PitStop, selling the right to walk in

An entrepreneur contacted me, as a freelancer, with a precise idea. Clean toilets already exist, in cafés. What's missing is the right to walk in without ordering a coffee or negotiating at the counter. PitStop sells that right in advance, a monthly pass at €1 or a 72-hour visitor pass at €1.50.

Three parties, only one of them pays. The subscriber gets a map, a pass and access instructions. The café owner opens their toilets and gets footfall, with no constraint and no cost. My job was to turn it into a product you can show and try, with a demonstration network of eight Paris venues and a simulated payment, labelled as such.

A café's sheet carries one thing above all, the access instruction. "Toilets downstairs, show the app at the counter." It's the most important field in the whole app, ahead of the hours, the walking distance and the photo of the entrance. In an emergency you don't compose a query, you tap once, so the list accepts one filter at a time.

The pass is a screen you hold up at the counter, full-screen teal like a boarding pass, with two proofs of life against screenshots, a clock ticking by the second and an animated badge. Everything works offline, except the rotating code, which is never persisted, because a stale code would only get you turned away.

Proving a visit, I ended up doing three ways. A QR sticker by the till, scanned with the native camera, which identifies the venue but not the person. A rotating code signed with HMAC-SHA256, which changes every 15 seconds, lives 20 seconds and works once. And a safety net, a four-digit code typed by the owner on the subscriber's phone, limited to ten attempts per minute.

No café-owner app, on purpose. No shopkeeper will install an application to say their toilets are being cleaned. They get a web page with no account or password, rendered in plain HTML, which opens on a 2015 phone sitting next to the till, with their footfall, a scanner in the browser and a pause of 24 hours at most.

The whole thing fits in 26 pages of documentation and six architecture decisions, so that someone else can pick the project up after me.