Product data

Keeping your product data consistent and in the right format is key when working with the Raptor Recommendation engine. Please read this article before you start integrating Raptor, and make sure that you are following our guidelines.

PIMCategory

In the PIMCategory field, you submit the product category path of your items according to the product categorization system of your PIM system. All the products within the category should be similar. 

A PIMCategory specifies the position of each product within the category hierarchy. It is a ; list of subcategories, beginning with the top-level category.

For example, a sports store may have the following category hierarchy:

Clothes
      Running-clothes
                   Running-jackets

Clothes;Running-clothes;Running-Jackets


Items with multiple categories

Sometimes, you may want to categorize products into multiple categories. For example, if our sports store sells t-shirts, which can be mapped as both Clothes;T-shirts&tops;t-shirts and Outdoor;Outdoorclothes;t-shirts. As such, the elements of lists should be separated by a comma (,) within the category field, and send like this Clothes;T-shirts& tops;t-shirts,Outdoor;Outdoorclothes;t-shirts.

Use cases of PIM categories

Raptor uses PIM categories to find synergies between your webshop's categories and products, which are an important building block in the GetSimilarItems and GetRelatedItems strategies.

GetSimilarItems (Other customers have also seen)

Raptor's most used recommendation strategy is GetSimilarItems. This module is typically seen on product description pages. Its function is to find similar items / look a like items to the product the user is visiting. This is done to provide the user with alternatives that ensure they find what they are looking for and drive conversion.
GetSimilarItems can be set to only show products from the same category as the product the user is visiting, these categories are feed from the PIMcategory.

GetRelatedItems (Other customers have also purchased)

For low-demand items, or webshops where the assortment is frequently changed, it can be difficult to build meaningful purchase relationships between products. This is why categories play an important role in the recommendation process. Categories are used to feed Raptor's recommendation engine with information about which categories are typically purchased with the product the user is visiting, and which categories are purchased together. The categories are essential for building a comprehensive data foundation that ensures best-in-class recommendations.

Best Practice for Tracking Category Structure

To ensure the optimal performance of the Raptor recommendation engine and modules such as GetSimilarItems and GetRelatedItems, Raptor recommend structuring your categories with a hierarchy of 3-4 levels. Below is more detailed explanation of why this range is effective, along with additional guidance on abstraction levels:

Why 3-4 Levels?

1. Too Few Levels (1-2):
   • Categories become overly heterogeneous, grouping products that differ significantly in function and use.
   • Example: A single "Tools" category might include screwdrivers, hammers, and drills, making it difficult to identify similar products.
2. Too Many Levels (5+):
   • Categories become overly granular, leading to very narrow groups where products are almost identical.
   • Example: Separate categories for "Flathead Screwdrivers" and "Philips Screwdrivers".

Abstraction Level for Categories

When defining your category hierarchy, consider the level at which products should be grouped:

• GetSimilarItems:
  • Designed to recommend items that are closely related. Therefore, categories should group items that serve a similar purpose or can be used in place of each other in certain contexts.
  • Example: All screwdrivers (flathead and Phillips) should fall under a single "Screwdrivers" category to ensure relevant recommendations within this group.
• GetRelatedItems:
  • Designed to inspire customers with complementary products from other categories.
  • Example: A "Screwdrivers" category might relate to "Screws," "Toolboxes," or "Power Tools" for cross-category recommendations.