Marchetti Group had grown from a single storefront into three offices, but their site had grown with it only by accretion — a new page bolted on every time someone joined, agent bios in four different templates, and a listings section that pulled from a nightly CSV somebody had to remember to upload.

The brief was blunt: make it fast, make it something the office manager can run without a developer, and stop losing buyers on the search page. We rebuilt the platform from the data model up rather than restyling what was there.

Listings come in over the RESO Web API on a fifteen-minute cycle, normalised into a single schema so a detached house and a downtown condo render through the same components. Photography is processed at upload into AVIF and WebP at five breakpoints, which is where most of the page weight on a listing site goes.

Every agent gets one profile template driven by a form: headshot, bio, specialties, languages, direct line. Adding a new agent is a five-minute job for the office manager instead of a ticket. Office pages assemble themselves from whoever is assigned to them, so a transfer between locations updates in both places automatically.

Everything above is arranged around one goal: the site stops being a dependency. Listings, agents, offices, and neighbourhood content are all editable in-house, so the work after launch is a periodic review rather than a support queue.

The performance work — image processing at upload, edge caching, a normalised schema — exists to keep listing pages fast as inventory grows, which is where brokerage sites usually degrade first.