One homepage, two audiences who want opposite things
A US shift-marketplace platform for residential buildings
The Brief
A platform connecting residential buildings with the staff who serve them — doormen, concierges, handymen, porters, and supers — needed its marketing site built in Next.js from the ground up.
The framework was the easy part. The hard part was that the site had to convert two groups who want opposite things, from the same pages. Buildings arrive looking for shift coverage. Individuals arrive looking for shifts to work. A hero that speaks to one of them is invisible to the other, and a hero that tries to speak to both usually reaches neither.
Then there was the vocabulary, which was a specification rather than a preference. The platform maintains a precise vernacular and needs clear distance from an adjacent category. Shifts, not jobs. Seekers, not workers. Language could not position the platform as a staffing agency, which carries different regulatory implications, and the platform could not describe a seeker as vetted.
That is not a style guide. It is a compliance boundary that happens to live in the copy.
The vocabulary, as specified
Applied site-wide rather than case by case.
| Not this | This | Why it matters |
|---|---|---|
| Jobs | Shifts | A shift is a unit of coverage, not an employment relationship |
| Workers | Seekers | Describes the person’s intent on the platform, not their status |
| Staffing agency framing | Marketplace framing | Agency positioning carries different regulatory implications |
| “Vetted” seekers | Only what can be substantiated | A claim the platform cannot stand behind is a liability, not a selling point |
The Approach
We connected GitHub and Vercel before page development started, so every change produced a live preview link. The client reviewed working software throughout instead of waiting for a delivery and reacting to it at the end, which is when structural feedback is most expensive to act on.
The vernacular was treated as a hard specification applied site-wide, not a per-page judgment call. Terminology rules enforced case by case drift, and drift here is not a tone problem — it is a positioning problem with regulatory weight behind it.
Review ran in structured rounds against a shared feedback document with items checked off as they were completed, so nothing was quietly lost between passes.
What we Did
Set up the project and the review pipeline first. Next.js scaffolded, header with logo and navigation built, GitHub and Vercel connected for live preview before any page development began.
Built the homepage around two separate paths. The core conversion sections were structured so buildings and seekers each get a distinct route through the page, rather than the two audiences competing for the same hero.
Applied the vernacular across the site. Shifts and seekers used throughout, staffing-agency framing avoided, and claims about seekers limited to what the platform can actually substantiate.
Changed the imagery to match the market. Photos updated to show the actual roles — doormen, concierges, handymen, porters, supers — and building-side imagery changed to residential buildings specifically. Generic commercial imagery would have signaled the wrong market to a building manager scanning the page for two seconds.
Built the three primary conversion actions. Post a Shift, Find a Shift, and Tell Us About Your Building. Two audiences, three clear next steps, no ambiguity about which one belongs to whom.
Scoped analytics across the subdomain boundary. Google Analytics and Tag Manager prepared with conversion tracking that follows through to the app subdomain rather than terminating at the marketing site edge — which is exactly where conversion data usually disappears, and why so many marketplaces cannot tell which audience is converting.
Removed the growth-metric banners. When the client confirmed the supporting data was not yet available to display accurately, the banners came out. A number on a homepage that cannot be backed up is worse than no number.
Worked feedback in batches and closed the loop. Copy corrections implemented against the shared document, favicon updated to the client’s logo, and a branding kit prepared covering logos, colors and color codes, fonts, and the imagery used to build the site.
One section, switched by audience. The toggle rewrites the three steps rather than showing both sets at once, so neither audience reads instructions written for the other. The vocabulary holds throughout: shifts and seekers.
Conversion action three. The building enquiry, for a manager not ready to post a shift yet but wanting a conversation
The seeker route stays separate.Self-onboarding in three steps rather than a form, because the two audiences need different amounts of hand-holding..
Results in Detail
Three distinct conversion actions built for two opposing audiences on a single site, giving each audience a clear next step rather than a shared one that fits neither.
This is the thing the brief was actually asking for. A two-sided marketplace lives or dies on both sides converting, and the usual failure is a homepage optimized for whichever audience the founder thinks about most. Splitting the paths means neither side has to work out whether the page is talking to them.
The vernacular applied consistently site-wide, with staffing-agency language avoided. A positioning requirement with regulatory implications, met as a specification rather than a preference.
Terminology discipline is invisible when it works. It becomes very visible the first time a regulator, a partner, or a competitor reads the site closely and finds the word that implies something the platform is not.
Conversion tracking scoped across the marketing site and the app subdomain, enabling accurate attribution by audience.
Attribution that stops at the subdomain boundary tells you people clicked. Attribution that carries through tells you which audience signed up, which is the only version of the data a two-sided marketplace can act on.
Delivered at approximately 95 percent complete and ready for domain cutover, with the client reviewing on live Vercel previews throughout rather than static mockups.
The remaining work is the cutover itself. Reviewing on live previews rather than flat mockups meant the client was signing off on the actual site behaving as it will behave, not on an image of it.
A branding kit was delivered covering logos, colors and codes, fonts, and imagery, so the client can extend the site independently. Growth-metric banners were corrected by removing those where supporting data was unavailable, keeping all displayed claims accurate. Favicon and copy corrections were also completed, worked through in structured review batches.
Tools & technologies
What the build runs on
verified