A couple of cookies keep the site running; analytics help us see what’s useful. You choose what we keep, read the policy.
The old site was desktop-only, which is a real problem when most contractor traffic is on a phone. It had near-zero organic search presence, and years of buildup had left it bloated: a content audit found 122 URLs, around half of them duplicates or dead pages carrying almost no traffic. Underneath all of it sat the real challenge: a catalog too complex for any off-the-shelf store.
Pricing lived on sheets printed from an inventory system and walked around the store by hand, so none of it was web-ready. And the new owners, who had just bought the business, had no way to manage their own catalog. The previous vendor even held the domain and email, which took real persistence to release.
“The person going on the website is more likely the homeowner, designer, and weekend hobbyist. But the person making the purchase is the contractor.”
Operations, Saint Charles Hardwoods
The millwork ends up in high-end homes, so the site had to earn trust with specifiers and homeowners while giving contractors the fastest possible path to a quote. The owners were clear about the brief: don't over-engineer it, ship a good first version, and let us run it ourselves afterward.
Phase 1 was scoped as a lead-generation website. The real engineering, and the center of the work, is the catalog system underneath it.
A custom admin where the owners manage everything themselves: categories, products, specifications, images, variants, and what is featured or in stock. Built custom rather than as an off-the-shelf store, because the product model does not fit one.
A proto-cart: customers add products to a quote basket and submit one request with the details attached. Custom products use a free-form quote form instead of the variant picker. A deliberate bridge toward the Phase 2 quoting work.
A new homepage and brand, product category and detail pages, and a dedicated “Our Mill” page as the differentiator. Click-to-call CTAs throughout, because most of the people buying are on a phone.
A local SEO foundation: clean titles and metadata, structured headings, analytics and Search Console, and a URL structure ready for Phase 2. Plus a CMS training handover so the owners can run it.
React with Zustand on the client, Lovable Cloud (Postgres, Storage, Auth) on the server. There is no custom server: the browser talks to the database directly, with row-level security deciding who can read and write what.
Sign-in by email or Google. Admin access is an allowlist: one database check gates every write, so granting or revoking it is a one-row change. Sensitive price tiers are hidden from the public read path entirely.
Categories form a tree. Variant axes like thickness and length generate their combinations automatically, and pricing units, board-foot, lineal-foot, square-foot, piece and sheet, are first-class. That last detail is what an off-the-shelf store gets wrong.
Edits apply optimistically and save in the background through a debounced, self-retrying queue, with a Saving and Saved indicator and a short-window undo. Images are resized in the browser before they upload.
Every meaningful admin action is written to an append-only audit log. The platform runs daily database backups with 14-day retention, and the image library is reproducible from the client's own Drive.
The quoting and pricing complexity, institutional-knowledge pricing, board-foot versus lineal-foot, many species and variants, and a deliberate buffer between a quote and a purchase, is more than an off-the-shelf store handles well. Modern AI-assisted tooling made building it custom genuinely faster than the manual QA such a system used to demand. The harder right thing for the business they actually run.
Truly custom software trades convenience for control. We were candid that the catalog system is the part most likely to need ongoing QA.
So we built the safety net first, snapshots, an audit log, backups, and a release protocol, to make that trade safe.
Rather than present static mockups, we defined four design directions and built three of them as live, clickable prototypes the client could test on their own phones. The direction was anchored to brands they reacted to, so we ruled out the wrong turns early instead of discovering them late.
We started from the client's real, messy pricing data instead of guessing, scoped a full QA week with their own staff before go-live, and put a release protocol in place so their catalog edits are never overwritten by a code push.
The catalog system is built and in the owners' hands. They manage categories, products, images, featured items and variants themselves, with publish and revert safety built in. QA was run jointly with their own staff before launch.
Between 12 August and 26 September 2026 the new quote flow received 60 requests. Search brought about four in five of the ones analytics could attribute, and about half of the real visitors were on a phone, which is why the site was built mobile-first.
Phase 2 started in September 2026: making the business easier to find in Google, local listings and AI assistants for people who don't know it yet. After that comes the real bottleneck, quoting. We have proposed mapping how custom work is priced today, the first step toward automating it with a person kept in the loop for judgment. The hard part is structuring the team's tacit pricing knowledge into something a model can use reliably, the same problem we solved on Fixlab.
The everyday walk-in quote, priced in minutes instead of by hand.
The large-project bid, where pricing by hand can take weeks.
Tacit pricing knowledge, structured so a model can apply it.
A small product studio. We prototype, automate, and ship.
Small is the feature: you work directly with the people building the thing, and we care more about something that holds up in production than something that looks good in a pitch.