Predicting next month's covers: a demand forecasting primer for independent operators

Start with the number you actually needLink

Forecast covers by site, day and service, not by month.

Stop saying, “It should be busy.” That is not a forecast. It is a guess dressed up as experience.

Guesses create bad rotas, excessive prep, wasted stock and poor service. They also make you spend marketing money on nights that were already going to sell out.

A proper demand forecast gives you a number. Not a perfect number. A useful number. You use it to plan covers, staff, purchasing, opening hours, promotions and pricing for the month ahead. The process is simple. The order matters.

Forecast covers by site, day and service. Do not forecast “September trade” as one lump. That number is useless.

A pub with 80 covers on a Friday night and 80 covers across Monday to Thursday does not have the same staffing problem. The total is identical. The operational reality is not.

For a hotel, forecast room nights separately from restaurant covers. A full house does not automatically mean a full dining room. Guests may eat elsewhere. Local demand may fill the restaurant even when occupancy is weak.

Your forecast must match the decision it is meant to support.

• Monday lunch
• Wednesday dinner
• Friday evening
• Sunday lunch
• Breakfast in a hotel or B&B
• Rooms sold by night
• Walk-ins
• Advance bookings
• Private functions
• Events and sports sessions

Pull 12 months of EPOS and booking dataLink

Not industry averages. Not a consultant's benchmark. Your actual covers.

Start with your own records. Pull at least 12 months of data. Twenty-four months is better where the business is highly seasonal.

Export the data from your EPOS, booking system and property-management system. If the exports are messy, fix that first. You cannot manage demand you cannot see. A forecast built on incomplete records is just a more sophisticated guess.

Create one row per day and service. Then add columns for day of the week, month, bank holiday, school holiday, weather category, local event, sport fixture, promotion and payday period, alongside actual covers.

Keep the spreadsheet plain. Fancy software will not rescue poor inputs.

• Covers by day
• Covers by service
• Revenue
• Average spend per head
• Bookings made
• Cancellations
• No-shows
• Walk-ins
• Discounts and promotions
• Private dining or function covers
• Room nights, if applicable

Build the baseline before adding adjustmentsLink

Compare like with like.

Your baseline is what you expect to sell under broadly normal conditions. The quickest method is to compare like with like: the same weekday, the same period last year, the nearest comparable trading days and your recent like-for-like trend.

For example, to forecast the first Saturday of next month, use the equivalent Saturday last year. Then compare the last three months with the same three months a year earlier.

If your comparable covers have increased by 6%, apply that trend to the baseline. If they have fallen by 4%, reduce it. Do not blindly apply market growth figures. Your own venue may be losing covers while the wider market grows, or it may be taking share from weaker competitors.

The calculation is straightforward: baseline covers = comparable historical covers × your like-for-like trend.

If last year's comparable Saturday produced 120 covers and your current trend is 5% up: 120 × 1.05 = 126 forecast covers. That is your starting point. Now you adjust it for reality.

Add weather without pretending you can control itLink

Calculate your own weather effect. Do not borrow someone else's.

Weather changes hospitality demand. The effect depends on your venue. A wet day may help a destination pub with a strong fireplace, parking and a good food reputation. It may damage a city-centre restaurant that relies on walk-ins. A heatwave may increase beer garden trade while reducing formal dining.

Tag historical trading days as dry and mild, wet, stormy, cold, very hot or normal. Then compare covers within the same service and weekday.

Suppose your autumn Saturday baseline is 100 covers. Your records show that wet Saturdays average 10% fewer covers. That is not “a 10% decline”. It is 10 covers missing. At £35 average spend, that is £350 in sales. Across four wet Saturdays, it is £1,400 in sales. The effect on contribution will be lower after variable costs, but the staff and fixed-cost burden remains.

Long-range forecasts are uncertain. Use broad scenarios: normal weather, wetter than normal, or warmer and drier than normal. Do not change the rota every time an app changes its forecast. Use the forecast as a planning range, then update the near-term view when the weather becomes more reliable.

as a reminder that weather affects the sector. Do not use it as a substitute for your own trading history.

• Use the RSM hospitality tracker

Layer in school holidays and bank holidaysLink

One blanket uplift is the wrong answer.

School holidays create demand shifts. They do not create the same shift everywhere. A family pub may see more daytime and early-evening covers during half-term. A commuter-area restaurant may lose weekday lunch trade because office workers are away. A hotel near a tourist attraction may see stronger occupancy but different booking lead times.

Compare last year's holiday periods with nearby term-time weeks and calculate the difference by service. Do not apply one blanket “school holiday uplift” to the entire venue.

