192 product pages, three custom apps, one testing programme.

Paid traffic landed on product pages that tried to say everything at once. No clear offer above the fold, proof buried three scrolls down, and a mobile layout where the add-to-cart disappeared as soon as you started reading. Checkout ended at the order, so nothing captured the buyer at the one moment they had already decided to spend.

Rebuilt the product pages around a single offer, with bundle and subscription selectors and a sticky mobile add-to-cart, then a standing library of advertorials and landing pages shipped as testable variants. Where the platform got in the way I wrote custom Shopify apps rather than working around it: the dental membership portal, a Buy X Get Y app that adds free gifts inside checkout so they survive Shop Pay and express checkout, and a discount engine that collapses the whole free-gift programme into a single rule. Separately, audited and cleaned a theme that had drifted into hardcoded IDs, duplicate scripts, and page-builder blocks carrying their own inline code. AI carried a real share of the throughput here: Claude Code and ChatGPT in the build, and the generation features in GemPages and Replo for page variants, with everything reviewed before it shipped.

The clearest result is a controlled one: adding express payment methods to the product page lifted conversion 24.89% in an A/B test, with the interval entirely above zero, so ad spend cannot explain it. Across the period I owned the storefront the store's conversion rate rose 59% to 3.59%, which is shared work, since offers and pricing move that rate too. Revenue on a store doing roughly $2M a month was a team result and I do not claim it. What is plainly mine: 192 product pages and pre-landers shipped as one testing programme, with the reasons-to-buy page reaching its eighteenth iteration, the membership, free-gift and discount apps in production, and a post-purchase upsell ladder on its own domain.