Key takeaways
- Your booking diary already holds a forecast: the share of final covers that are usually booked by a given day before service.
- Forecast = covers on the books today divided by the share normally booked at this lead time, plus expected walk-ins.
- Demand moves sharply with the season: 30.7% of accommodation and food businesses said demand fell from August to September 2026[1].
- Check forecast accuracy every week; a forecast you never score will not improve.
Why the diary beats gut feel
Most managers forecast next Saturday by remembering last Saturday. That works until it does not: a school holiday, a heatwave or a match changes everything, and memory gives you no way to tell how unusual this week is. Your booking system, whether it is a dedicated reservations platform or a paper diary, records when each booking was made as well as when it is for. That timing is the key to a better forecast.
Bookings build up in a pattern. A Saturday dinner might be half full a week out and nearly full by Thursday; a Tuesday lunch might get most of its trade on the day. Once you know your own pattern for each day and session, the number of covers already on the books tells you roughly how many you will end up with.
What a pick-up curve is
A pick-up curve shows, for one day and session, what share of the final covers had been booked at each point before service. You build it from history and use it to project the future. The table shows an Example for a neighbourhood restaurant.
| Days before service | Saturday dinner: share of final booked covers | Tuesday dinner: share of final booked covers |
|---|---|---|
| 28 | 35% | 10% |
| 14 | 55% | 20% |
| 7 | 75% | 40% |
| 3 | 90% | 65% |
| 1 | 97% | 85% |
| Day of service | 100% | 100% |
The two curves tell different stories. In this Example Saturday is booked early, so a slow Saturday is visible weeks ahead. Tuesday is booked late, so a quiet-looking Tuesday a week out is normal and not a reason to panic or cut staff yet.
Building your own curve
Export at least eight weeks of history
You need each booking's created date, service date, session and party size. Most reservation systems export this; with a paper diary, start recording it now.
Group by day of week and session
Saturday dinner and Tuesday lunch behave differently, so build a separate curve for each session you trade.
Count covers on the books at fixed points
For each past service, count booked covers 28, 14, 7, 3 and 1 days before, and the final booked total. Count cancellations as they happen, so a booking cancelled two days out drops out of the later counts.
Convert to shares
Divide each count by the final booked total and average across the weeks. Exclude weeks with something exceptional, such as Christmas or a private hire.
Record walk-ins separately
Walk-ins are not in the curve. Note the average number of walk-in covers for each session from your EPOS or door counts.
Refresh it each quarter
Booking habits change with the season and with your marketing. Rebuild the curve every three months, and separately for December.
Turning the curve into a forecast
The formula is simple: forecast booked covers = covers on the books now divided by the usual share at this lead time. Then add expected walk-ins and subtract expected no-shows.
Example: it is Monday and Saturday dinner, five days away, has 48 covers booked. Halfway between the seven-day point (75%) and the three-day point (90%), the curve gives about 82%. Forecast booked covers = 48 / 0.82 = 59. Add 6 walk-ins, subtract 2 no-shows: forecast 63 covers.
Interpolating between the fixed points, as in the Example, is fine; you are not trying to be exact. Round to the nearest five covers when you hand the number to the kitchen, and give a range for sessions with lots of walk-ins.
Adjust for what the curve cannot see
Pick-up curves assume this week is like an average week. Several things break that assumption, and they are worth a manual adjustment.
- Season. Demand turns quickly after the summer. 30.7% of trading accommodation and food businesses said domestic demand fell between August and September 2026, while only 4.4% saw it rise[1].
- Weather. 10.6% of accommodation and food businesses said weather was affecting their turnover in October 2026[1]. Hot spells pull people outdoors: restaurants' at-home sales slowed in summer 2026 as consumers headed out during heatwaves[2]. A garden pub and an indoor restaurant need opposite adjustments.
- Events. Big tournaments, concerts and local festivals shift demand. July 2026 was managed pubs' best month of the year, boosted by the men's football World Cup[3].
- School holidays and bank holidays: build a separate curve for holiday weeks if you have the history.
