Reference models
Luigi's Box offers a wide scale of standard recommender models. Each model needs to be trained first. Once everything is set up, you will be given the recommender_type to use in recommendation requests described in API section. Please contact our support to discuss the options in detail. If needed, these models can be customized to your specific use case, or, we can prepare fully customized models. The table below lists the standard models.
Each recommender is personalized for users who granted cookie consent. The models consider user preferences based on sent user identifier (user id) and boost products with characteristics similar to user's past interactions. The most of the models accept also product / category / brand resource identifier (product / category / brand id - typically url or id).
| Recommender type | Short description | Location | Input |
|---|---|---|---|
test_reco |
Simple, most popular recommender to test client requests. | Arbitrary | None |
item_detail_alternatives |
Similar products / alternatives recommender preferring slightly more expensive products compared to input product (upsell). Product similarity is determined based on selected metadata. | Product detail page (PDP) | product id, user id (optional) |
item_detail_visual_alternatives |
Visually similar products / alternatives recommender preferring slightly more expensive products compared to input product (upsell). Similarity is determined based on product images, whose characteristics are crucial. | Product detail page (PDP) | product id, user id (optional) |
item_detail_complements |
Related products usually bought together (cross-sell) with an input product. By default, recommender prefers cheaper products (complements). Products are diversified by category to offer a variety of different products if possible. The general product detail complements model fits the most industry segments. Some specific segments have their own model (described below). | Product detail page (PDP) | product id, user id (optional) |
basket_popup |
Related products usually bought together (cross-sell) with a product most recently added into the basket. By default, recommender prefers cheaper products (complements). Products are diversified by category to offer a variety of different products if possible. The general basket popup model fits the most industry segments. Some specific segments have their own model (described below). | Basket pop up window | product id, user id (optional) |
basket |
Related products usually bought together (cross-sell) with all products already in the basket. By default, recommender prefers cheaper products (complements). Products are diversified by category to offer a variety of different products if possible. The general basket detail model fits the most industry segments. Some specific segments have their own model (described below). | Basket detail | product ids, user id (optional) |
favorites |
Favorite products most likely to be repeatedly bought by a user. This model is suitable for segments with consumable products, where users are likely to repeat purchases (e.g. groceries, pharmacies, b2b supplies). | Arbitrary, usually basket, homepage | user id |
user_conversion_based |
Recommender offers products related (cross-sell) to previously purchased or trends (if user has no past interactions). The idea is to help a user find more interesting products to buy. | Arbitrary, usually homepage, 404 | user id |
user_click_based |
Recommender offers products similar to those clicked but not bought recently by the user / or trends (if user has no past interactions). The idea is to help a user finish the purchase. | Arbitrary, usually homepage, 404 | user id |
category |
Most relevant trending products, from the input category for the user. Results are diversified to as many next level subcategories as possible. | Product listing page (PLP) | category id, user id (optional) |
brand |
Most relevant trendy products from input brand for the user. Results are diversified to as many categories as possible. | Product listing page (PLP) | brand id, user id (optional) |
top_categories |
Most relevant categories for the user. Used as a shortcut to user's preferred PLPs. | Arbitrary, usually homepage, 404 | user id |
trends |
Most popular products. Could be used on homepage, in PLP to offer products from this category/brand only, in PDP (in combination with some filter) to offer e.g. popular products from the same collection. Could be ordered by a different attribute than popularity (e.g., price_amount). | Arbitrary, usually homepage, 404 | user id (optional), category / brand / product id (optional) |
news |
Novel products based on date of introduction (date attribute _introduced_at). Could be used on homepage, in PLP to offer products from this category/brand only, in PDP (in combination with some filter) to offer e.g. novel products from the same collection. |
Arbitrary, usually homepage, PLP | user id (optional), category / brand / product id (optional) |
news_by_label |
Popular products with a defined label indicating novel products. By default it expects allowed products to contain "label": "is_new" in attributes. Could be used on homepage, in PLP to offer products from this category/brand only, in PDP (in combination with some filter) to offer e.g. novel products from the same collection. |
Arbitrary, usually homepage, PLP | user id (optional), category / brand / product id (optional) |
discount |
Discounted products based on discount amount (numeric attributes price_amount and price_old_amount). Could be used on homepage, in PLP to offer products from this category/brand only, in PDP (in combination with some filter) to offer e.g. discounted products from the same collection. |
Arbitrary, usually homepage, PLP | user id (optional), category / brand / product id (optional) |
discount_by_label |
Popular products with a defined label indicating discounted products. By default it expects allowed products to contain "label": "is_sale" in attributes. Could be used on homepage, in PLP to offer products from this category/brand only, in PDP (in combination with some filter) to offer e.g. discounted products from the same collection. |
Arbitrary, usually homepage, PLP | user id (optional), category / brand / product id (optional) |
last_seen |
Products most recently visited by a user, excluding already bought ones. | Arbitrary | user id |
recently_ordered |
Products most recently ordered by a user. | Arbitrary | user id |
article_items |
Most relevant products based on the input article (blog post). | Article detail page | article id, user id (optional) |
article_articles |
Most relevant articles based on the input article (blog post). | Article detail page | article id, user id (optional) |
random |
Random products. It gets more useful when combined with some filters, e.g., new products added within the last month, discounted products, products from same the brand as an input product. | Arbitrary | user id (optional), category / brand / product id (optional) |
random_news |
Random novel products based on date of introduction (date attribute _introduced_at). An example of random recommender usage. It recommends primarily random products up to 30 days old, then 60, 90. could be combined with more filters, e.g. from a certain category. |
Arbitrary | user id (optional), category / brand / product id (optional) |
As some industry segments have their specifics, we identified several segment specific models adjusting abovementioned models.
