Build a Recommendation Scheme

You can build your recommendation scheme from the Content tab of the recommendation studio by adding one or more recommendation blocks as required.

Note: Recommended items in a recommendation block are displayed based on the selected preview user, so it is important to select the right preview user.

• Recommendation blocks are displayed on the right side of the Content tab.
• When you select a recommendation block, you can see all the options to refine the recommendation.
• If you have multiple recommendation blocks, drag and drop a block to reorder.
• You can also clone or delete a block.

Add a recommendation block

A recommendation block provides product recommendations based on a particular recipe. When you add a recommendation block to your scheme, you can select one of the many pre-configured recipes as the base for the block. You can then refine the recommendations based on your requirements.

1. Click NEW BLOCK to add a recommendation block.
2. Click a recipe to select it.
3. Click SELECT RECIPE to add it to your scheme.
4. Save the scheme. You can now see the newly added recommendation block on the right side of the Content tab.
5. Select the recommendation block and update any of the inputs, item exclusions, item filters, or output settings based on your requirements.

Inputs

The input for the recommendation block can be an event, a user attribute, or a product catalog attribute.

• View the attribute that sets the context for this recommendation block. 
• The settings for the input depend on the type of input. You can tweak the settings as required.

Example: Abandoned cart items block where the input is the timeline on add_to_cart event.

Example: Matching items block where the input is a user attribute, in this case a favorite author.

Example: Popular items in a category block where the input is a category name.

Item Exclusions

You can exclude certain items from the recommendation based on the following criteria:

• Event activity in past 366 days
• Messaging activity in past 31 days

Min number of previous messages is the threshold on how many times a specific item was recommended (sent) previously in the last X days to the same user. For example, if Min number of previous messages is set to 2, then the same item cannot be recommended to a user more than 2 times (2 sends) in the last X days.

Item Filters

You can use various item attribute filters to control which items are displayed to your users.

You can set the filters to:

• Matches All of These
• Matches Any of These
• Matches None of These

For example, 

• Limit the items to a particular category of items by using the Matches All filter
• Eliminate a particular item based on the SKU or a book based on the title by using the Matches None filter.
• Include particular brands by using the Matches Any filter.

Note: To filter using the category attribute, use the contains operator instead of the equals operator. This is because the catalog category is saved as an array.

Output

With the output settings you can determine the rank and order in which the items are listed in a recommendation block. You can also define the number of items to include in the block.

Rank by:

• Product Aggregate
  • Shuffle
  • Rank by popularity within a time period
  • Auto Optimize with AI
• Product Attribute
  • Price
  • Start_date
  • num_reviews

Order by:

• Ascending
• Descending

Use the “Fail if Empty” option to suppress messaging users if this particular block is empty. For more information, see Handle empty recommendations.

Recommendations based on parent product

Parent Items offers a powerful abstraction for organizing your catalogs such that collections, bundles, variants can be modeled with their relationships fully represented. 

Consider a media streaming company where subscribers can watch shows that have many seasons and episodes. The parent item is the show that is streaming. If a customer watches some episodes from a particular comedy series, it might not make sense to recommend the next episode of the same series. Instead, you could recommend other similar comedy series to your customer.

Another example is a retailer that has many color variants of the same base item. The parent item is the base item/bundle. You can recommend a similar product to the customer.

• Parent Item in Inputs: The parent item of the item in the input event is considered when items are recommended. To consider the actual item and the parent item, add one row for each item.
• Parent Item in Event based item exclusions: When you exclude an item based on an event, the parent item or category is excluded. To exclude both the item and the parent item, add one row for each item.
• Parent Item in Output: The parent items of the final results are recommended in the recommendation block.

Handle empty recommendations

When the “Fail if Empty” option is selected, messages to users are suppressed if this particular block is empty. The recommendation is not executed and an error code is sent to suppress messages from being sent. 

Use this flag in conjunction with the “Do not message users if any of the recommendation blocks are empty” option in messaging templates. By combining both flags, you can have better control over when to send messages. 

Caution: If you are unsure, don't select the “Fail if Empty” option.

Example

Consider the following example to understand how the selected options affect message delivery.

User, Jane Smith, qualifies for Campaign A with an email trigger which uses Template Y. Template Y uses Recommendation Scheme X that consists for two recommendation blocks, B1 and B2.