Prep from forecast, not from yesterday's prep

Most kitchens set tomorrow's prep by looking at what they prepped today and adjusting it a little. It feels sensible, it is quick, and it quietly repeats every mistake the kitchen made yesterday. If today was over-prepped, tomorrow starts from too much. If today ran out of the fish at nine o'clock, tomorrow's adjustment is a guess about how much more would have sold.

To prep from forecast is to start somewhere else: with how many guests are expected, what they are likely to order, and what is already in the fridge. The arithmetic is simple, it takes a quarter of an hour once the habit is in place, and it gives the kitchen a number it can defend when the night is busier or quieter than expected.

This guide sets out the steps, from covers to dishes to prep quantities, with a worked example. It also covers the two kinds of dish where forecasting always loses, and what to do about them instead. Every figure in it is an example; use your own history.

Yesterday's prep is a record of yesterday's decision, not of yesterday's demand. It tells you what the kitchen chose to make, which may have been too much or too little, and it says nothing about why. A kitchen that copies it forward carries the error with it.

There are three more problems. Days are not alike: a Tuesday lunch and a Saturday dinner are different businesses, and so are the same Saturday in a holiday month and in the quiet weeks after it. Leftovers distort the picture: if yesterday ended with six portions of braise in the fridge, today's prep should be six smaller, but the copied list rarely knows that. And copying hides the one thing worth learning, which is how far the kitchen's guesses miss.

A forecast does not need to be clever to beat this. It needs to start from demand rather than habit, and to be checked against what actually sold.

A prep forecast answers three questions for each service: how many guests, what will they order, and when. The first comes from bookings and history. The second comes from the menu mix, the share of guests who order each dish. The third matters for items with a short holding time, which need to be ready in batches through the night rather than all at once.

You will also need a few figures the kitchen probably already has: the recipe quantity of each component per portion, the batch sizes that make sense for each item, how long each prepared item keeps, and what is already prepared and in date.

Our guide to sales forecasting for next week in twenty minutes covers the wider forecast that this one sits inside. Here the focus is on the daily step that turns it into a prep list.

Begin with the covers you already know about: the bookings on the books for tomorrow, by service. Then estimate the guests you do not know about yet, the walk-ins and the late bookings, from your own history.

A simple way to estimate them is a ratio. Look at the last few weeks of the same service on the same day of the week, and compare the bookings that were on the books the evening before with the covers you actually served. If, as an example, Thursday dinners have ended up at 1.25 times the bookings held the night before, a Thursday with 80 covers booked forecasts at 100.

Keep the ratio by service and by day of the week, because a weekday lunch fills from walk-ins in a way a Saturday dinner does not. Our guide to a covers forecast your kitchen will believe goes further into building and testing these ratios. The method is a small version of what larger businesses call demand forecasting: using what happened before, adjusted for what you know is different, to estimate what will happen next.

The booking list holds more than a number of covers, and a few details are worth reading before you apply the ratio.

None of these needs precision. A quick look at the list while setting the prep is enough to catch the service that is about to be unusual.

Covers alone do not tell the kitchen what to make. The menu mix does. For each dish, work out what share of guests ordered it over recent comparable services. If, over the last four Thursday dinners, 18 per cent of guests ordered the braised short rib, then a forecast of 100 covers suggests about 18 portions.

Use recent weeks rather than the whole year, because the mix shifts with seasons, menu changes and specials. Use the same service and day where you have enough data, because a lunch mix is not a dinner mix. Our guide to menu mix explains how to read it and why it differs from what you think you sell.

Two refinements help. Count dishes per guest rather than per table, so that a table of eight is not treated like a table of two. And watch for dishes that are usually shared, such as sides and sharing boards, which follow tables more than guests.

The history describes an ordinary day. Tomorrow may not be one. Before turning dishes into prep, adjust for what you already know about it.

Adjust in steps and write down why. "Covers up 10 per cent for the match" is a note the kitchen can learn from next time. A silent change to the numbers is not.

Now turn each dish forecast into the components the kitchen has to prepare. For each component, multiply the portions forecast by the quantity per portion from the recipe, and add the same component's use across every dish that shares it.

An example. The short rib forecast is 18 portions, and each uses 220 grams of braised rib, so the kitchen needs about 4 kilograms of cooked rib. The same braising liquor is used in a pie on the lunch menu, which forecasts 10 portions at 80 grams of liquor each, so the liquor needed is the rib's share plus 800 grams.

Then round to sensible batches. A braise is made in the batch size the pot and the oven allow; a sauce in the quantity that keeps its quality. Rounding up to a batch is fine as long as the extra is accounted for in the next day's prep.

Before writing the final number, look in the fridge. Anything already prepared, in date and good for tomorrow's service comes off the total. This is the step copied prep lists miss most often, and it is where much over-preparation starts.

Then add a buffer: a little extra, so that a busier night than forecast does not run out at the first rush. The size of the buffer should depend on the dish, not be a flat 10 per cent across the board. Two questions decide it: what does it cost to run out, and what does it cost to throw away?

The worked example in full: 18 short rib portions forecast, a buffer of 3 because it is a signature dish that keeps for two days, and 6 portions already cooked from today. The prep is 18 plus 3 minus 6, which is 15 portions, or about 3.3 kilograms of rib to braise.

A daily forecast tells the kitchen how much to make. For items with a short holding time, it also matters when they are made. A sauce that holds all night can be made in the morning. Fried garnishes, cut herbs, dressed salads and some breads lose quality within an hour or two and are better made in waves.

