Rapariga Social Club — Case Study
backThe question
Starting over costs money at the worst possible moment. Someone leaving a house has to furnish another one in a week, on whatever is left. The idea was an invite-only club where women pass on the furniture, the clothes and the things a restart needs — a marketplace where the community, not the catalogue, is the product.
What I built
An MVP of eight pages, a design system at v1.2, and a finished brand: a chrome tulip, "grow together, glow together", a deep purple that clears AA, copy in European Portuguese. The plan was deliberately staged — design and build, then a beta with five testers, then fifteen to twenty members, and only then money, trademark and a permanent domain. Product first, spend last.
- Eight pages plus a standalone flow, built against a documented token set
- Brand, emblem and voice closed before a line of marketing
- Payments, moderation and reporting obligations mapped before building any of them
- A written go/no-go date, set months in advance
Technical decisions
The riskiest part was never the code — it was holding other people's money. Taking payment between members means identity checks on every seller, invoicing the fee, and a legal shape for the whole arrangement. The plan was to run the first transactions by hand, with me as the clearing house, before writing a single line against a payments API.
Points that never leave the club — a closed, non-transferable balance keeps the thing from becoming something I am not licensed to run. That constraint came from reading the rules first, not from a feature meeting.
Moderation is a liability decision — a wall where members post reduces what I am responsible for in the transaction and increases what I am responsible for in the content. Those are different problems and they wanted different answers.
What I learned
At the go/no-go I went back to the file that said validation was complete — twenty contacts, around eighty-five per cent positive. It wasn't true. The twenty were twenty posts I had read, not twenty people I had spoken to. The three things I had genuinely verified were the brand, the domain and the handle: availability, not desire. The figure was even labelled "estimate" in my own notes, and the file meant to hold the feedback was empty.
Written down plainly: I was treating as validated what I had never tested, and as tested what I had only checked. The honest move was to stop, and the cost of stopping was one summer of design instead of a year of build. Everything I made — the system, the brand, the regulatory map — still exists and still works. What was missing was five real conversations, and that is a gap no amount of design closes.
What I did about it
So I went and got the conversations. raparigas.club is live and recruiting twenty beta testers — not a launch, a validation page. It asks for one thing, an application, and it says so at the top of the screen.
Building it forced the product to get specific, which the design system never had. Trust is a plant two members grow together, not a five-star average: a rating can be farmed in an afternoon, a plant takes two people and time. Entry is by a member's invite or a questionnaire read by hand. Handovers happen at public meeting points, so nobody has to give a stranger their address. And listings marked Recomeço — starting over — get priority and fair prices, because the person furnishing a flat in a week on whatever is left is the person this was for in the first place.
Those four rules are answers to the question I had skipped: what, concretely, makes this safer than the marketplace that already exists? None of them came out of the design system. They came out of writing a page that had to convince someone to apply.