Bank holidays need the same treatment. Compare the relevant bank holiday weekend with ordinary weekends across Friday evening, Saturday lunch, Saturday dinner, Sunday lunch, Monday lunch and Monday evening. Some sessions gain. Others simply move.

• Check the official school term and holiday dates on GOV.UK

Map local events before you set the rotaLink

Events can create demand. They can also steal it.

A concert may fill your restaurant before the doors open, then empty the town afterwards. A football fixture may help a wet-led pub and hurt a quiet dining room. A village festival may create traffic but not necessarily customers.

List the events within a practical travel radius, then check what happened last time. Separate pre-event demand, event-time demand, post-event demand and displacement from nearby venues. Apply only an adjustment your data supports.

If a previous local festival added 25 covers but reduced your normal Saturday lunch by 15, the real net gain was 10 covers. Record the net result. Do not celebrate the headline number while ignoring the lost trade elsewhere.

• Concerts
• Festivals
• Race days
• Sporting fixtures
• Weddings
• Business conferences
• University term dates
• Local fairs
• Theatre performances
• Road closures
• Major construction disruption

Replace opinion with a simple forecast modelLink

A transparent spreadsheet beats a black box.

You can manage this in Excel or Google Sheets, with a table for date, service, baseline, weather adjustment, holiday adjustment, event adjustment and forecast.

Example: baseline 126 covers, wet weather adjustment –13 covers, local event adjustment +8 covers, school holiday adjustment +5 covers. Forecast: 126 − 13 + 8 + 5 = 126 covers.

The forecast remains 126. The reasons have changed. That matters. It tells you where the risk sits.

You can later add regression analysis using day, month, weather, holiday, event and promotion variables. But start with a transparent model. Everyone should understand why the number changed. If your manager cannot explain the forecast to the head chef in two minutes, it is too complicated.

Turn the forecast into operating decisionsLink

A forecast that stays in a spreadsheet is worthless.

Staff to expected demand. Plan the rota around covers by service, not weekly habit. A quiet Tuesday should not carry the same labour structure as a busy Saturday. Use historical covers, average spend and expected prep requirements to identify the minimum safe staffing level. Do not cut so hard that service collapses. Saving £100 in wages while losing £500 in repeat custom is not efficiency. It is vandalism.

Purchase to expected covers. Convert forecast covers into purchasing quantities using your actual attach rates: main courses per cover, desserts per cover, chips or sides per cover, breakfast items per room night, and drinks per wet-led cover. Build a normal case and a downside case. Buy for the likely range, not the fantasy peak.

Price and promote around weak sessions. Do not discount Friday night because it is visible — Friday usually needs no help. Use the forecast to identify underperforming sessions such as Tuesday dinner, early Wednesday seating, Sunday evening, shoulder-season rooms and midweek breakfast, then build a targeted offer, set menu or event around that capacity.

This is where revenue management fits. Marketing creates demand. Revenue management decides where demand is needed and what it should be worth. For pubs, session-by-session planning matters even more: separate food, wet trade, functions, events and rooms instead of pushing one vague “visit us” message.

Protect the booking path. A forecast is only useful if customers can act on it. Check your Google Business Profile, booking buttons, menus and event pages. The rule is simple: if customers cannot find the session or book it quickly, your forecast will not save you.

• See our restaurant marketing services
• and our pub marketing agency approach

• For a pub-focused example of fast mobile pages and clear booking paths, see PubLandlord
• For broader website and local-search principles, see JetAds
• and for a plain-English audit of local visibility and conversion problems, see VU1

Review forecast accuracy every monthLink

A forecast is a control system, not a one-off document.

At the end of each month, compare forecast with actual. Track forecast covers, actual covers, variance, revenue variance, variance by weekday, variance by service, weather impact and event impact.

If you forecast 126 covers and served 102, investigate the gap. Was the weather worse? Did a road closure reduce access? Did a competitor open nearby? Did your booking page fail? Did you overestimate walk-ins? Did your offer attract low-value covers?

Do not simply change the number next month. Find the cause. A forecast is not a document you produce once. It is a control system. It gets better when you measure its errors.

Stop guessing and start planningLink

Pull the data. Build the baseline. Then decide.

Your next month's trade is not unknowable. You already have most of the evidence. It is sitting inside your EPOS, booking system, calendar and local knowledge.

Pull the data. Build the baseline. Add weather, holidays and events. Then turn the number into a rota, purchasing plan and revenue decision. That is the order.

Do it backwards and you waste money creating demand you cannot serve. Do it properly and you protect margin before the month starts.

• Start with our free marketing audit