How to write a café roster that doesn't fall apart

Most café rosters are not hard to write. They are hard to write again — after someone drops a shift, after a casual goes on holiday, after the week you built on Sunday stops matching the week that actually happened. This is the order that survives that.

1. Collect availability before you open the calendar

The single biggest cause of a roster being rebuilt three times is starting to place shifts before you know who can work. You place Mina on Thursday, she tells you she has class, and now every shift after Thursday moves too.

Get availability in one place first, and get it as a standing pattern rather than a weekly conversation. Most café staff have a repeating shape — a student who cannot do Tuesday days, a parent who needs to leave by three on school days — with occasional one-off exceptions on top.

• Standing availability: the normal week for each person.
• Exceptions: the specific dates that break the pattern.

If you keep only the exceptions, you rebuild the pattern from memory every week. If you keep only the pattern, you roster someone onto the week they told you they would be in Sydney.

Why the group chat does not count

Availability sent by message is availability you have to re-read. It is scattered across a thread, it has no dates attached to half of it, and when there is a dispute later the record is a wall of messages. Anything is better than that: a shared sheet, a form, a proper tool. The test is whether you can see one person's whole availability without scrolling.

2. Build the week around your busy hours, not your people

Start from what the venue needs, then fill it. Owners who start from people ("Jay always does Monday") end up with a roster that matches last year's trade.

1. Write the hours you actually need covered, by day — open, the rush, the lull, close.
2. Mark which of those need a specific skill: someone on coffee, someone who can close, someone who can handle a function.
3. Now place people, most-constrained first — the closer, the barista, the one with the narrowest availability.

3. Check the wage cost before you publish, not after

A roster is a spending decision. The version most cafés use is a total that appears after the fortnight, in payroll, when nothing can be changed about it.

Look at the number while the week is still a draft, and look at it by day. A single day that is 30% above the others is usually one shift that starts an hour too early, and it is trivial to fix before you publish and impossible after.

4. Publish once, and make the published version the only version

The worst rostering failure is not a bad roster. It is two rosters — the one on the wall and the one in the group chat, which stopped matching on Wednesday.

Whatever you use, there has to be exactly one place where a staff member finds out when they work, and it has to be the place that updates. A printed roster with handwritten changes on it is two rosters. A published roster plus a chat message saying "actually can you come at 4" is two rosters.

Keep the draft private while you are moving things around, then publish deliberately. People should not watch you think.

5. Expect the week to change, and decide now how

Every café week changes. What separates a calm venue from a frantic one is whether the change has a route.

• Shift swaps: decide whether staff can arrange them directly or whether a manager approves. Either works. Ambiguity does not.
• Sick calls: know who the callable people are for each skill before 6am, not during.
• Extra cover: agree what triggers calling someone in, so it is not re-argued every Saturday.

The mistakes that cost the most

Timesheets nobody checks until payday

Three specific failures are worth guarding against, because each one reaches payroll looking like real data:

• The 0-hour shift. Someone clocks in, realises they are early, clocks out ten seconds later. That record is now in the timesheet and it looks legitimate.
• The forgotten clock-out. Someone goes home without clocking out. Next shift they cannot start, because the system thinks they are still on. It gets fixed by a manager typing in a number, which is exactly how hours end up wrong.
• The impossible shift. A finish before a start, a twenty-nine hour day, a break longer than the shift. These come from typos and they save silently unless something refuses them.

Whatever system you use, ask it what happens in those three cases. If the answer is "the manager notices", the answer is that it reaches payroll.

Rostering to last week instead of next week

Copying last week is the right starting point and the wrong finishing point. Copy it, then change the three things that are different about the coming week — a public holiday, a function, someone away, a delivery day moving.

Building it where staff cannot see it

If checking the roster requires asking someone, you will be asked. Every week, by everyone. A roster staff can open on their own phone removes more manager interruptions than any other single change.

Spreadsheet or software?

A spreadsheet is genuinely fine for a small, stable team where one person builds the roster and nothing changes much. It is free and you already know how to use it.

It stops being fine at the point where any of these are true:

• Staff need to see the roster themselves, on their phones.
• Availability changes often enough that you cannot hold it in your head.
• You want the hours worked to come from clocking in, not from what the roster said.
• More than one person edits it, and you have started emailing versions around.

The honest version of that trade-off, with what it costs, is on the cost page.

Try it on your own week

RosterPilot does the sequence above: standing availability with exceptions, a draft you build privately, estimated wage cost from the rates you configure, one published version staff open on their phones, and timesheets that refuse records that cannot be true.

Fourteen days, no credit card, your own staff and your own week.

Start the free trial →