We built Revlance because rule-based merchandising was failing in plain sight

I spent four years managing merchandising rules at a mid-market DTC platform. We had a 40,000-SKU catalog and five static rules — written in a sprint four years before I joined — that decided what every shopper saw on the homepage. The rules never changed because nobody had time to change them. And the result was that a shopper who spent 12 minutes hovering on outdoor furniture saw the same twelve homepage tiles as someone who clicked immediately on seasonal decor.

I watched that catalog get served as if every visitor had identical preferences. The analytics showed the problem — 65% of visitors left the homepage within 30 seconds — but the solution always felt like a data science project we couldn't staff. I left that role convinced the problem was solvable for teams without ML headcount. You just needed the right signals and a model that started working in minutes, not months.

I met Dawson at a Shopify Partners meetup in Atlanta in early 2024. He'd spent years building recommendation infrastructure at a logistics platform — he understood behavioral signal modeling from the infrastructure side. We started building Revlance that summer.