The brand's commercial engine was the weekly drop: a limited release of new pieces announced in advance, sold online and backed by 12 physical showrooms. That model concentrates a week of demand into a few minutes, and the templated Shopify store underneath was never designed for traffic that shape. Roughly 1 in 3 drops ended with collection pages crashing at the exact moment buyers arrived, turning the launch minutes into an apology.

Even when the pages stayed up, the buying experience worked against the brand. Carts held no reservation, so several buyers could carry the same limited piece towards checkout at once. Sometimes only the fastest won; sometimes every order went through and the piece oversold, and the team then had to unwind orders for customers who believed they had secured something scarce. For a fine jewellery brand, that is a damaging conversation to keep repeating.

Inventory allocation compounded the damage. Pieces were split across regions with no demand signal behind the split, so allocation was effectively a guess. Some markets sold out within minutes of a drop opening while others held identical stock for weeks, and sell-through in the first 48 hours swung anywhere from 55% to 100% depending on the market. Sold-out pieces captured nothing: buyers who missed out simply left, taking their demand signal with them.

We rebuilt the storefront on Shopify Plus and treated the drop itself as the thing being engineered, not an event the site merely had to survive. The custom theme is built around countdown launches, so every release opens at a known moment on pages designed for the first-minute rush, and collection pages are cached and pre-warmed through Cloudflare before each countdown ends, so the heaviest traffic of the week lands on pages that are already fast.

Overselling was fixed at the platform level rather than in the theme. Cart-level reservations built with Shopify Functions hold a piece from the moment it enters a cart, so checkout stopped being a race between buyers carrying the same item. The interactive storefront runs on React and Tailwind CSS over the Shopify Storefront API, keeping availability states live and honest while a drop unfolds.

Allocation moved from guesswork to signal. Each market's waitlist signups now inform how much of each piece that region receives, so stock follows demonstrated demand rather than a flat split. And every sold-out piece feeds waitlist and back-in-stock flows in Klaviyo, so the demand a drop could not serve is captured for the next one instead of walking away.