The main recommendation algorithms available through the Recommendation API are categorized by type and methodology. Each algorithm can be called via its dedicated endpoint.
Algorithm | Definition | Endpoint | Abbreviation |
|---|---|---|---|
An automated recommendation algorithm that automatically brings the best-mixed strategy combination by testing most popular items, top sellers, user-based, highest discounted, new arrivals, and trending algorithms for better conversion rates. | /v2/chef | chef | |
Recommends products that complement the one currently viewed (e.g., a belt for trousers, a case for a phone, a lamp for a sofa). Uses a catalog-aware map of complementary category relationships generated by a language model from the partner's category tree, so it works even on catalogs with little or no purchase history. Works best on product detail and cart pages. | /v2/complementary | cp | |
Recommends items by sorting items based on their discount ratio. Recommends products ordered from the highest discount to the lowest. The discount ratio is calculated separately for each currency type. | /v2/highest-discounted | hdop | |
Brings details of manually specified products. Only in-stock products are returned. Enables showcasing of specific products or content from a curated list specified in campaign configuration. | /v2/manual-merchandising | mm | |
Creates a customized mixed recommendation strategy using different recommendation types. Allows the use of multiple algorithms for each slot of the Recommendation Widget in a single request. | /v2/mixed | mixed | |
Recommends items by analyzing the most popular products by page views. Generates recommendations based on page view counts during the last 30 days. Works best on main, category, and product pages. | /v2/most-popular | mvop | |
Recommends items by their contribution to total revenue. Recommends products that generate more revenue across your site based on contribution to revenue and revenue per visit. | /v2/most-valuable | mvpop | |
Brings products newly added to the website. Recommends products in order of their publish date. For the Publisher vertical, the updated time is used for newly released articles. | /v2/new-arrivals | naop | |
Recommends complementary products purchased by other users alongside the user's purchases. Generates recommendations based on products purchased in the same sessions and locale during the past 30 days, ordered by purchase frequency. | /v2/purchased-together | btb | |
Allows users to create campaigns highlighting recent product views. Enables users to re-engage with products based on their historical behavior. Returns only the user's recently viewed products. | /v2/recently-viewed | rvp | |
Recommends products that match the description of the one currently viewed by comparing product text and attributes, such as name, brand, category, material, and other configured metadata. Replaces the Substitute Products strategy, works best on product detail and cart pages, and requires no prior user activity. | v2/similar | sim | |
Recommends similar products using a collaborative filtering approach. Considers product similarity and price proximity to help with product discovery on product and cart pages. | /v2/substitute | sp | |
Recommends products in order of their purchase counts for the last 30 days. Works best on the main page. Falls back to your most purchased category. | /v2/top-sellers | mpop | |
Recommends items using a scoring system. Identifies this week's trending items compared to those in the previous week by scoring items based on weekly view and purchase information. | /v2/trending | tpop | |
Recommends items by finding similar users to the current user. Generates recommendations based on user behavior and product popularity. Uses a user-product-rating matrix based on visits, purchases, and add-to-carts within the last 30 days. | /v2/user-based | ub | |
Recommends products by analyzing the most recent interactions of the current user. Generates personalized recommendations informed by real-time user behaviors and evolving preferences using a deep learning-based transformer model. | /v2/user-engagement | ue | |
Recommends items by finding similar products to those visited by the user. Generates recommendations based on products visited in the same sessions and in the same locale within the past 30 days, ordered by visit frequency. | /v2/viewed-together | vtv | |
Recommends products that look like the one currently viewed by comparing product images using a multimodal AI model, refined with light product metadata. Works best on visually-driven verticals such as fashion, home décor, and accessories. | v2/visually-similar | vs |