I started in kitchens at fifteen, making pasta in Turnpike Lane and selling it at farmers’ markets on the weekend.
I cooked until 2011. Pasta first, then chef de partie at Market on Parkway in Camden, when it first opened. Stock arrived on Monday against a week nobody had predicted, and by Thursday half of it was wrong.
Then head chef and unit manager at The Fine Food Company, running units at Lord’s and the RHS shows. My job description said to limit wastage and carefully monitor wage costs. It did not say how. So the rota was a guess. I wrote it on a Thursday from memory and the weather forecast, then spent the week fixing it: overtime I had not budgeted for, gaps I covered myself and never logged. At the time I thought I had a staffing problem. I had a forecasting problem.
Then two years as an account manager at Seafood Holdings, with 100+ trading accounts. Michelin-starred restaurants, five-star hotels, hospitals, contract caterers. That is where I started seeing the same forecasting problem everywhere. You can tell which accounts are guessing, because their orders are jagged. Big Tuesday, nothing Wednesday, then a panic order on Friday with a chef tearing their hair out down the phone. From the outside you can see it across a hundred businesses at once. From the inside nobody can see it about themselves.
Then five and a half years at Memotech, selling EPOS, ordering and back-office systems to Irish hospitality. By then the data was already sitting there in every one of those businesses. But most tills had no link to stock ordering at all. The chef controlled one, front of house ran the other, and the two never talked. Sales data on one side, orders on the other, and nobody joining them up.
Kitchen, unit, supplier, software. Four completely different-looking problems. It is one problem.
The bet nobody writes down
Every rota and every stock order is a bet on what you are going to sell. In most multi-site groups that bet is made by a GM, from memory, and it is not written down, not compared against what actually happened, and made differently at every site you own. Twelve sites, twelve forecasting methods, no way to see any of them from group level.
So it surfaces as unplanned overtime, waste, a quiet Tuesday fully staffed and a busy Saturday short. You see it in the monthly numbers, weeks after the money is gone, and you treat it as twelve separate operational failures instead of one systemic one.
The GM absorbs the rest with unpaid hours. That never appears in a report.
None of this was fatal when margins were fatter. Labour was cheaper and easier to find, food inflation was not what it is now, and the gap between the guess and the reality got quietly swallowed. That cushion is gone.
It also gets worse with scale, which is the part operators do not expect. More sites means more independent guesses and less group visibility, not more.
Run it on your own numbers
2% of revenue on a €10m group is €200k a year. That is not a cost-cutting exercise. That is the difference between what you guessed and what was going to happen anyway.
The groups I have seen fix it did not squeeze rotas harder. They made the forecast a number everyone worked from, instead of something living in one person’s head. Sometimes that is software. Sometimes it is a spreadsheet with a rule that everyone actually follows. The tool matters less than the discipline of writing the bet down and checking it against what happened.
If this is your P&L
If you run multiple food and drink sites and this is familiar, we should talk. Astraeus works with multi-site hospitality operators on exactly this kind of operational problem, and the first conversation is a free consultation of 45 to 60 minutes: where you are losing money on labour and stock, roughly what that is costing you a year, and whether fixing it needs software at all.
I have been on every side of this industry. You will get a straight answer.
Read how we work with hospitality groups or book the free consultation.
Frequently asked questions
Why do our rotas and stock orders keep going wrong?
Often it is a forecasting problem rather than a staffing or stock problem. Every rota and every stock order is a bet on what you are going to sell. In many businesses that bet is made from memory, not written down and not checked against what actually happened, so it shows up as unplanned overtime, waste and shifts staffed for the wrong day.
Why does forecasting get harder as a hospitality group adds sites?
More sites mean more independent guesses and less visibility from group level. In most multi-site groups, each GM makes the rota and stock bet from memory, differently at every site. Twelve sites can mean twelve forecasting methods, and the result appears as twelve separate operational failures in the monthly numbers instead of one systemic problem.
Why is forecasting a bigger problem for restaurants now than it used to be?
The cushion that used to absorb it has gone. When margins were fatter, labour was cheaper and easier to find and food inflation was lower, the gap between the guess and the reality was quietly swallowed. Now it shows up as unplanned overtime and waste, plus unpaid hours absorbed by the GM that never appear in a report.
Why doesn’t our EPOS data help with stock ordering?
In many hospitality businesses the till has no link to stock ordering at all. The chef controls ordering, front of house runs the till, and the two never talk. The sales data is already sitting in the business, but it sits on one side and the orders on the other, with nobody joining them up.
How can a multi-site group improve its labour and stock forecasting?
Make the forecast a number everyone works from, instead of something living in one person’s head, and check it against what actually happened. Squeezing rotas harder is not the fix. Sometimes the tool is software and sometimes it is a spreadsheet with a rule everyone actually follows. The tool matters less than the discipline of writing the bet down.