This document provides an overview of how the Barilliance on-site product recommendations work and how the recommendations in each widget (a block with the recommended items) are determined.

Data Sources
There are 3 main sources of data that power the recommendations:

1. Aggregated data that the system collects by analyzing the website traffic
2. User specific data that is used to personalize the recommendations
3. Static product data that is supplied by the client in the product feed

The recommendations engine fetches the relevant data from the system, personalizes it for the visitor and filters it to make sure that only the most relevant products for the current context are displayed.

Aggregated  data
During the set-up process Barilliance adapts the software to your site so it can correctly identify visitor activity. By visitor activity we mean things like:

● Product/category page views (identifying the exact category and product)
● Adding a product to the cart
● A completed purchase
● Using the internal search function
● Etc…

The system analyzes the data and uses it to build a relationship map between the different products (the classic example being “customers who bought this product also bought…”)

User  specific  data
This refers to things such as:
● The users recently viewed products
● The categories they showed interest in
● Which products they’ve added to the cart
● Which product they purchased
● Etc…

Static  product  data
​ 
This is data that is usually supplied to Barilliance through the product feed and includes things like:
● The product category
● Price
● Availability
● Brand
● Etc..

Default widget logic
The following describes the logic used to power the widgets in the most common locations typically used for product recommendations. In most cases the actual products selected will internally be a mix of more than one algorithm in descending priority until the system has enough products to fill the widget.

Homepage
​Visitors to the homepage are either new or returning visitors. New visitors will see the website’s top sellers. Returning visitors will see personalized recommendations based on their recent activity. This includes:
- Products related to their recently purchased items
- Products related to their recently viewed products
- Top sellers from their recently viewed categories

A recent activity widget that shows what people are buying/adding to cart and the location of each user can be added

Product  Page
​The product page will typically have two widgets:
a. Alternatives - displays products of the same or similar category to the currently viewed product. This is the more relevant widget on the product page since at that point the visitor still hasn’t yet decided to purchase the product he is viewing. The widget includes algorithm such as:

- Products viewed by other visitors who viewed the same product
- Product of the same category with similar price range
- The category’s top sellers

- What do customers ultimately buy after viewing this product

b. Cross-sell - products that are of a different category to the currently viewed product. These are mostly products that were purchased or considered for purchase by visitors who viewed or purchased the current product

Category  page
If the visitor already showed some interest in products of that category then personalized recommendations from that category will be displayed, otherwise the top sellers of the category will be recommended.

Note about category pages: in order to limit recommendations to the products of the same category the category data in the product feed must match the category name listed on the category page. If that is not the case then more general recommendations will be displayed (similar to the homepage recommendations)

Cart  page

- Products that were purchased or considered for purchase by visitors who viewed or purchased products in the user's cart (could be also just based on the last item added to the cart)

- Top sellers from the categories of the products in the user's cart

Search  page
​Barilliance tracks the search term used and the products that were clicked after a search using that term. The default logic will recommend the most clicked products for that search term with a fallback to personalized recommendations based on the visitor’s recent activity.

Order confirmation page
Products that were purchased or considered for purchase by visitors who viewed or purchased the current product

All pages

- User recently viewed
- User recently added to cart
- User recently purchased