Overview

A recommendation strategy controls which catalog items are shown to each user and in what order, based on signals such as user activity, item popularity, or content attributes. It connects your catalog to your campaigns, sources eligible items, and delivers personalized recommendations through engagement. Once created, a strategy can be reused across multiple campaigns.

When you create a strategy, you define:

Which strategy type to use
Which Catalog is used to generate product recommendations
How user activity links to catalog items based on the strategy type selected

This allows you to define recommendation logic once and reuse it across multiple engagements.

Prerequisites

Ensure the following are in place before creating a strategy:

Catalog: At least one Catalog must exist in your account with all required attributes, such as Item ID, Name, Description, Category, Genre, Language, Price, and Brand or tags. If no suitable catalog exists, you can create one during setup. For more information, refer to Catalog.
Event Tracking: If your strategy depends on user behavior, ensure relevant events are tracked in your project, such as Product Viewed, Added to Cart, Charged, or Video Viewed.
Identifier Mapping: Events and catalog data must share a common identifier so the system can link user actions to items. For example, if an event Product Viewed includes a Product ID property, its value must match the corresponding item identifier in the Catalog.

Recommendation Strategies Workflow

Every recommendation strategy is built on the following three layers that work together in sequence:

Layer	Description	Decision it Drives
Catalog	Provides the pool of items available for recommendation	What items to generate recommendations on
Rules	Defines item eligibility and behavioral signals	Which user actions matter?
Strategy logic	Selects items for generating recommendations and ranks them per relevance	Which items must be recommended as part of personalization

These layers feed into the recommendation engine, which processes them at runtime.

OTT: Recommend Content Based on Past Viewing Activity

The following example shows each layer of a strategy configured to recommend content to users based on their recent viewing activity:

Layer	Configuration
Catalog	Movies Catalog
Strategy	Based on what you viewed
Event rule	`video_viewed` in the last 7 days
Mapping	`video_viewed`. Content ID maps to Catalog identity
Result	Recommends content based on recent viewing activity

E-Commerce: Recommend Products Based on Viewing Activity

The following example walks through each layer of a strategy configured to recommend products based on a user's recent browsing activity:

Layer	Configuration
Catalog	Products catalog
Strategy	Based on what you viewed
Event rule	`Product Viewed` in the last 7 days
Mapping	`Product Viewed`. Product ID maps to Catalog identity
Result	Recommends relevant products based on recent browsing behavior

Subscription: Recommend Popular Articles

The following example shows each layer of a strategy configured to recommend content to users based on the popular articles:

Layer	Configuration
Catalog	Articles catalog
Strategy	Popular
Event rule	`article_read` in the last 7 days
Mapping	`article_read`.Article ID maps to Catalog identity
Result	Recommends articles that have been read the most in the last 7 days

Create Strategy

Creating a strategy involves the following three key steps:

Select Strategy Type

Choose a strategy type that matches your use case. Each type uses different signals to generate recommendations, such as what a user has recently viewed, what is trending across your audience, or what is similar to items they have engaged with.

You can search by name or browse available options. For a detailed comparison to help you pick the right type, refer to Choose Right Strategy.

Configure Data Source and Recommendation Rules

This step defines where the system sources items from and what user behavior it should consider when generating recommendations. You select a catalog, set rules based on user activity, such as views, add to carts, or purchases, and specify which event data should be used to match user actions to catalog items.

Select Catalog

Select a catalog from the dropdown. A preview table displays sample items and attributes. If no suitable catalog is available, you can create one directly from this step. For more information, refer to Create Catalog.

Define Criteria

Criteria determine which items qualify for recommendation based on user activity. Each criteria block includes:

Event name: The user action to track (for example, Product Viewed, Added to Cart, Video Viewed.
Time window: How far back the system looks (for example, last 7 days)
(Optional) Filters: narrow the event scope by event properties.

For example, CRITERIA A: Did Video Viewed in the last 7 days.

