Data Insights

Restaurant Sales Forecasting: Data Instead of Flying Blind

Updated on 9 min read

A reliable restaurant sales forecast combines the history of each weekday with weather, local events and seasonal patterns, including school holidays. Forecast deviation depends on data quality, concept and forecast horizon; there is no general guarantee. The forecast is measured against actual sales; how to calculate them properly alongside seven other values is shown in restaurant KPIs with formulas.

Many businesses still schedule staff without a written sales forecast. How large that share is across the whole market cannot be derived in good faith from our operating experience. So what matters is not a blanket market figure, but a measurable comparison of your own forecast with the actual sales that follow.

Not because operators are bad at business. Because gut feel is fighting three external variables at once that it cannot know: weather, local events and statistical patterns in historical data.

Contents

What a sales forecast actually is

A forecast is not a guarantee. It is a reasoned expectation based on data.

In practice I distinguish three levels of quality:

Level 1: gut feel. "Next Friday will be good, it's the weekend." No data, no documentation. Sometimes right, often wrong.

Level 2: historical averages. "The last four Fridays averaged €7,200, so I'll plan for €7,000." Better. But external factors are missing.

Level 3: forecast weighted by data points. Historical values combined with weather, public holidays, local events, school holidays and seasonal patterns. For a reliable assessment, the measurement period, data basis and forecast horizon must always be stated.

The economic effect does not follow directly from the MAPE. It only arises when a business demonstrably adjusts its hours budgets, orders or prep to the improved forecast. That is why forecast errors and the decisions they trigger should be measured separately.

The four decisive data points

1. Weekday and historical pattern

The most important data point is the simplest: what did your restaurant take on this weekday over the last eight weeks?

Not the average of all days. The average of exactly this weekday. A Wednesday and a Friday have fundamentally different starting values.

And not just the average, but also the spread. If your Friday sales swing between €5,400 and €9,200 (a 70% swing), a forecast is uncertain. If they sit between €7,000 and €7,800 (an 11% swing), you can plan very precisely.

The spread is the underrated indicator. It tells you how predictable your business is.

2. Weather

Weather is a massive sales factor in hospitality, but most operators do not plan with it because they do not know the historical correlation.

Hypotheses a business can test with its own data:

  • A beer garden can see significantly more demand on sunny Saturdays than in overcast weather.
  • An urban location can see fewer spontaneous walk-ins in heavy rain.
  • A lunch business can show a different demand pattern on very hot days than on normal office days.

These correlations are specific to each location. You can only know them if you link weather data with your historical sales data.

If you do that, you can check staffing specifically where there is a reliable weather correlation and adjust capacity in time. The change must never be derived from a generic weather rule alone.

3. Local events

A football match three kilometres away. A street festival in the next neighbourhood. A concert at the arena that takes away parking from your evening guests. A trade fair that brings 8,000 business travellers into town.

All of this affects your sales, positively or negatively. And all of it is in the city's events calendar.

Businesses that actively connect their events calendar with the forecast reduce mis-planning on event days by 30-40%. These are often the most expensive days to get wrong, because the deviation is largest.

A concrete case: a restaurant near a concert venue in Cologne was regularly 20% understaffed at major concerts without event planning, which meant overloaded staff, slow service and a lower average check. With an events feed in the planning tool: corrected straight away.

4. Seasonal patterns and school holidays

School holidays do not only affect family restaurants. They change the movement patterns of entire cities. The business-lunch restaurant loses regulars because more office workers are on holiday. The family restaurant gains because families eat out more.

If you do not factor these patterns into your forecast, you will be surprised every time, even though school holidays have had the same effect for decades.

The same goes for the Christmas season, Valentine's Day, Mother's Day and local festivals. These dates are predictable. So is their effect on your sales, once you have measured it.

How accurate is your forecast today?

Here is a simple way to measure it:

Take the last eight Fridays. For each Friday: what did you expect (in euros), and what did you actually take?

