Explore Salesforce Recommenders
Learning Objectives
After completing this unit, you’ll be able to:
- Define core Salesforce Personalization terminology.
- Differentiate between Rule-Based and Objective-Based recommendation strategies.
- Explain how the parts fit together to power real-time, cross-channel decisioning.
Before You Start
Before you start this badge, make sure you complete this content. The work you do here builds on the concepts and work you do in that content.
Salesforce Personalization Home
In the Salesforce Personalization in Marketing Cloud Next: Quick Look badge, you learned how Salesforce Personalization uses AI and Data 360 to deliver more relevant customer experiences, from personalized website banners to intelligent recommendations.
You can access Salesforce Personalization using the Personalization app, which is the central hub where you build, manage, and monitor everything from recommendation logic to channel delivery to analytics. This image shows the main areas on the Personalization home page.
The top navigation bar (1) gives you access to every object you'll work with. Each tab maps directly to a concept you'll build. The guided tour (2) helps you get started quickly. Click Start the Tour to learn the three key steps to create your first personalization: Configure a recommender, define a personalization point, and map the two together. All Recommenders (3) and All Personalization Points (4) list views show all your recommenders and personalization points at a glance. Both are empty for now, you start building in the next unit.
Here’s a rundown of the main navigation areas and what you can do.
Object |
Description |
|---|---|
Recommenders |
Build and manage recommendation logic using rule-based strategies, calculated insights, or AI-driven objectives. |
Personalization Points |
Configure channel endpoints that deliver personalized decisions using a content schema and optional targeting rules. |
Data Graphs |
Set up Profile and Item Data Graphs which is the data foundation for personalization and recommendations. |
Attributions |
Measure the business impact of your personalization decisions. |
Experiments |
Configure and manage A/B experiments on Personalization Points. |
Personalization Intelligence |
Access Pipeline Intelligence and Attribution Intelligence dashboards to measure recommendation performance and ROI. |
Personalization Content Schema |
Formerly known as Response Templates, defines the fields returned in each decision response such as product name, image, price, CTA link to render personalized experiences in a consistent format. |
Batch Personalizations |
Generate personalization decisions in bulk for a Data 360 segment and activate results to external systems like MC Engagement. |
Calculated Affinities |
View and manage numerical preference scores that indicate a customer's affinity for specific catalog items or attributes. |
Engagement Signal |
Act as both the input for model training and the foundation for the metrics that a recommender can optimize. For each engagement signal that is created, a simple count metric is automatically generated, for example, the number of purchase events recorded. These metrics can be used directly as objectives in recommenders. |
Experience Templates |
Create channel-specific templates to render personalization decisions in your web and mobile channels. |
Personalization Campaigns |
Use a guided workflow to configure dynamic content personalization experiences on your website. |
Refer to Salesforce Help to learn more about Terminology in Salesforce Personalization.
A Business Scenario
At Cloud Kicks, a fast-growing custom sneaker brand, every shopper who lands on the website sees the exact same thing—the same featured products, the same bestseller list, the same homepage experience. Whether it’s a first-time visitor browsing high-tops or a loyal customer who just bought trail runners last week, the site treats everyone identically.
The result? Missed revenue opportunities and generic experiences that fall short of what today’s shoppers expect.
Linda Rosenberg is the administrator at Cloud Kicks. She’s been tasked with making the website smarter. This means she wants the site to surface the right sneaker to the right shopper at the right moment, based on what they actually care about.
She knows that Salesforce Personalization’s recommenders make this possible. But first, she needs to learn what they are, the different types available, and how they work.
Join Linda as she explores the building blocks of Salesforce Personalization recommenders, and lays the foundation for a smarter Cloud Kicks experience.
Recommendation Types
When you build a recommender, the first step is to choose a recommendation strategy. Salesforce Personalization offers two types, and the right choice depends on the business goal and what’s already set up in Data 360.
Rule-Based Recommendations
A rule-based recommender uses rules-based business logic—such as sort items by price, priority, or recently viewed—or Calculated Insights, such as Top Sellers or co-browse/co-buy from Data 360 to generate results.
Choose rule-based when you want to control which products appear first. This option is also the fastest way to launch a recommender because it doesn’t require model training or historical engagement data. If your Calculated Insight is already set up in Data 360, you can start building right away.
