Rostering to Forecast, Not to Habit: Staffing Around Real Demand

Open your rota for next week. Chances are it started life as a copy of the week before, which was a copy of the week before that, with a name swapped out where somebody booked a holiday. Four on the floor Tuesday because there are always four on the floor Tuesday. Six on Saturday because Saturday is busy.

That system survives because it mostly works and it takes twenty minutes. The problem is what it costs you on the days it is wrong, which is more days than you think. A quiet Tuesday with four people on costs you real wages for covers that never arrive. A Thursday that turns out to be a school holiday with three people on costs you something harder to see on a spreadsheet: slow service, a table that waits eleven minutes for a first drink, and two reviews that mention it.

You already have most of what you need to do this better. It is sitting in your booking system and in last January.

Start with the two numbers you already own

Forecasting sounds like a spreadsheet exercise for people with a finance team. In practice, for a single venue, it comes down to two inputs and a bit of judgement.

The first is your booking book for the day in question. On a Wednesday morning, look at Saturday. You have a number of confirmed covers already in the diary. That number is not your Saturday, but it is the floor, and it is knowable.

The second is what happened on the same day last year, adjusted for how your business has moved since. If last year's second Saturday in February did 140 covers and you are running roughly ten percent up on last year, your starting estimate for this year is somewhere near 155.

The useful part is combining them. Track, over a few weeks, what proportion of your final covers were already booked at the same point in the week. Most venues find this is remarkably stable. If you consistently have around sixty percent of Saturday's covers on the book by Wednesday morning, then forty booked covers on Wednesday points to a Saturday of roughly sixty-five. That ratio is your own private forecasting tool and it costs nothing to build. Write the booked number down every Wednesday, write the final number down every Sunday, and after a month you have it.

Then layer on the things a spreadsheet cannot know. School holidays. The match at the stadium two streets away. The conference that fills every hotel in town the second week of March. The roadworks that closed your street to parking for a fortnight. A bank holiday Monday behaves like a Sunday, so treat it as one. Bad weather on a night where half your covers are walk-ins hits you hard; on a night that is ninety percent booked it barely registers.

The cost of overstaffing is obvious, so people overweight it

One extra person on a shift costs you their hourly rate plus employer contributions, so call it somewhere between forty and eighty euros for an evening depending on where you are and who they are. It appears on your wage bill as a clean, visible, uncomfortable number. Everybody in the business can see it.

Because it is so visible, it dominates the decision. Managers get told to hold wage percentage under a target and the easiest lever is to cut one person from a shift. The saving lands this week and it is easy to point at.

The cost of a bad service is invisible, so people underweight it

Understaffing costs more, but nothing on your P&L is labelled "the Thursday we got caught out."

Work through what actually happens. Service slows down. Starters take longer, so tables sit longer, so your second sitting slides and the 20:30 booking waits at the bar. Guests who wanted another bottle do not get asked, because nobody has time to ask. Your average spend per head drops on the exact night you were trying to save money. Somebody who would have come back does not book again, and you never find out.

Then there is the part that costs you next year. Understaffed shifts are miserable to work. Your best server, the one who knows the regulars by name, does not quit over one bad Thursday. They quit after the fifth one, and replacing them costs you weeks of recruiting plus the productivity gap while somebody new learns the room. Hospitality turnover is punishing everywhere, and understaffing is one of the few causes entirely within your control.

Set both numbers side by side and the asymmetry is clear. Overstaffing costs you a known amount, once. Understaffing costs you an unknown amount, repeatedly, in reduced spend, lost repeat visits, public reviews, and staff who leave. When your forecast is genuinely uncertain, err upward. The error is cheaper.

One more thing worth pricing before you touch the rota: no-shows sit directly on top of your forecast. If a chunk of your booked covers routinely goes to people who never turn up, you are rostering for a room that does not arrive. Work out your own rate from your own book before you trust anyone else's number. Our no-show calculator puts a number on what that is costing you across a year, which is usually the argument that gets a deposit or reminder policy approved.

Where split shifts help and where they quietly hurt

If your demand has two clean peaks (a lunch rush and a dinner rush with a dead middle), a straight eight-hour shift means paying people to fold napkins for three hours. The split shift is the obvious answer and it is genuinely the right answer for some venues.

It is also one of the most reliable ways to lose staff, so be honest about when it works.

Splits work when the gap is short enough to be a real break, roughly two to three hours, and when the person lives close enough to go home. They work when they are voluntary and when there is something in it for the person working one, whether that is a premium on the hours or a four-day week instead of five. Some people love them. A parent doing a school run in the middle of the gap may prefer a split to anything else you can offer.

Splits fail when the gap is four hours and the commute is forty minutes each way. That is not a break, it is unpaid time you have taken from someone. Do that twice a week to somebody who commutes and you are training them to look at other job ads. They fail hardest when they are assigned to whoever is least likely to object, which is usually your newest and cheapest staff, which is exactly the group with the highest turnover already.

Two alternatives are worth trying before you reach for splits. Stagger start times so people arrive across a ninety-minute window matched to when covers actually land, rather than everyone at 17:00. And use short shifts deliberately, a five-hour dinner-only shift for people who want exactly that, instead of stretching one person across the whole day.

Publish the rota further ahead than feels comfortable

Ask people who left hospitality why they left and scheduling comes up constantly, usually described as the inability to plan a life around the job. If you find out on Friday what you are working on Monday, you cannot commit to a course, a football team, a friend's birthday, or a second job.

Three weeks ahead is a genuinely good standard. Two weeks is workable. Publishing on Friday for the week starting Monday is the thing that quietly costs you people, and it costs you the ones with options first, because they are the ones who can get a job elsewhere with a fortnight's notice.

The usual objection is that you cannot forecast three weeks out, and that is fair. But you are not committing to a perfect roster. You are committing to a shape: who is working which days, and roughly when. Publish that early, then adjust the fine detail as bookings firm up in the final week. Most people can plan around "Thursday evening, starting somewhere between five and six." Almost nobody can plan around silence.

A few things make early publishing survivable. Collect availability once a month rather than chasing it weekly. Keep a small pool of people who genuinely want extra shifts and will pick up on short notice, and treat them well. And accept that the rota will change. Nobody expects it to be set in stone. They expect to know the shape of their month.

What the loop looks like once it runs

Wednesday morning, pull the booked covers for the coming weekend. Compare against the same weekend last year, adjusted for your growth. Apply your booked-to-final ratio. Check the calendar for events, holidays, and anything local that moves people around.

You now have a number. Roster to that number, rounding up rather than down when you are unsure. Note what you forecast.

Sunday night, write down what actually happened, and if you were off by a lot, write down why. Do that for a season and your forecast gets sharp. More usefully, so does everyone else's: a head chef who has seen the numbers for three months starts prepping to the forecast instead of to memory, which pulls your food cost down alongside your wage cost.

None of this requires a workforce management platform. It requires that the rota gets built from what you know about next week rather than from what you did last week.