BLIK for the bill, and staff who set themselves up

We work in monthly cycles. Each has a written scope with a start and an end date, and closes with a signed acceptance protocol and a build published to the App Store, joined by a web deployment once the web work came to us. Since July 2025, 14 have run back to back without a month skipped.

That rhythm is most of the story. It is what let a client who hired us for one feature keep handing over the next thing, and the next, until the guest app, the staff app and the web products were all being built in the same place.

Loyalty was built in three passes, each on top of the last. Points first, earned on in-app payments, with a balance and a history a guest can check. Then membership tiers, with a discount attached to each. Then a rewards catalogue to spend the points on. GlobalTips says the programme raised how often guests come back.

Next to it, an Inbox, so a restaurant can send an offer to the guests who have already paid there: formatted messages with images and links, grouped by sender, push notifications, and a mute switch per sender for the guest who has had enough.

We took the client's Figma design system and rebuilt the Restaurant and Digital Menu screens on iOS and on the web, replaced SF Symbols with a platform-independent icon set, and left behind a reusable component library, so the next screen does not need a designer to draw it from scratch. iOS 26 requirements including Liquid Glass are supported while iOS 17 still works.

Around the same time we put in event tracking with batching, session tracking across background and foreground, and offline persistence that sends the deferred events once the network comes back. A/B testing sits on top of it: experiment configuration comes from the backend, variants apply to behaviour as well as to interface, and a user's assignment persists so their experience stays the same between sessions. After that, an argument about which version is better has somewhere to go.

Between features we replaced the legacy Combine networking layer with async/await, added retry with exponential backoff, and built a per-endpoint mocking layer so previews and tests stop depending on a live backend. Localisation moved to Xcode String Catalogs with a code generator that turns keys into compile-time-checked symbols, under a naming convention shared across iOS, Android and web, so a missing translation key now stops the build before it can reach a screen. Both changes were our own proposals.

On the web side: payment page work including 3DS and custom domains, the admin, business, sales and staff panels, German and Ukrainian localisation, and a Vite-based build.

Every cycle ends the same way: unit and integration testing on what was built, then a pass over the screens and flows that were already there, before the release goes out.