Building a Custom Product Recommendation Engine That Increased AOV by 28%

The Challenge

A DTC health supplements brand with $8M annual revenue was hitting a growth ceiling. Their Shopify store used basic "customers also bought" recommendations powered by simple co-purchase data, which surfaced obvious pairings but missed the nuanced cross-sell opportunities in their 340-SKU catalog. Cart abandonment sat at 71%. Average order value had been flat at $67 for two years despite adding new product lines. The marketing team was running blanket email campaigns to their 85K subscriber list with no personalization beyond first name — open rates had dropped to 14% and trending downward.

Our Approach

We built a custom recommendation system that goes beyond purchase history. The engine combines browsing behavior patterns, quiz responses (the brand runs a "find your routine" quiz), purchase frequency data, and product attribute matching (ingredients, health goals, dietary restrictions) to generate personalized bundles. We integrated it at four touchpoints: product pages ("complete your routine"), cart drawer ("frequently paired with your selections"), post-purchase email sequences, and a personalized storefront section for returning visitors. The engine runs on a lightweight ML model deployed as a serverless API that queries against their Shopify data via a custom app. We also rebuilt their email segmentation: subscribers are now grouped by health goal, purchase recency, and predicted next-purchase timing.

The Results