A Storefront Taken Back Off A Platform They Could Not Edit

The site looked finished and was not. Placeholder Latin sat on live product pages. A review widget appeared on every product page and had never been connected to anything. Copy was truncated where a field had outgrown its box. None of it was quite anyone's fault, because nobody on the team could fix any of it: the storefront ran on a third party headless platform, and the deploy, the content model and the vendor app's permissions all belonged to the vendor. A wording change was a request, not an edit.

Then the brand renamed itself on a Sunday evening. The site said the old name. The wordmark said the old name. All 270 images in the email account said the old name. And the old domain could not simply be dropped, because QR codes pointing at it were already printed on product and on boxes sitting in a warehouse. Whatever got built had to keep answering at an address chosen months earlier by a print run.

What they had before we started was not a prototype in the usual sense. It was a vendor built site they could not open, plus a large capture of it that we used as the visual specification, plus an abandoned rebuild plan from earlier in the summer. So the first thing we agreed was not an architecture, it was a date: a deployed preview with a working home page and one product page by a named day, with a written fallback that would fire if we missed it. Having the fallback in writing is what made the date real rather than optimistic.

The build is a Next.js storefront in the client's own repository, deployed to their own hosting account, with the commerce platform behind it doing two jobs. Products, cart and checkout handoff come through its storefront API. Long form product content lives in structured content objects, which matters more than it sounds: it is the difference between a marketing lead editing a paragraph and a marketing lead filing a ticket. We also built a first party app to replace the outgoing vendor's permissions, 49 of them, verified one at a time rather than copied across.

Both domains cut over on one day. The old address forwards, so the printed codes still land. Afterwards we crawled 33 pages with controls and found no broken links, and the placeholder review widget came off all 12 live product pages in a single commit.

They own it. Their repository, their commerce account, their domains. The honest state is that the work is not finished: a live storefront generates defects faster than a dead one, and there are open items on the board today. What changed is who is allowed to close them. Within a day of the preview going up, the client's marketing lead came back with a list of content gaps and one question: can we go in and make some of these adjustments ourselves. That question was the whole point of the rebuild.