Persona voice“I just want next week done before Sunday dinner.”
- Build and publish the rota
- Keep labour under 30% of sales
- Run payroll without re-typing hours
Crewday helps cafés, bars and restaurants with 8 to 40 staff run their week. In the brief, managers juggle a spreadsheet rota, a payroll provider and a group chat. Staff have no single place to see shifts or pay.
The goal: one product where the rota, clock-ins, tips and payroll feed each other, and a staff app people actually want to open.
Business problem (assumed): small venues trial scheduling tools, then churn when payroll still lives somewhere else and hours get re-typed. Design goal: a published rota becomes clock-ins, and clock-ins become payroll, with no re-typing in between.
These figures frame the brief. A real engagement would replace them with interview and rota data before any design decision leaned on them.
A self-initiated concept, framed as a client brief so the decisions have real stakes. I worked solo, so everything below is mine, including the parts a real team would share.
Team: solo concept. In a real engagement I would expect a PM, two engineers and a payroll specialist.
About 7 weeks, split into five phases. This is how I would plan the work as a real engagement, not a record of time spent.
Persona voice“I just want next week done before Sunday dinner.”
Persona voice“Tell me when I work and what I earned. That’s it.”
Personas are assumptions built from the brief, not interview subjects.
A floating dock and a full-width weekly shift board, like a wall planner on screen. Try it below.
Sofia, a barista, opens the app twice a day: to clock in and to see what she earned. Everything else is one tap away.
Frontline work happens in basements, on old phones and at 23:00. The happy path is the easy part, so these are the states I designed for first.
The clock-in is stored on the device with the device time, then synced. Sam sees a small “synced late” tag on the board, not an error, so nobody is marked late for a basement with no signal.
A soft check, not a block. Hard geofences fail on bad GPS and push people back to paper. An honest “clock in anyway” keeps the record complete and puts the decision with the manager.
Availability comes from the same rota the manager edits, so the app never offers someone who is already on shift. Picking a busy person explains why and points to who is free, instead of failing after Sam approves.
Payroll cannot be confirmed while a shift has a missing or edited time. The manager fixes it in the row, and every edit is logged with who changed it and why, which is what an accountant asks for later.
The swap conflict is interactive in the staff app above. The other three are designed states, not yet wired into the prototypes.
Owners are not software buyers. The site leads with warmth and shows the product working: stickers of everyday moments, an auto-filling rota, a tips splitter and a savings calculator. Proof is framed as pilot goals, not invented testimonials.
Managers live in one screen, the week. A small dock on the left gives the shift board the whole width, and it reads like a wall planner.
Rejected: A standard left sidebar, as in Gusto and Deel. It costs over 200 px of the width a seven-day board needs, for navigation managers use a few times a week.
Dragging is fiddly on a laptop trackpad during service. Picking a shift and tapping a free day is faster and works the same on a tablet.
Rejected: Drag and drop as the only way to move shifts. It is fiddly on a trackpad and unusable with a keyboard or switch access. Drag can come later as a shortcut on top.
Every change updates the cost bar and the percent of sales, so managers see the effect before they publish the rota.
Rejected: A separate labour report. Managers would see the cost after publishing, when it is too late to change. Known gap: the figure is wages only; an “include employer costs” toggle is next.
A slide gesture avoids accidental clock-ins from a pocket and feels like a small ritual at the start of a shift.
Rejected: A plain tap button. Faster, but pocket taps create false clock-ins that a manager then fixes in payroll. The slide is a native range control, so keyboards and screen readers can operate it (see Accessibility).
Review hours, check pay, confirm. Each step answers one question, then ends in a celebration because payday should feel good.
Rejected: One long payroll table with a Pay button. Fast for experts, but it hides the two questions that cause mistakes: are the hours right, and is the pay right.
The brief’s sharpest problem is late changes. Publishing a changed rota first shows who is affected. Each person gets one notification per change and taps Got it, so the manager knows it was seen. Designed, not yet in the prototype.
Rejected: Saving edits straight to everyone’s phone. It feels fast, but silent changes are how a moved shift becomes a missed one.
Frontline staff are every age, use whatever phone they own and often read in a second language. These decisions are in the prototypes unless marked as spec.
It is a native range control with a label, so a keyboard (End key) or a screen reader’s adjust gesture can complete it. The spec adds a tap-and-confirm button when VoiceOver or TalkBack is on.
Role colours always come with the role name and the shift times on the chip, so the board works in greyscale and for colour-blind managers.
Every control in the staff app is at least 48 px tall, and the main actions sit in the bottom half of the screen.
Confetti, bouncing stickers and sheet animations stop when the system asks for reduced motion. Nothing important is only shown through animation.
Payslips and the clock-out sheet show the sum as rows (hours, break, tips, taxes) so people can verify pay without doing maths in their head.
White text on the Peach buttons measures about 2.4:1, below WCAG AA. The fix is Ink text on Peach (about 7:1), which keeps the brand colour.
Soft and friendly: warm cream instead of white, round shapes everywhere, one colour per role so a shift is recognisable from across the kitchen. Every colour has one job.
Scope is a design decision. Each cut keeps v1 focused on one loop: rota, clock-in, payroll.
The brief is single venues. Roll-ups and head-office permissions would slow the core loop for the people it is for.
v1 copies last week and lets the manager edit. Auto-filling needs real availability and sales data first, and trust comes from seeing the week.
v1 flags overtime only. Rest periods, maximum hours and under-18 rules vary by country, so they come next with legal review.
Crewday prepares the monthly file for the accountant. Filing itself needs per-country compliance work.
The board shows wages as a share of sales. Employer contributions add a lot in EUR markets, so a toggle is first on the v1.1 list.
Swaps, notifications and time off cover what clogs group chats today. A full chat is a product of its own.
Concept project, not measured. No pilot was run. These are the targets I would agree with a client for a 6-week pilot in three venues.
Counter-metric: manual edits to clock-in times per payroll run. If edits go up while missed clock-ins stay at zero, the clock-in flow is failing quietly and managers are patching it.
The slide to clock in. It protects against pocket taps, but it is the one gesture every new starter must learn on day one. I would test it against a press-and-hold button with five baristas before committing.
Making the numbers agree across three surfaces. Hours, breaks, tips and take home have to match from the rota to the payslip. An early version of the staff payslips did not add up, which is the first thing a payroll buyer checks.
The riskiest assumption: staff will open a work app weekly without being told to. Cheapest test: a clickable staff app with one venue’s team for two weeks. Success means 6 of 8 open it unprompted each week.
I can walk you through the prototypes and the decisions behind them on a short call.