Using item recommendations in your messaging

After your recommendation is trained, you can use Liquid to fetch and display recommended items in your messages. The key here is working directly with the product_recommendation Liquid object. This article covers the product_recommendation Liquid object and includes a tutorial to help you put that knowledge into practice.

For additional guidance on using AI item recommendations in Braze, check out our Braze Learning Course on Crafting Personalized Experiences with AI. This course covers industry use cases, step-by-step instructions, and an additional use case for creating an in-app message with AI-driven recommendations.

Anatomy of the recommendation object

The product_recommendation object represents the set of items recommended by the model. It provides data directly from the associated catalog, structured as an array of objects, where each object represents a recommended item.

• Structure: Each item is accessed as items[index], where index starts at 0 (for the first item) and increments for subsequent items.
• Catalog fields: Each item in the array contains key-value pairs corresponding to fields (columns) in the catalog. For example, common catalog fields for product recommendations include:
  • name or title
  • price
  • image_url

Liquid tags

The product_recommendation object contains dynamically generated product recommendations. To access these in Liquid, you must first assign the data to a variable before using it in your message.

Assigning recommendation data

Always start with the assign tag to fetch the product_recommendation data and store it in a variable.

1
{% assign items = {{product_recommendation.${RECOMMENDATION_NAME}}} %}

• RECOMMENDATION_NAME: Replace this with the name of the AI recommendation you created in Braze.
• items: The variable storing the recommended items array.

Accessing individual items

After the recommendation data is assigned, you can reference specific items and their fields using array indexing and dot notation:

1
2
{% assign items = {{product_recommendation.${RECOMMENDATION_NAME}}} %}
{{ items[0].name }} for {{ items[0].price }}

To include multiple items, reference each item individually by its index. .name and .price pull the corresponding field from the catalog.

1
2
3
4
{% assign items = {{product_recommendation.${RECOMMENDATION_NAME}}} %}
{{ items[0].name }} for {{ items[0].price }}
{{ items[1].name }} for {{ items[1].price }}
{{ items[2].name }} for {{ items[2].price }}

AI recommendations return multiple products as an array, where items[0] is the first item, items[1] is the second, and so on. If a recommendation only returns one item, attempting to reference items[1] will result in an empty field.

Adding images

If the catalog your recommendation uses includes image links, you can reference those images in your message.

• drag-and-drop
• html

1
2
{% assign items = {{product_recommendation.${RECOMMENDATION_NAME}}} %}
{{ items[0].IMAGE_URL_FIELD }}

1
2
{% assign items = {{product_recommendation.${RECOMMENDATION_NAME}}} %}
{{ items[0].IMAGE_URL_FIELD }}

1
2
{% assign items = {{product_recommendation.${RECOMMENDATION_NAME}}} %}
<img src="{{ items[0].IMAGE_URL_FIELD }}" alt="{{ items[0].name }}">

• Replace MY_RECOMMENDATION_NAME with the name of your recommendation
• Replace IMAGE_URL_FIELD with the name of the field in your catalog that contains image URLs.

Tutorial: Create abandoned cart email

In this tutorial, you’ll learn how to create a dynamic email that recommends products to users based on their preferences or behavior using Braze AI item recommendations.

Let’s say you’re a marketer at “Flash & Thread,” an online clothing retailer. You want to re-engage customers who have left items in their carts and upsell additional products. Your goal is to create an email that displays the abandoned items and personalized recommendations.

Step 1: Prepare your catalog

Your recommendation will pull items from a catalog. Follow the steps for Creating a catalog. Make sure your catalog includes these fields:

Example catalog

Step 2: Set up your recommendation

1. From your catalog, select Create recommendation.
2. Follow the steps for Creating an AI item recommendation.
3. For the recommendation type, select AI Personalized.
4. Use the catalog you just created to train the recommendation. This might take some time—you’ll get an email when the training is complete.

Step 3: Create an email

When the recommendation has finished training, you can use it in your messaging.

1. Create an email with the drag-and-drop editor.
2. In the message body, add an image block wherever you want to pull in a recommendation from the catalog.
3. Select the image block and turn on Image with Liquid in the Image properties panel.
4. Paste this Liquid snippet in the Dynamic URL field.

1
2
{% assign items = {{product_recommendation.${abandoned_cart}}} %}
{{ items[0].image_url }}

5. Below the image, add a paragraph block. Here’s where you’ll add the product name and any supporting details.
6. Paste the following Liquid snippet in the block. This pulls the name, category, color, and price of the first recommendation from the catalog, and adds them as separate lines.

1
2
3
4
5
{% assign items = {{product_recommendation.${abandoned_cart}}} %}
{{ items[0].name }}
{{ items[0].category }}
{{ items[0].color }}
${{ items[0].price }}

7. For both snippets, replace abandoned_cart with your recommendation name in Braze.
8. Double-check that item field names (``) match column names in your catalog.
9. Increment the array by one each time you repeat the block to pull in the next recommended item from the catalog. For example, the array starts with , so the next item would be.

Step 4: Preview your message

To see how your message looks like for a real user:

1. Go to the Preview & Test tab in your editor.
2. Select Random User from the dropdown.
3. Select Get Random User to fetch a user from your audience and preview how the email will appear with their data.

The preview will fully render Liquid, including AI recommendations, as long as the selected user has the required attributes or event data tied to the recommendation.

If the recommendation doesn’t appear in the preview, check the following:

• The user has interacted with relevant products or events that trained the recommendation model
• The recommendation itself has been successfully trained
• The Liquid code correctly references the correct recommendation and fields