personalization-engine

Install LimeSpot Personalizer or Frequently Bought Together by Code Black Belt from the Shopify App Store. These apps analyze order history automatically and place recommendation widgets on product pages, cart, and homepage without custom code.

Pre-compute a co-purchase matrix nightly in PostgreSQL by counting product pairs within the same orders, then serve Frequently Bought Together queries ordered by co-purchase count. For browsing-history recommendations, build product vectors from category, price, and tags, then rank candidates by cosine similarity.

LimeSpot provides a full personalization suite covering homepage, PDP, cart, and post-purchase placements with ML-based algorithms. Frequently Bought Together by Code Black Belt is purpose-built for the PDP bundle section and requires no manual configuration.

Recommendations must exclude out-of-stock products, the currently viewed product, and recently purchased items. Filter inventory at query time rather than model build time, and pass excludeProductIds containing the current product to every recommendation function.

Cold-start visitors should see bestsellers as a fallback. Once a visitor views even one product, constrain recommendations to that product's category to keep results relevant until more browsing signals accumulate.

Only build custom when your catalog size, traffic volume, or recommendation logic exceeds what app-based solutions support. Apps like LimeSpot handle cold start, exclusions, and model refresh edge cases that take weeks to replicate correctly.