Prompt
What recommender should I use for product pages?
Latest observation
Nov 13, 2026
OpenAI APIWeb search: off
For product pages, the best recommender is usually a hybrid item-to-item recommender, often with a few modules depending on your goal.
Best default choice
“Frequently bought together” / “Similar products” / “You may also like”
- Shows items related to the current product
- Works well when you have sparse user data
- Easy to explain to users
- Strong for product detail pages (PDPs)
Recommended approach by situation
- New or small catalog / little user data:
Use content-based recommendations (based on title, category, brand, attributes, embeddings) - Enough interaction data:
Use item-item collaborative filtering or matrix factorization - Best overall at scale:
Use a hybrid system combining:- content similarity
- user behavior signals
- co-purchase / co-view data
- business rules (stock, margin, availability)
What usually works best on product pages
- Similar items — if the user wants alternatives
- Frequently bought together — if you want basket building
- Complementary products — accessories, add-ons, bundles
Practical recommendation
If you’re unsure, start with:
- Item-to-item recommender based on product embeddings + co-view/co-purchase signals
- Filter by:
- in stock
- same category or compatible category
- price range
- margin/priority rules
UI tip
On a product page, one recommender is often not enough. A common setup is:
- Similar products
- Frequently bought together
- Customers also viewed
If you want, I can suggest the best recommender architecture for your catalog size and data maturity.