Restaurant Sales Forecasting Methods: Predict Revenue With Precision in 2026

You ordered 140 pounds of chicken thighs last Tuesday because that is what you always order. Sixty-two pounds went unused, sat in the walk-in for four days, and ended up in the waste bin. Your Tuesday labor schedule had nine cooks on the line because last Tuesday was slammed. This Tuesday was not. You burned $680 in wasted food and $420 in unnecessary labor in a single shift.

This is what happens when restaurants run on gut instinct instead of forecasting. And it happens constantly. Estimates cited by the Environmental Defense Fund put restaurant kitchen loss at 4% to 10% of purchased food, which translates to $28,000 to $70,000 annually for a restaurant doing $700,000 in food purchases. Labor overscheduling quietly adds thousands more every month.

The fix is not complex. It is not expensive. It is sales forecasting, and the restaurants that do it well operate with tighter margins, less stress, and significantly more profit. This guide covers seven forecasting methods ranked by accuracy, complexity, and the type of restaurant they work best for.

Why Most Restaurants Forecast Badly (or Not at All)

Before diving into methods, it is worth understanding why forecasting remains so rare in an industry that desperately needs it.

Few independent restaurants use any formal sales forecasting method, and many of those that do rely on a simple spreadsheet updated monthly or less. The rest use what the industry politely calls "experience-based estimation," which means the owner or manager guesses based on how busy they feel last week was.

Here is the thing: experience-based estimation is not worthless. Veteran operators develop strong intuition. But intuition cannot account for the 47 variables that actually influence daily restaurant sales, from weather patterns and local school schedules to gas prices and competing restaurant openings. Structured forecasting can.

Let's break down the methods.

Method 1: Simple Moving Average (SMA)

The simplest forecasting method that actually works. You take the average sales from the same day of the week over the previous 4 to 8 weeks and use that as your prediction for next week.

Formula: Forecast = (Week 1 Tuesday + Week 2 Tuesday + Week 3 Tuesday + Week 4 Tuesday) / 4

If your last four Tuesdays did $4,200, $3,800, $4,500, and $4,100, your SMA forecast for next Tuesday is $4,150.

When to use it

• Restaurants with stable, predictable traffic patterns
• Quick-service and fast-casual concepts with low seasonal variation
• As a starting point when you have limited historical data (minimum 4 weeks)

Limitations

SMA treats all weeks equally. A freak snowstorm three weeks ago drags down your average just as much as a normal week. It also cannot detect trends. If your sales are growing 3% month over month, SMA will consistently underpredict. Typical accuracy: 82-88% at the daily level.

Method 2: Weighted Moving Average (WMA)

A smarter version of SMA that gives more weight to recent weeks. The logic is straightforward: last week's Tuesday is a better predictor of next Tuesday than a Tuesday from six weeks ago.

Common weighting: 40% most recent week, 30% two weeks ago, 20% three weeks ago, 10% four weeks ago

Using the same numbers: ($4,100 x 0.4) + ($4,500 x 0.3) + ($3,800 x 0.2) + ($4,200 x 0.1) = $4,170. The difference from SMA is small in this example, but it compounds over time, especially when your business is trending upward or downward.

WMA is the workhorse of restaurant forecasting. It is easy to implement in any spreadsheet, requires no statistical software, and delivers meaningfully better accuracy than SMA. Most restaurants that switch from guessing to WMA see an immediate improvement in purchasing accuracy. For a deeper look at how sales patterns evolve over time, see our restaurant sales trend analysis guide.

Typical accuracy: 86-91% at the daily level.

Method 3: Year-Over-Year Comparison With Growth Adjustment

This method uses the same week from the previous year as the baseline, then adjusts for your growth or decline rate.

Formula: Forecast = Last Year Same Week Sales x (1 + Annual Growth Rate)

If the first week of June 2025 generated $31,400 and your year-over-year growth rate is 7%, your forecast for the first week of June 2026 is $33,598.

Why this method matters

Seasonal restaurants, tourist-area operations, and any concept with strong calendar-driven patterns benefit enormously from year-over-year forecasting. A beachfront seafood restaurant in Myrtle Beach cannot predict July sales from April data. But last July's numbers, adjusted for growth, are remarkably accurate.

The catch is that you need a full year of clean data. New restaurants or concepts that have undergone significant changes (new menu, new chef, renovations) may find last year's data misleading. Our seasonal menu performance analysis explains how to account for menu changes in your year-over-year calculations.

Typical accuracy: 88-93% at the weekly level for established restaurants.

Method 4: Regression Analysis With External Variables

Now we are getting into serious forecasting territory. Regression analysis builds a mathematical model that quantifies how specific variables impact your sales.

Here is a real example. A casual dining restaurant in Denver ran a regression on 18 months of daily data and found these relationships:

• Rain: -8.4% average daily sales impact
• Temperature above 85°F: -5.2% (fewer people walk to restaurants)
• Local sports event (Broncos home game): +22.7% for restaurants within 2 miles of the stadium
• Holiday weekends: +14.1% Friday, +18.3% Saturday, -11.6% Sunday
• School in session vs. summer break: -6.8% during summer for family-oriented concepts

With these coefficients established, you can forecast next Saturday by starting with your WMA baseline and then adjusting for known variables. If next Saturday has a Broncos home game and clear weather, your baseline of $8,900 becomes $8,900 x 1.227 = $10,920.

