Time tracking for cafes and restaurants — what actually works on a busy Saturday

A friend of mine runs a coffee shop in Lisbon. Twelve staff, three managers on rotation, a constantly-shifting student crew. Last summer he sent me a long voice note around 2 a.m. on a Sunday. The gist of it: their cafe’s WiFi had crapped out at 11 a.m. on Saturday, taking down the time-tracking app that runs on the iPad by the till. Nobody could clock in for four hours. Nobody knew whether to wait it out or scribble names on a napkin. By the end of service the shift log had gaps in seven different places, and reconstructing payroll for that day cost him most of his Sunday.

That’s the story I keep thinking about whenever someone asks me to recommend a time-tracking app for a small food-and-bev operation. The tool doesn’t have to be brilliant. It has to survive a bad day.

What restaurants actually need from a time clock

I’ll spare you the marketing checklist. Here’s what I’ve seen people actually use, in order of how often it comes up:

1. A way to clock people in fast — five seconds, hands free if possible, because hands are wet
2. Something that works when the internet doesn’t
3. A clean export to whatever payroll system the bookkeeper uses
4. Some way to handle split shifts and shifts that cross midnight
5. Editable records, with an audit trail, because mistakes happen

The last one is the one most apps undersell and most managers care about most. Somebody forgot to clock out at 2 a.m. Somebody clocked in for someone who hadn’t shown up yet. Somebody’s split shift looks like two clock-ins because their break went over the allowed window. You will spend more time fixing the records than recording them. So the app’s job is partly to make those fixes painless.

Why the WiFi thing matters more than people think

Restaurant WiFi is a category of its own. It’s never the WiFi you put in. It’s the cheap router behind the till, mounted at chest height, surrounded by metal, sometimes sharing a band with the customer guest network and 14 phones. It works most of the time and falls over precisely when you wish it didn’t.

Cloud-based time-tracking apps treat connectivity as a checkbox: “works offline.” But “works offline” usually means “the app opens, you can clock in, and we’ll sync when you’re back online.” Which sounds fine until the device itself dies — and then you discover the offline buffer was held in RAM, not disk, and three hours of clock-ins are gone.

A truly offline-first app stores every clock-in to disk before it acknowledges it. There is no “syncing” — there’s just the local database, and an export when you ask for one. If you’ve never thought about this distinction, that’s a sign you’ve never been bitten by it.

The setup we use, end to end

This is the setup for the Lisbon shop, more or less unchanged for nine months now:

• One Android tablet, mounted next to the door at chest height. Not next to the till. (Cluster too much hardware in one spot and one wet hand kills the night.)
• FaceClock installed on the tablet. Free, no account.
• Twelve registered employees. Each took about thirty seconds to register — the app asks for three reference photos at slightly different angles, then it’s done.
• Maximum shift length set to 14 hours. Auto-clock-out runs after that, with a notification.
• 90-day data retention (the default). Older shifts get pruned automatically, which keeps the database small and matches the legal advice he’d been given by a Lisbon labor lawyer.
• A weekly CSV export, dropped into a shared folder, opened by his bookkeeper Monday morning.

That’s it. No subscription, no admin portal, no per-user fee. The total monthly cost of the time-tracking system is whatever the electricity to keep the tablet alive costs.

What it actually costs to run

I asked him to add it up. For comparison, here’s what he had been paying with the previous system, a popular SaaS product:

• $4.50 per active user per month × 12 employees = $54/month
• One year: $648
• Plus the iPad they bought for it (one-off): ~$330

With FaceClock:

• Tablet (used Samsung Tab A8, bought refurbished): $120 one-off
• Power: about $1.50/month, generously
• Subscription: $0

The annual delta is roughly $640. That’s a couple of weeks of his most junior barista’s wages. The numbers are obviously different at scale — past about 25 staff, the SaaS model becomes more attractive on convenience grounds — but at his size it’s not close.

The handful of things to actually watch out for

I want to be honest about where this kind of setup gets awkward, because if you skip past these you’ll hit them in the wild.

The device is the system. This is the trade-off. With cloud-based tools, your records live on someone else’s server. If your iPad cracks, you’re inconvenienced. With on-device tracking, if the tablet falls off the wall on a Sunday, you’ve lost what you didn’t export. Mitigation: weekly CSV exports, stored somewhere else. Treat the device as ephemeral, not as the source of truth long-term.

Bad lighting can hurt recognition rates. A dim entrance corridor with strong backlight from a window is the worst case. We solved it by adding a $9 LED strip above the tablet — about ten minutes of work. After that, recognition went from “occasionally fails” to “I can’t remember the last time it failed.”

Twins and lookalikes. Two of his staff are sisters, not twins, but close enough that the app sometimes confused them in the first week. Solved by raising the confidence threshold and re-registering both with more reference photos. Hasn’t been an issue since.

Manager edits matter. Inevitably someone forgets to clock out and you need to fix it Monday morning. Make sure whoever you trust to do that has a clear, obvious way to do it without becoming a power user of the app. The fewer steps the better.

When this approach doesn’t fit

If you’re running multiple locations and want central oversight in real time, on-device tracking is the wrong shape — you’ll end up shuttling CSVs and wondering why you didn’t just pay for a multi-location SaaS. If you’re a chain of even five cafes, the math probably flips.

If you have a payroll workflow that requires real-time API integration into your scheduling system (think Deputy ↔ Square ↔ Toast type setups), you need something that lives in the cloud and exposes hooks. FaceClock won’t do that. It outputs CSVs. That’s the contract.

But for a single-location restaurant with under thirty staff, the calculation is usually: “do I need any of the SaaS conveniences enough to justify $50–$200 a month, every month, forever?” In a lot of kitchens, the honest answer is no.

The thing I keep coming back to

The fanciest time-tracking app in the world doesn’t help if it’s offline when you need it. And restaurants need it during the worst possible moments: peak service, when the chef is yelling and the bartender’s iPad is wet and the host is double-seating tables. That’s when you want a tool that’s already done its job five seconds before anyone thought about it.

A face camera by the door. A tablet. A CSV at the end of the week. That’s enough for most restaurants. It’s been enough for the Lisbon shop for almost a year now, and the only voice notes I get from the owner these days are about espresso machine repairs.

Which feels like a win.