Objective-Based Recommendations
An objective-based recommender uses a deep learning model to generate recommendations for each customer based on a selected objective. Unlike Rule-Based recommendations, the results are personalized at the individual level because the model learns from engagement data and optimizes over time.
Choose objective-based when you want AI to find the best items for each customer. You select the outcome you want, such as maximizing revenue or clicks, and the model does the rest. To use this option, you need a Real-Time Profile Data Graph with engagement history and at least three recorded engagement activities for initial training.
Linda starts with a rule-based recommender as the Top Sellers Calculated Insight is already set up in Data 360. This lets her quickly add a working recommender to the Cloud Kicks homepage. Once enough customer engagement data accumulates, she plans to add an objective-based maximize revenue recommender for returning customers. This helps surface the sneakers each shopper is most likely to add to their cart.
Building Blocks of a Recommender
Before you start building, let's explore what a recommender is made of. Every recommender—regardless of type—is built from the same three core ingredients.
Data Foundations
A recommender needs two Data Graphs to do its job. The Profile Data Graph tells Salesforce Personalization about the person—their purchase history, browsing behavior, affinities, and segment memberships. The Item Data Graph tells it about the items—the product catalog, including product names, categories, prices, and stock status. Together, they give the recommender everything it needs to match the right sneaker to the right Cloud Kicks customer.
While this badge focuses on a retail scenario, Salesforce Personalization can recommend any content defined in an item data graph in Data 360. Other common examples include offers, promotions, articles, blogs, destinations, and hotels.
Strategy
The strategy determines how to select and rank items. For rule-based recommenders, the strategy can be either a Calculated Insights from Data 360 or item/profile based sort criteria, for example, ranking sneakers by total units sold. For objective-based recommenders, the strategy is a business objective, either a standard objective, such as maximize revenue, or a custom objective such as maximize add-to-cart.
Filters
Filters are optional but powerful. They let you define guardrails around which items are included or excluded from results. Three filter types are available.
Filter Types |
What It Compares |
Example |
|---|---|---|
Decision Context Filters |
Item attributes against an anchor item currently being viewed |
Show sneakers in the same category as the one currently being viewed. |
Static Filters |
Item attributes against manually defined values |
Only show items that are currently in stock, or items priced under $200. |
Profile Data Graph Filters |
Item attributes against the customer’s profile |
Exclude sneakers the customer has already purchased. |
Start with the minimum filters needed to meet the business requirement. Overly restrictive filters can reduce the number of items returned, and if too few items pass the filters, customers may see an incomplete recommendation block if a fallback recommender is not defined on the recommender definition.
Once you have configured the filters, you can validate the recommender using the Recommender Simulator. It’s a built-in testing tool that lets you preview exactly which items would be returned for a specific customer before going live.
Execute Real-Time Recommendations
So far, you've explored the building blocks of a recommender—the Profile Data Graph, Item Data Graph, strategy, and filters. But how do they actually work together in the moment a customer lands on the page? Here's what happens.
When a customer visits the Cloud Kicks homepage where a personalization point is configured to return recommendations, a request is made to Personalization for a decision. Salesforce Personalization retrieves that customer’s Profile Data Graph such as their browsing behavior, purchase history, and affinities, and uses it alongside the recommendation strategy to return the most relevant sneakers to them at the 1:1 level.
Consider two shoppers visiting the Cloud Kicks homepage at the same time.
- Shopper A is a returning customer who has been browsing trail runners for the past week. Their Profile Data Graph reflects that behavior. Salesforce Personalization uses it to return a carousel of top-selling trail running sneakers at the top of the homepage.
- Shopper B is a first-time visitor with no browsing history. Their Profile Data Graph has no affinities yet. Salesforce Personalization falls back to the overall top sellers across all categories acting as a strong, relevant starting point for someone new to the brand.
Same homepage. Same recommender. A different experience for every shopper.
For Linda, this means every Cloud Kicks customer now sees a homepage that feels like it was built just for them.
Next Up
In this unit, you explored the Salesforce Personalization home page, and learned two recommendation strategies. In the next unit, learn to build your first recommender.