Contextual Algorithms

In Contextual Algorithms, product recommendations are based on the current context of the user event (e.g. product category of the currently visited page), and product relations (e.g. products that are viewed or purchased together). 

The algorithms covered under Contextual Filtering are:

• Viewed Together
• Purchased Together
• Location-based Top Sellers
• Checkout Recommendation
• Most Popular Items of the Category
• Most Valuable Items of the Category
• Substitute Products
• Complementary Products
• Recently viewed
• Purchased with Last Purchased

Viewed Together

The algorithm generates recommendations based on the products that have been visited in the same sessions and the same locale (the language of the website that the user visits) during the past 30 days. After generating recommendations, it arranges the results according to their visit frequency or popularity. This way, users can discover complementary or alternative products related to the one they are currently viewing. This increases discovery rates and the chance to grab users’ attention when they don’t have a target product.

• Page Type: Product/Article Page, Cart Page
• Example Use Case: Showing items that have grabbed visitors' interests during previous sessions creates the potential to bring the visitors closer to adding the viewed or recommended products to their cart and eventually increased conversions.
• Fallbacks: Most popular items of the subcategory, Most popular items
• Prerequisites: 30 days of product views
• Maximum Number of Products to Recommend (in the same variant): 50

Purchased Together

The algorithm generates recommendations based on the products that have been purchased in the same sessions and in the same locale (the language of the website that the user visits) during the past 30 days. After generating recommendations, the algorithm orders the results according to their purchase frequency. With this algorithm, you are able to apply the purchase patterns of their users to their strategies.  

• Page Type: Product/Article Page, Cart Page
• Example Use Case: Expanding user's cart amount by showing products that have been bought together by other users and increasing cross-category selling options. With a feeling of “Users who purchased this item also purchased…” you can use both product popularity impact and preferable price/quality balance impact on their users.
• Fallbacks: Top Sellers of Category, Top Sellers
• Prerequisites: 30 days of product purchase.
• Maximum Number of Products to Recommend (in the same variant): 50

Location based Top Sellers

Location-based top sellers algorithm allows you to serve product recommendations to each visitor of the website, analyzing the current location of a user (IP address geo-location) and serving the most sold products in that location.

• Page Type: Home Page, Product/Article Pages, Category Pages, Cart Pages
• Example Use Case: Display the products that are able to complete the user flow of discovery, add to cart, and purchase. Especially for local products and demographic trends, (e.g. student populations around campus) you can increase purchase rates with the help of geographical personalization.

• Fallbacks: Top Sellers Of Parent Category
• Prerequisites: 30 days of product purchase data.
• Maximum Number of Products to Recommend (in the same variant): 90

Checkout Recommendation

Taking basket amount into account, this algorithm recommends only the products that fulfill the campaign amount along with Purchased Together algorithm. 

For instance, a user has 2 items in their basket, amount up to $100 and there is free a shipping campaign for orders exceeding $150. In this case, items purchased together with the items in the basket that are at least $50 are recommended. If the basket amount of the user already exceeds $150, there will be no product recommendation. 

• Page Type: Cart Page
• Example Use Case: In cart pages, in order to increase AOV, you can organize some campaigns such as free shipping campaigns - "Order at least $X, get free shipping". Taking basket amount into consideration, only products that are fulfilling the campaign amount are recommended with checkout recommendation.
• Fallbacks: NA
• Prerequisites: NA
• Maximum Number of Products to Recommend (in the same variant): 90

Most Popular Items of the Category

The algorithm generates recommendations based on page view counts during the last 30 days in the same locale (the language of the website that the user visits). Most Popular Items of the Category works with the same logic but brings results from the same category with the product or category that is currently being viewed. After generating recommendations, the algorithms order the results with descending page view counts and place them on the smart recommender widget.

• Page Type: Product/Article Pages, Category Pages
• Example Use Case: Display the most popular products and promote the hottest products that are viewed or sold most on the website. Display the products that gather the attention of the users from the same category, and apply filters to show the products that have a higher price to create upsell opportunities from both the same category or cross categories.

• Fallbacks: Top Sellers Of Parent Category
• Prerequisites: 30 days of product views.
• Maximum Number of Products to Recommend (in the same variant): 90

Most Valuable Items of the Category

This algorithm recommends products that generate higher revenue across your site, considering both the contribution to revenue and revenue per visit. All users see the same recommendation.