- Your own marketing: a newsletter or a social post brings bookings forward, which makes the curve look ahead of normal when final covers may not be.
From covers to rota and prep
A cover forecast is only worth making if it changes what you do. Agree a standard conversion with the head chef and the floor manager so the forecast flows straight into the plan.
- Front of house: covers per server per session for your style of service, and the point at which you add a runner or a second bartender.
- Kitchen: covers per chef on each section, and the forecast at which you open a second fryer or grill station.
- Prep: par levels for the main mise en place items per 10 covers, adjusted on the morning of service.
- Ordering: perishable orders placed against the forecast for the next delivery window, not against last week.
Seasonal swings in staffing are large across the sector. Not seasonally adjusted, UK accommodation and food workforce jobs rose from 2,566,000 in March 2025 to 2,657,000 in June 2025, a spring and summer rise of about 91,000[4]. A good forecast lets you plan those extra hours in advance rather than calling people in at short notice.
Example: forecasting a week in October
Example (fictional). On Monday morning the manager of an 80-cover restaurant opens the diary for the week. Tuesday has 26 covers booked, Wednesday 22, Thursday 31, Friday 52 and Saturday 64. Using the session curves, the forecast for booked covers comes out at roughly 30, 40, 48, 64 and 78, before walk-ins. A local half-term starts on Friday, which last year added about ten covers to each weekend session, so she adds that by hand. The forecast for the week is about 300 covers including walk-ins, against 330 in the same week last year.
That gap changes three decisions. She drops one front-of-house shift on Tuesday and keeps Wednesday as it is, because Wednesday's curve is late-booking. She asks the kitchen to cut Tuesday's prep par by a fifth. And she sends a short newsletter on Tuesday aimed at Thursday, the weakest night relative to last year. On Friday morning she checks the forecast again with the latest bookings and adjusts the weekend rota once more.
Scoring the forecast
After each service, write the forecast and the actual side by side. Once a week, work out the average error as a percentage of actual covers for each session. A forecast within a tenth of actual covers for most sessions is good enough to plan staff and prep. If one session is always over or under, the curve for that session needs rebuilding or walk-ins are being misjudged.
- Error under a tenth of actual covers: keep the method.
- Error between a tenth and a fifth: check walk-in and no-show assumptions first.
- Error above a fifth for three weeks: rebuild the curve from recent weeks only.
- Big misses on specific days: check for events or weather you did not note.
Common traps
- Large parties distort a curve. Treat any booking over ten covers as a known event and add it on top rather than letting it inflate the share.
- Cancellations and no-shows differ by session; Friday and Saturday often have more late changes than midweek.
- Bookings taken by phone but entered late make the curve look slower than it is. Enter them at the time of the call.
- Opening a new session or channel makes old curves useless for that slot until you have six to eight weeks of new history.
Frequently asked questions
How many weeks of history do I need?
Eight weeks gives a usable curve for each session; 12 to 16 weeks is better. Use the same season from last year if you have it, because summer and winter booking patterns differ.
What if most of my trade is walk-in?
Use sales history instead: the average covers for the same session over the last four to six comparable weeks, adjusted for weather and events. Track walk-ins by hour so you can staff the busy window rather than the whole session.
Can I do this without special software?
Yes. A spreadsheet with one row per service and columns for covers booked at each lead time is enough. Fill in the counts each morning from your booking system and the curve builds itself over a couple of months.
References
Office for National Statistics (2026). Business insights and impact on the UK economy, wave 166 dataset (opens in a new tab) ↑
NIQ (CGA by NIQ) (2026). Restaurants' delivery sales strengthen in September after tough summer (Hospitality at Home Tracker) (opens in a new tab) ↑
NIQ (CGA by NIQ) with RSM (2026). Bright September gives hospitality groups their best growth of 2026 (NIQ RSM Hospitality Business Tracker) (opens in a new tab) ↑
Office for National Statistics (2025). JOBS02: Workforce jobs by industry (September 2026) (opens in a new tab) ↑