An online retailer with a large but generic catalog was showing every visitor the same "you might also like" shelf. We built a recommendation engine that actually learns from behavior.
0%
Average order value uplift
2.1M+
Items scored
0%
Recommendation accuracy
The challenge
A catalog too big to browse by hand.
With over 30,000 SKUs and a static "related products" widget, most of the catalog was invisible to most shoppers. Merchandising by hand couldn't keep up, and conversion on browse pages had been flat for a year.
The approach
Recommendations that learn per visit.
We trained a recommendation model on browsing and purchase history, then wired it into every product page, cart, and post-purchase email — each recommendation shelf re-ranks itself in real time as a visitor browses.
What we built
1
Real-time scoring pipeline
Every product gets re-scored against live session behavior, not last night's batch job.
2
On-site recommendation widgets
"You might also like," "frequently bought together," and cart cross-sells, all powered by the same engine.
3
Email personalization
Post-purchase and abandoned-cart emails now recommend products based on individual behavior, not generic bestsellers.
4
Merchandising dashboard
The marketing team can see what the model is recommending and why, and override it when needed.
★★★★★
Our catalog finally feels alive. Products that used to sit on page 12 are now showing up in front of the right shoppers.
Your email client should have opened with everything pre-filled — hit send and we'll get back to you within a day. If nothing opened, email us directly at hello@stckd.studio.