• Page Type: All Pages, Product/Article Pages, Category Pages, Cart Pages
• Example Use Case: Promote more revenue-generated products on your website. 
• Fallbacks: Most Valuable Of Parent Category
• Prerequisites: "Product Value Scoring" and "Most Valuable Products" should be enabled for your account to be able to use the Most Valuable Products algorithm.
• Maximum Number of Products to Recommend (in the same variant): 90

Substitute Products

It recommends products that are similar and helps the product discovery process. Substitute Products algorithm considers product similarity and price proximity.

• Page Type: Product/Article Page, Cart Page
• Example Use Case: Substitute products that can be purchased instead of each other. Substitutable products are those that are interchangeable—such as one t-shirt for another.
• Fallbacks: Viewed Together, Most Viewed Of Category
• Prerequisites: The price of the items recommended should be within 0.5-1.5x.
• Maximum Number of Products to Recommend (in the same variant): 90

Complementary Products

This algorithm utilizes a unique approach to provide product recommendations that complement users' existing preferences and purchases. It examines products frequently bought together or identified as complementary through user interactions, proposing items to enrich the overall shopping journey for users. It considers item performance to create complementary products.

• Page Type: Product/Article Page, Cart Page
• Example Use Case: Complementary products that can be purchased in addition to each other. For instance, laptops and mouse are complementing each other, also sugar and tea can be another example.
• Fallbacks: Top Purchased Of Category, Top Purchased
• Prerequisites: NA
• Maximum Number of Products to Recommend (in the same variant): 90

Recently Viewed Items

The Recently Viewed Items algorithm is one of the personalized algorithms. It tracks the user’s product view behavior that is collected from the Unified Customer Database (UCD). It means it collects data from both Web and Mobile events. It is a user-based algorithm and available on Web Smart Recommender, API-based Recommender, and App Recommender. 

• Page Type: All Available Pages, Product/Article Page, Cart Page, All Pages, Category Page
• Fallbacks: NA. We highly suggest you to use a combination of products and enable the “Hide the recommendation if there are not enough products to recommend” option. If you want it to be displayed in all cases, (when Hide option is not selected), then, when a new (or non-login) user displays even a single product, the widget would be visible with a single product. 
• Example Use Case: On the cart screen, you can display the user’s recently viewed products so that you can increase the AOV through the end of the purchasing funnel.
• Prerequisites: Having product view activity in the last 30 days.
• Maximum Number of Products to Recommend (in the same variant): 50

This algorithm is updated once a user visits a new product and the update is reflected to the UCD. It takes the last 30 days of activity. Products that have been viewed before this period are not taken into consideration. The user’s previous product views as well as the current session views are combined in the algorithm. 

You can shuffle the recommendations. However, if you do not shuffle them, then the most recently viewed product is displayed first. All filters and exclusions that Insider has on the Smart Recommender are available except for the Exclude Recently Viewed Items option since its main purpose is to track this specific event (Web and API-based only).

For the App Recommender, filters, and exclusions are limited to App Recommender’s capabilities. 

Purchased with Last Purchased

The Purchased with Last Purchased algorithm is one of the personalized algorithms. It recommends the products that are purchased along with the user’s last purchased product. Purchase events can be collected from Web, Mobile, and Offline (CRM) UCD events. It is available on Web Smart Recommender, API-based Recommender, and App Recommender.

• Page Type: All Available Pages ( Product/Article Page, Cart Page, All Pages, Category Page)
• Fallbacks: Top Sellers of the Category, Top Sellers
  • Top Sellers of the Category is the primary fallback. It uses the Category that the user is browsing at the moment of recommendation creation.
  • If the user is not in a specific category, then, the algorithm will use the secondary fallback, the “Top Sellers”.
• Example Use Case: You can display products that are purchased together by the user’s last purchase to provide personalization on a page.
• Prerequisites: 30 days of product purchase.
• Maximum Number of Products to Recommend (in the same variant): 50

The algorithm takes the latest purchased product ID and creates recommendations based on this item (A single product). It checks if there is a purchase in the last 30 days. Products that have been purchased before this period are not taken into consideration.

All filters and exclusions that Insider has on the Smart Recommender are available (Web and API-based only). For the App Recommender, filters, and exclusions are limited to App Recommender’s capabilities.