Split the day's forecast for those items by the shape of the service. If, as an example, a dinner service takes 30 per cent of its covers between six and seven-thirty, 50 per cent between seven-thirty and nine, and the rest after that, prepare short-life items in matching batches, with the second and third batches started as the first runs low. The total is the same; the quality and the waste are not.

The same split helps with staffing the prep. A morning team that knows exactly what must be ready by the first sitting, and what can wait for a quieter hour in the afternoon, works more calmly than one told to prepare everything at once.

Some dishes cannot be forecast well from history, however careful the method. Two kinds come up in every kitchen, and it is better to recognise them than to keep trying.

A dish with no history has no mix to forecast from. The first few nights of a new menu item, a weekly special or a seasonal dish are guesses, and treating them as forecasts gives them a precision they do not have. Prep them to a fixed, deliberate quantity that you are prepared to sell out of, tell the floor team how many there are, and record what sold. After a few services the dish has a history and can join the forecast.

A dish that sells three portions on one night and eleven the next has an average, but the average describes no real night. Its sales depend on a few tables, and one large party can double them. Forecasting it from the mix will be wrong most nights in one direction or the other.

For these dishes, use a par rather than a forecast: decide how many portions to hold at the start of each service, prep back up to that number, and accept that the dish will occasionally sell out. If it is long to prepare, consider whether it can be finished to order from a component that keeps. Our guide to par levels explains how to set them by value and by how much each item varies.

A forecast that never runs out is a forecast that over-prepares. For some dishes, running out near the end of service is the right result: it means the kitchen prepared close to demand and threw nothing away.

Decide in advance which dishes may run out and which may not. A signature dish, a children's dish, or the vegetarian main on a small menu probably should not. A special, a premium cut or a dish with good alternatives probably may. When a dish does run out, take it off sale at once, so that the floor team stops offering it and guests are not disappointed at the table. Record the time it ran out, because that is the most useful single fact for the next forecast.

A forecast that is never checked never improves. Each morning, write the forecast for the previous service next to what actually sold, dish by dish, and note the error.

A simple measure is enough: the difference between forecast and actual, as a share of the forecast. If short rib was forecast at 18 and 15 sold, the error was 3, or 17 per cent over. Over a few weeks, the pattern shows which dishes are reliably over-forecast, which under, and which are simply volatile. Our guide to scheduling from forecast uses the same covers forecast for the rota, so checking it pays twice.

Do not chase every error. A forecast within 10 to 15 per cent on most dishes, in this guide's example, is good enough to prep from. What matters is that the errors are not all in the same direction.

The forecast is most useful if it is done at the same time every day, by the same role, in the same format. A common pattern is for the chef in charge of the evening to set the next day's prep list during the last quiet half hour of service, when the bookings for tomorrow are mostly in and the fridge can be checked before close.

The routine itself is short:

Our guide to the prep list that ends the morning guesswork covers how the list itself is laid out so the morning team can work from it without questions.

In a group with a central kitchen, the same steps run one level up. Each venue forecasts its covers and dishes, the central kitchen adds them together into a kitchen production plan, and the venues finish on site. The discipline is the same, and so is the checking, but the cost of a bad forecast is larger, because it moves through a production run and a delivery before anybody notices. Our guide to running a central kitchen for three venues covers that production plan in detail.

A prep forecast starts with two kinds of history: covers and dish sales. servearound holds both, and the prep list itself stays on your own sheet or board.

In reservations, tomorrow's bookings are on the day's view with their party sizes, so the covers already on the books are there to start from. Analytics shows covers day by day, with booked guests and walk-ins side by side, which is the history behind a walk-in ratio. The menu report shows units sold for every dish over any range of dates, which is the menu mix the forecast needs. When a dish runs out during service, 86 it on the till and it shows as sold out on every table's menu at once.

servearound does not write forecasts or prep lists today, so the arithmetic and the list stay with the kitchen. If you would like to see where the numbers come from, get in touch.

A day ahead for most items, set during the previous evening's service. Items that take longer to make or keep longer, such as stocks and braises, can be planned two or three days ahead and checked daily.

There is no general figure. It depends on your location, your booking policy and the service. Work out your own from the last few weeks of the same service and day, and update it as the season changes.

Not a flat one. Set it dish by dish, from the cost of running out against the cost of waste. Signature dishes that keep well deserve more; perishable, expensive dishes with alternatives deserve less.

You cannot, at first. Prep a deliberate, fixed quantity you are willing to sell out of, tell the floor team how many there are, and record sales. After a few services it has enough history to forecast.

For venues with outdoor seating, passing trade or a seasonal menu, yes. Keep a note of the adjustments you make and how they turned out, so the next adjustment is based on your own experience rather than a hunch.

Accurate enough that the kitchen rarely runs out of its core dishes and rarely throws away prepared food. In this guide's example, within 10 to 15 per cent on most dishes is good enough to work from.

Take it off sale immediately so guests are not offered it, record the time, and use that time when you forecast the next comparable service.

Copying yesterday's prep repeats yesterday's mistakes. Instead, prep from forecast: start with tomorrow's bookings and a walk-in ratio from your own history, turn covers into dishes with the recent menu mix, adjust for weather, events and last year, then turn dishes into components, subtract what is already in the fridge and add a buffer set dish by dish.

Treat new dishes and low-volume, lumpy dishes differently: prep them to a fixed quantity or to a par, and accept that they will sometimes sell out. Record every forecast and check it against sales the next morning. A forecast that is checked gets better; one that is not is just yesterday's prep with more steps.