Quickstart: Choosing recommendation models

Introduction

This guide provides strategic advice on which recommendation models to use on different parts of your website. Choosing the right model for the right context is key to maximizing engagement, increasing order value, and creating a better user experience.

While the technical implementation is covered in other guides, this document focuses on the "why" and "where" of using Luigi's Box's powerful recommendation models.

What you'll learn

• Which models are best suited for product pages, shopping carts, and homepages.
• How to use different models to achieve specific business goals like upselling, cross-selling, and personalization.
• A framework for thinking about your overall recommendation strategy.

Who is this guide for

• Developers who are planning their recommendation strategy.
• Users looking to understand the business value of different recommendation types.

Choosing models by page type

The most effective recommendation strategies are context-aware. A recommendation that works well on a product page might not be the best choice for the homepage. Here's a breakdown of the best models for key pages on your site.

1. The product detail page (PDP)

Goal: Keep the user engaged with your catalog and encourage them to either upgrade their choice (upsell) or add more items to their cart (cross-sell).

• For "You might also like" (alternatives):

  • Model: item_detail_alternatives
  • Why: This model is perfect for showing similar products. It's trained to find alternatives that are often slightly more expensive, which can help increase the average order value. It helps users find the perfect product if the current one isn't quite right.

• For "Frequently bought together" (complements):

  • Model: item_detail_complements
  • Why: This is the classic cross-sell model. It recommends items that are commonly purchased with the product being viewed, like accessories, batteries, or matching items. This is a highly effective way to increase the number of items per order.

2. The shopping cart / basket page

Goal: Increase the final order value just before the user begins to check out.

• Model: basket
• Why: This powerful model looks at all items currently in the user's cart and finds products that are complementary to the entire order. It's smarter than a simple PDP complement model because it considers the combination of products, leading to more relevant last-minute suggestions.

3. The homepage

Goal: Immediately capture the user's interest, whether they are a new or returning visitor.

• For new visitors ("Trending Now"):

  • Model: trends
  • Why: New visitors don't have a history with your site, so you can't personalize their experience yet. The trends model shows them your most popular products, providing social proof and a great starting point for exploration.

• For returning visitors ("Picked for you"):

  • Model: user_click_based or favorites
  • Why: For returning users, personalization is key. The user_click_based model recommends products similar to what they've recently viewed but haven't purchased. The favorites model is excellent for consumable goods, reminding users to repurchase items they've bought before.

4. 404 and "No search Results" pages

Goal: Recover a broken user journey and guide the user back to relevant products.

• Model: trends or user_conversion_based
• Why: When a user hits a dead end, you want to show them something engaging. Displaying your most popular products (trends) is a safe and effective bet. Alternatively, for a known user, user_conversion_based can show them products related to their past purchases, offering a personalized path forward.

Next steps

Now that you have a basic understanding of recommender models.