Recommendation Rules

Recommendation Rules determine which products are recommended to users and how those products are selected.

A recommendation rule acts as the decision-making logic behind a recommendation widget. Whenever a recommendation widget is displayed in an email, in app or onsite campaign, the widget uses a recommendation rule to decide which products should be shown to each user.

For example:

• Show the best-selling products in a category.
• Show products similar to the one a visitor is viewing.
• Show products that are frequently purchased together.

The recommendation rule defines this behavior.

Prerequisites

Before creating recommendation rules, ensure that:

• The SDK is integrated.
• Analytics Event Definition is enabled.
• Product integration is completed.

How to Access

Go to Content > Assets > Recommendation Engine.

From this screen, you can create, edit, and manage recommendation rules.

How Recommendation Rules Work

A recommendation rule defines how the system selects and displays products to users.

Every time a recommendation is shown, the system follows a clear step-by-step process.

Step 1: Read the Recommendation Rule

The system first identifies the recommendation rule connected to the widget or product box.

This rule contains instructions such as:

• Which recommendation approach to use
• Which products or data to consider
• Which conditions and filters to apply

Step 2: Apply the Strategy

The system checks the Strategy in the rule.

The strategy defines how recommendations are generated:

• Rule-Based: Uses product catalog data (e.g. top sellers, category best sellers,new arrivals, category new arrivals, discounted products, category discounted products, trending products)
• Predictive: Uses AI and user behavior (e.g. similar items, frequently bought together, frequently viewed together, recommended items (user-based) )

Step 3: Apply the Model Type

After selecting the Strategy, choose the appropriate Model Type. The model type defines the logic used to generate recommendations.

Rule-Based Models

Rule-Based models use product catalogue data and do not rely on individual user behaviour. They focus on product performance or category-level data. These models include:

• Top Sellers
• Category Best Sellers
• New Arrivals
• Category New Arrivals
• Discounted Products
• Category Discounted Products
• Trending Products

Predictive Models

Predictive models use AI and user behavior to generate personalized recommendations. These models include:

• Similar Items
• Frequently Bought Together
• Frequently Viewed Together
• Recommended Items (User-Based)

Some models work independently, while others depend on a specific product or category to generate meaningful results. This is why some models require a context, which is explained in the next step.

Step 4: Check the Context (if required)

Some models require a Context Source to generate recommendations. The context defines what the system uses as the starting point for recommendation generation.

In simple terms:
Context is the input that tells the system what the recommendations should be based on.

Context is required when the model depends on a specific product or category.

Category Context (Rule-Based)

Category context is used for category-based models such as Category Best Sellers, Category New Arrivals, and Category Discounted Products.

These models generate recommendations only within a selected category, meaning the system ranks or filters products that belong to that category.

Category context can be defined in three ways:

• Static category: You manually enter category path.
• User attribute: Based on user data such as Favourite Category or Last Viewed Category. This makes recommendations more personalized.
• Event attribute: Based on real-time behavior such as Current Category the user is viewing.

Product Context

Product context is used for predictive models such as Similar Items, Frequently Bought Together, and Frequently Viewed Together.

These models generate recommendations based on a specific product, by analyzing relationships between products or user interactions.

Product context can also be defined in three ways:

• Static product: You manually enter a product ID.
• User attribute: Based on user behavior such as Favourite Product, Product in Cart, or Last Viewed Product.
• Event attribute: Based on real-time behavior such as Current Product the user is viewing.

Step 5: Apply Output Settings and Filters

In this step, you can configure how the final recommendation list is displayed and how products are selected. These settings control both the number of products shown and the filters used for product selection.

Product Selection Filters

You can apply filters to control which products are included in the recommendation results. These filters help refine product selection and improve relevance.

• Shuffle products: Changes the order of products so the same ranking is not always shown
• Exclude recently purchased products: Removes products the user has already bought
• Exclude recently viewed products: Prevents showing products the user has already seen
• Exclude products in cart: Excludes items already added to the cart
• Show only in-stock products: Ensures only available products are shown

These filters help ensure users see more relevant and useful product suggestions.

You can further refine product selection using advanced filtering conditions. These allow you to narrow down which products are eligible for recommendation based on specific criteria.If you apply too many restrictive filters at the same time, there may not be enough products left to display. In this case, the final result may return fewer products or be handled by fallback logic depending on the configuration.

Step 6: Generate Recommendations

After all previous steps are applied (strategy, model type, context, and filters), the system generates the final list of recommended products.

At this stage, the system evaluates the filtered product pool and prepares the final output that can be used in different channels such as web, email, or in-app experiences.

Step 7: Fallback Behavior

After generating recommendations, the system checks whether any products were found.

If no products match the selected strategy, model, context, and filters, you can define what should happen next using Fallback Behavior.

You can choose one of the following options:

• Show Fallback Model: Uses an alternative recommendation model, such as Top Sellers, New Arrivals, Discounted Products, or Last Visited Products.
• Hide Recommendations: Returns no recommendations when no matching products are found.

For example, if the primary model is Frequently Bought Together and no products are found, the system can automatically show Top Sellers as the fallback model.