Forecast error = |expected − actual| / expected × 100

If you had no written expectation, estimate retrospectively: "What would I have expected on the Monday if someone had asked me?"

Calculate the average of all eight errors.

Below 8%: you plan well. Your system works. 8-15%: average. There is room for improvement. Above 15%: you are planning blind. The costs are real.

Most businesses land at 18-25% the first time. That is not personal failure; it is the statistical result of planning without data.

What forecast errors actually cost

Forecast too high → overstaffing

You expect €7,500, and €6,000 comes in. You have scheduled too many people.

Cost: assume four hours of overstaffing with three employees at €14 each: €168 per evening. With ten such evenings a month: €1,680. Per year: €20,160 per location.

Forecast too low → understaffing

You expect €5,800, and €8,200 comes in. You have too few people.

The costs are harder to measure, but real: service slows down, the average check drops (fewer upsells because the team has no time), guests wait longer and come back less often.

A conservative estimate: with five such evenings a month and an 8% loss of sales due to poor service, €2,600 of additional monthly sales are left on the table.

No forecast → no purchasing plan

If you do not know what is coming, you either buy too much or too little. Too much ends in write-offs. Too little ends in 86'd items on the menu, sold out by 8:30pm.

Write-offs, model calculation (assumed values, not an industry figure): a business with €80,000 monthly sales and poor purchasing planning writes off 1.5 to 3.5% of its food cost. At a 30% food cost percentage, that is €360 to €840 a month.

From gut feel to a structured forecast

Five steps you can put in place within three weeks:

Step 1: prepare historical data Export POS data for the last 90 days. If possible, by weekday and hour. That is your raw material.

Step 2: calculate weekday averages For each weekday: average and spread. That gives you the baseline forecast.

Step 3: add an events calendar Enter all relevant local events for the next four weeks: football matches, concerts, trade fairs, public holidays. For each event, estimate an uplift or reduction factor based on past experience.

Step 4: factor in the weather Check the weather forecast for the next seven days. Adjust your baseline forecast up or down based on the historical correlation in your business.

Step 5: compare forecast vs. actual Every Monday: what did you forecast last week, and what happened? Record it in a table. After four weeks you will spot systematic patterns in your mis-planning.

At the start this process takes 30-45 minutes a week. After three months, once the data basis is larger and the patterns are clear: 10-15 minutes.

Compare that with the €20,000 to €40,000 of planning errors you reduce as a result.

Conclusion: those who plan, win

A forecast is not a crystal ball. It reduces uncertainty; it does not eliminate it. But the difference between a 20% error rate and an 8% error rate is real, and measurable.

The three most important takeaways:

First: historical data alone is not enough. If you only take the average of recent weeks, you ignore weather, events and patterns.

Second: forecast accuracy can be measured. Start documenting your forecasts this week, even if they are only estimates at first.

Third: the costs of poor planning can be quantified. Overstaffing, understaffing, write-offs: all of it can be expressed in euros.

Kiosa automatically creates sales forecasts from your historical POS data, weather data and a local events feed, updated daily, for every location. You can also query forecast and actual sales in conversation: the Kiosa AI plugin brings your Kiosa data into Claude and ChatGPT. Book a 15-minute demo.


Sources


From analysis to action

What this is about: method and accuracy measures of a documented sales forecast.

Put it into practice: Sales forecast in Kiosa One View

More on KPIs, sales and forecasting

Sven Schneider

About the author

Sven Schneider — Founder of Kiosa and Ragazzi Concepts. 15 years in multi-unit restaurants; helped grow Mama Trattoria from 0 to 10 locations in Hamburg.

Kiosa Dashboard

All your numbers. One view. Instantly.

What you just read, Kiosa does automatically: every day, for every location, without you having to piece anything together.

Book a demo call
All articles

Data: Kiosa Gastro-Index (in German): 269,448 venues across Germany, Austria and Switzerland, measured.