You can define multiple criteria blocks. For more information on combining criteria, refer to Filter Rules.

Map Catalog and Event Properties

Events and catalog data must share a common identifier so the system can link user actions to items. For each event defined in your criteria, use the dropdown to map its properties to the corresponding catalog identifier.

For example, if a Video Viewed event includes a show_id property, map it to the catalog column that holds the same identifier value so the system can correctly attribute that interaction to a catalog item.

Event	Event Property	Catalog Identifier
Video Viewed	show_id	Content ID
Product Viewed	product_id	Item ID

The following image shows how to create a Recommendation Strategy:

Add Basic Details

Enter the following details and click Create to save the strategy:

Field	Description
Name	Use a clear and descriptive name that reflects the use case, Catalog, and logic. For example, Popular Shoes by Purchases, Trending Hindi Movies, Viewed Electronics Last 7 Days, and so on.
Description	Add additional context if needed.

Recommendation Strategy States

A strategy moves through a defined set of states from the moment it is created until it is deleted. Understanding these states helps you track what a strategy is doing and take the right action at each stage.

📘
Strategy Management Constraints
A strategy cannot be deleted while it is Active, it must be stopped first. Restarting a Stopped strategy does not reset its interaction data; it resumes serving recommendations based on existing data. A strategy in the Error state can be restarted directly without stopping first.

The following image shows a strategy moving through five states, from creation to deletion:

The following table describes each strategy state:

State	Description
In Progress	The system is computing recommendations for the first time. No recommendations are served until this process completes.
Active	The strategy is live and actively serving recommendations.
Active But Refresh Failed	The strategy is still serving recommendations based on its last successful run, but the most recent scheduled refresh failed. Recommendations may be stale. Monitor the next refresh cycle to confirm recovery, or stop and restart the strategy to trigger an immediate retry.
Failed	Activation failed. The strategy could not reach the Active state, typically due to a misconfigured catalog mapping or event rule. Review the error details and click Start to retry without losing your configuration.
Stopped	The strategy has been manually paused. Recommendations are no longer served. Click Start to restart, the strategy re-enters the In Progress state and returns to Active once processing completes.
Deleted	The strategy has been removed. A strategy can only be deleted from the Stopped state. Deletion cannot be undone.

Manage Recommendation Strategies

Go to the Strategies tab to view and manage all recommendation strategies present in your account.

The strategies list displays the following information for each strategy:

Column	Description
Name	Strategy name and ID. Click the copy icon to copy the strategy ID.
Status	Current state of the strategy, displayed as a color-coded badge.
Algorithms	Recommendation algorithm types used by the strategy.
Catalog	The catalog on which the strategy is built.
Created On	Date the strategy was created.
Created By	Team member who created the strategy.
Engagements	Number of active engagements using this strategy. Click the count to view a list of all engagements.

Each strategy has the following actions available:

Action	Description	Available When
Clone	Creates a copy of the strategy with the same configuration.	Always
Open in new tab	Opens the strategy detail view in a new browser tab.	Always
Start	Activates the strategy. It re-enters the In Progress state and moves to Active once processing completes.	Stopped, Failed
Stop	Pauses the strategy. Recommendations are no longer served. If the strategy is used during engagements, a confirmation dialog is displayed before your action.	Active, Active But Refresh Failed, In Progress
Delete	Permanently removes the strategy. This action cannot be undone.	Stopped

Filter Strategies

Click the icon to open the Filters panel. You can filter by:

Type	Description
TIME PERIOD	Filter by creation date. Defaults to All Time.
STATUS	Filter by one or more strategy states.
CREATED BY	Filter by the team member who created the strategy.
TYPE	Filter by algorithm type. Use the search bar or select all available options.
CATALOG USED	Filter by the Catalog used to build the strategy.

Best Practices

Always validate catalog selection before defining rules so that the correct Catalog is used for generating product recommendations
Ensure event-to-catalog mapping is correct for activity-based strategies
Use descriptive strategy names so they are easy to identify when reusing across engagements