| Industry segment | Recommender type | Short description | Location | Input |
|---|---|---|---|---|
| Books & Games | item_detail_complements |
Related products usually bought together (cross-sell) with an input product. Products are diversified by category to offer a variety of different products if possible. In comparison to the general complements model, it doesn't prefer cheaper products, also it allows recommendations from exactly the same category. | Product detail page (PDP) | product id, user id (optional) |
| Books & Games | basket |
Basket detail recommender modified for books & games industry segment. Related products usually bought together (cross-sell) with all products already in the basket. Products are diversified by category to offer a variety of different products if possible. In comparison to the general basket model, it doesn't prefer cheaper products, also it allows recommendations from exactly the same category. | Basket detail | product ids, user id (optional) |
| Books & Games | basket_popup |
Basket popup recommender modified for books & games industry segment. Related products usually bought together (cross-sell) with a product most recently added into the basket. Products are diversified by category to offer a variety of different products if possible. In comparison to the general basket popup model, it doesn't prefer cheaper products, also it allows recommendations from exactly the same category. | Basket pop up window | product id, user id (optional) |
| Food & Beverages | item_detail_complements |
Related products usually bought together (cross-sell) with an input product. Products are diversified by category to offer a variety of different products if possible. In comparison to the general complements model, it prefers previously bought products and doesn't filter out expensive products. | Product detail page (PDP) | product id, user id (optional) |
| Food & Beverages | basket |
Basket detail recommender modified for fast-moving consumer goods, typically from food the and beverages industry segment. Related products usually bought together (cross-sell) with all products already in the basket. Products are diversified by category to offer a variety of different products if possible. In comparison to the general basket model, it prefers previously bought products and doesn't filter out expensive products. | Basket detail | product ids, user id (optional) |
| Food & Beverages | basket_popup |
Basket popup recommender modified for fast-moving consumer goods, typically the food and beverages industry segment. Related products usually bought together (cross-sell) with a product most recently added into the basket. Products are diversified by category to offer a variety of different products if possible. In comparison to the general basket popup model, it prefers previously bought products and doesn't filter out expensive products. | Basket pop up window | product id, user id (optional) |
| Food & Beverages | last_seen |
Products most recently visited by a user, including already bought ones. | Arbitrary | user id |
| Luxury goods & Jewelry | item_detail_complements |
Related products usually bought together (cross-sell) with an input product. Products are diversified by category to offer a variety of different products if possible. In comparison to the general complements model, it prefers products of similar price. | Product detail page (PDP) | product id, user id (optional) |
| Luxury goods & Jewelry | basket |
Basket detail recommender modified for luxury goods & jewelry industry segment. Related products usually bought together (cross-sell) with all products already in the basket. Products are diversified by category to offer a variety of different products if possible. In comparison to the general basket model, it prefers products of similar price. | Basket detail | product ids, user id (optional) |
| Luxury goods & Jewelry | basket_popup |
Basket popup recommender modified for luxury goods & jewelry industry segment. Related products usually bought together (cross-sell) with a product most recently added into the basket. Products are diversified by category to offer a variety of different products if possible. In comparison to the general basket popup model, it prefers products of similar price. | Basket pop up window | product id, user id (optional) |
| Pharma / Medical, Cosmetics & Body care | item_detail_complements |
Related products usually bought together (cross-sell) with an input product. By default, recommender prefers cheaper products (complements). Products are diversified by category to offer a variety of different products if possible. In comparison to the general complements model, it prefers products from more different categories. | Product detail page (PDP) | product id, user id (optional) |
| Pharma / Medical, Cosmetics & Body care | basket |
Basket detail recommender modified for medical, cosmetics & body care industry segments. Related products usually bought together (cross-sell) with all products already in the basket. Products are diversified by category to offer a variety of different products if possible. In comparison to the general basket model, it prefers products from more different categories. | Basket detail | product ids, user id (optional) |
| Pharma / Medical, Cosmetics & Body care | basket_popup |
Basket popup recommender modified for medical, cosmetics & body care industry segments. Related products usually bought together (cross-sell) with a product most recently added into the basket. Recommender prefers cheaper products (complements). Products are diversified by category to offer a variety of different products if possible. In comparison to the general basket popup model, it prefers products from more different categories. | Basket pop up window | product id, user id (optional) |
| Pharma / Medical, Cosmetics & Body care | last_seen |
Products most recently visited by a user, including already bought ones. | Arbitrary | user id |