morning email send.

Step-by-Step Setup

1. Create or confirm your segment in Data Cloud:

   • Navigate to Data Cloud → Segments.
   • Create a segment called "Active Shoppers" targeting Unified Individuals who have browsed products in the last 30 days.

2. Confirm your personalization point has an active decision with a recommender:

   • Navigate to the Personalization app → Personalization Points.
   • Verify that your point (e.g., "Email Product Recs") has at least one active decision with a trained recommender.

3. Create a Scheduled Flow:

   • Navigate to Setup → Flows → New Flow → Schedule-Triggered Flow.
   • Set the schedule to run daily at 8:00 PM.
   • Add a Get Records element to retrieve Unified Individual records from the "Active Shoppers" segment.
   • Add a Loop element to iterate over each individual.
   • Inside the loop, add an Action element using the "Get Personalization Decision" invocable action.
     • Pass the individual's Unified Individual ID as the Individual ID input.
     • Pass the personalization point API name (e.g., email_product_recs).
   • After the action, add logic to store or route the decision response:
     • Option A: Update a custom object record with the recommendation data for later consumption by your email platform.
     • Option B: Use a Send Email action (if using Marketing Cloud on Core) to send the personalized email inline.
   • Activate the Flow.

7. Marketing Cloud Integration

For customers using Marketing Cloud Growth or Marketing Cloud Advanced (both built on the Salesforce Platform with Data Cloud), personalization integrates directly into the email authoring experience.

How It Works

1. When building an email in Marketing Cloud Growth or Advanced, use the email builder's personalization component to add a personalization block.
2. Behind the scenes, this creates a personalization point in the Personalization app with a source of "CMS Content".
3. The email builder provides a visual interface for selecting:
   • The personalization point to use
   • The response template and decision configuration
   • Fallback content for recipients who don't qualify for any decision
4. At send time, the email system calls the Personalization decisioning pipeline for each recipient to generate individualized content.

8. Output & Consumption

Batch personalization results can be consumed by downstream systems in several ways, depending on the integration pattern:

Decision Response Storage

When using a Flow or Apex to call the Get Personalization Decision invocable action, the decision response is returned as a structured JSON object. You control where and how this data is stored:

Decision Logging in Data Cloud

Regardless of how you consume the results, every decision made by the Personalization pipeline is automatically logged in Data Cloud's personalization log. This logging is critical for:

• Attribution analytics — Understanding which personalized experiences drive revenue and conversions.
• Pipeline Intelligence dashboards — Monitoring total personalization requests, unique individuals targeted, and decision distribution across personalization points.
• Experiment analytics — If the personalization point has an active experiment, batch decisions are included in experiment cohort analytics.

9. Best Practices

Manage Decision Credit Consumption

• Segment-first approach: Always target a specific segment of customers rather than running batch personalization against your entire customer base. Focus on individuals most likely to engage.
• Limit personalization points per request: Each personalization point in a request consumes a separate decision credit. Only request what you need.
• Monitor usage: Use the Pipeline Intelligence dashboard to track total personalization requests and unique individuals targeted over time.

Optimize Data Freshness

• Calculated Insight refresh: For rules-based recommenders, the fastest automated refresh window for a batch CI is 1 hour. If new items need to appear in recommendations quickly, consider manually triggering CI refreshes via Data Cloud APIs or Flow orchestration.
• Item data graph refresh: The standard NRT (near-real-time) data graph refresh cycle is 30 minutes. New items added to a DMO will be available for rules-based recommender decisions after both a CI refresh and a DG refresh complete.
• Recommender training: Objective-based recommenders train on a 24-hour cycle. New items require at least one training cycle to generate the embeddings needed for objective-based decisioning.

Plan for Fallbacks

• Fallback content: Always configure fallback or default content on your personalization decisions for individuals who don't qualify for any targeting rules. This ensures every individual in your batch run receives some content rather than a blank response.
• Error handling in Flow: Add fault paths to your Flow to handle scenarios where the invocable action fails (e.g., the personalization point is misconfigured, the recommender hasn't trained, or the individual doesn't exist in the profile data graph).

Security Considerations

• Authenticated personalization points: If your personalization point is marked as authentication required, batch requests via the invocable action must include proper credentials. The invocable action handles authentication within the Salesforce platform context, so this is typically seamless for server-side Flow execution.
• Data access: Ensure that the user running the scheduled Flow has the appropriate Personalization and Data Cloud permission sets assigned. See Setup & Permissions for details.

10. Putting It All Together — End-to-End Example

Here's a high-level walkthrough of setting up batch personalization for a weekly personalized email campaign:

Scenario

You want to send a weekly email to customers in the "Loyalty Members" segment with personalized product recommendations based on their browsing and purchase history.

Steps

1. Data Foundation (already in place):

   • Web SDK is capturing browse, cart, and purchase events → mapped to DMOs in Data Cloud.
   • Identity resolution is deployed → Unified Individual records exist.
   • Profile data graph includes engagement objects (Product Browse, Shopping Cart, Product Order).
   • Item data graph is configured for Goods Product with fields: name, image URL, price, product URL.
   • An objective-based recommender is trained and active.

2. Create the Personalization Point:

   • Navigate to the Personalization app → Personalization Points → New.
   • Select your data space, profile data graph, personalization type = Recommendations, and the appropriate response template.
   • Name it loyalty_email_recs.

3. Add a Decision:

   • On the loyalty_email_recs personalization point, add a decision.
   • Optionally add targeting rules (e.g., only target individuals with segment membership = "Loyalty Members").
   • Select your trained recommender.
   • Configure a fallback decision with top-selling products for individuals who don't qualify for the primary decision.

4. Build the Scheduled Flow:

   • Create a Schedule-Triggered Flow set to run weekly (e.g., every Monday at 6:00 AM).
   • Retrieve Unified Individual records from the "Loyalty Members" segment.
   • Loop through each individual and call the Get Personalization Decision IA with loyalty_email_recs.
   • Store each individual's recommendation results in a custom object (e.g., Personalized_Email_Content__c).

5. Consume in Email Campaign:

   • Your email platform (Marketing Cloud or other) reads from the Personalized_Email_Content__c records to populate each recipient's email with their personalized product recommendations.
   • Send the campaign.

6. Measure Performance:

   • Review the Pipeline Intelligence dashboard to see total batch decisions made.
   • Set up an attribution model to measure whether personalized email recipients convert at a higher rate than non-personalized recipients.
   • If you configured an experiment on the personalization point, review experiment analytics after the 24-hour data processing window.

Summary

Batch personalization extends the power of Salesforce Personalization beyond real-time web and mobile experiences. By leveraging personalization points, the Decisioning API, invocable actions, and Salesforce Flow, you can deliver individualized recommendations and content through email, push notifications, CRM workflows, and Agentforce conversations.