Regression requires more data and slightly more analytical skill, but the payoff is substantial. Typical accuracy: 90-95% at the daily level when external variables are properly modeled.

Method 5: Daypart Decomposition

Most forecasting methods predict total daily sales. Daypart decomposition breaks the day into segments, typically breakfast, lunch, afternoon, dinner, and late night, and forecasts each independently.

Why does this matter? Because a restaurant can have a strong overall sales day while badly misstaffing individual dayparts. If your forecast predicts $9,200 for Thursday and you staff evenly across lunch and dinner, you might have too many cooks during a $2,800 lunch and too few during a $5,400 dinner.

Implementing daypart forecasting

1. Define your dayparts based on natural traffic breaks (when your POS shows distinct volume changes)
2. Calculate the historical revenue split by daypart for each day of the week
3. Apply your chosen forecasting method (WMA or regression) to each daypart independently
4. Use daypart forecasts to drive staffing, prep quantities, and reservation capacity

KwickView automatically segments your sales data into customizable dayparts, so you can see exactly when revenue peaks and valleys occur. Our guide to predicting busiest hours from POS data walks through the setup process step by step.

Typical accuracy improvement over whole-day forecasting: 4-7 percentage points for staffing precision.

Method 6: Item-Level (Bottom-Up) Forecasting

Instead of predicting total revenue and working backward, bottom-up forecasting predicts how many of each menu item you will sell and then calculates revenue, food cost, and prep requirements from those predictions.

This is the gold standard for purchasing accuracy. Here is why.

Knowing you will do $8,500 on Friday tells you nothing about whether to order salmon or chicken. But knowing you will sell approximately 42 salmon entrees, 67 chicken dishes, and 31 pasta plates tells you exactly how much of each protein to order, how much prep time to allocate, and which stations will be busiest.

The data requirements

Bottom-up forecasting requires item-level sales data, which means a modern POS system that tracks every item sold, not just revenue totals. You also need at least 8 to 12 weeks of item-level history to establish reliable patterns. Restaurants running on older POS systems that only capture check totals cannot use this method without upgrading.

The KwickOS POS captures item-level data automatically, and KwickView transforms that data into item-level forecasts that feed directly into prep sheets and purchasing guides.

Typical accuracy: 85-90% at the item level (daily), which translates to 93-96% accuracy at the total revenue level.

Method 7: Machine Learning Ensemble Models

The most advanced forecasting method combines multiple models, using algorithms that learn from your data and continuously improve their predictions. Ensemble models typically blend time-series analysis, regression, and pattern recognition into a single forecast.

You do not need a data science degree to use them. Modern restaurant analytics platforms handle the complexity behind the scenes. What you see is a daily forecast that accounts for day of week, seasonality, weather, local events, historical trends, and even the impact of your own marketing campaigns.

How to judge whether it is working

Do not take any vendor's accuracy claim on faith, including ours. Measure it in your own operation with four numbers:

• Forecast accuracy: forecast sales versus actual sales, by day, tracked as a rolling average
• Food waste: recorded waste as a share of purchases, before and after forecasting
• Labor cost: scheduled hours versus hours the forecast said you needed
• Inventory on hand: average dollars sitting in the walk-in and dry storage

Run the arithmetic on your own numbers. For a restaurant spending $25,000 per month on food, every 10% of waste you eliminate is worth roughly $2,500 a month, and a few fewer over-scheduled shifts each week adds up quickly on top of that.

How to Choose the Right Forecasting Method

The best method depends on three factors: your data maturity, your operational complexity, and your tolerance for manual work.

• Just starting out (less than 6 months of data): Use Simple or Weighted Moving Average. Get comfortable with the process before adding complexity.
• Established single location: Weighted Moving Average with year-over-year adjustment covers 90% of your needs. Add weather and event variables if you want to push accuracy higher.
• Multi-location or high-volume: Daypart decomposition combined with item-level forecasting. The operational gains are too large to ignore at scale.
• Data-driven operations: ML ensemble models via platforms like KwickView. Maximum accuracy with minimum manual effort.

Regardless of which method you choose, the single most important factor is consistency. A simple forecast updated weekly outperforms a sophisticated model updated monthly. Build the habit first, then upgrade the methodology.

Turning Forecasts Into Action: The Operational Playbook

Purchasing alignment

Your purchasing orders should flow directly from your sales forecast. If your forecast predicts 380 covers on Saturday with a known mix of 35% chicken, 25% beef, 20% seafood, and 20% vegetarian, you can calculate exact protein quantities with a safety margin of 10-15%. Compare this to the typical approach of ordering "about the same as last week" and you see why forecasting restaurants waste dramatically less food.

Labor scheduling

Map your staffing template to forecasted revenue by daypart. A common rule of thumb: schedule one front-of-house employee per $125-$175 in forecasted hourly revenue, and one back-of-house employee per $150-$200 in forecasted hourly revenue. Adjust these ratios based on your concept and service model. Our peak hour staffing analytics guide provides detailed benchmarks by restaurant type.

Prep planning

Use item-level forecasts to generate daily prep sheets. If you predict selling 55 Caesar salads on Friday, your morning prep should include enough romaine for 60 to 63 salads (a 10-15% buffer). This approach eliminates both under-prepping, which slows service, and over-prepping, which creates waste.

Cash flow management

Weekly and monthly revenue forecasts feed directly into your cash flow projections. Knowing that July will generate 18% more revenue than February helps you plan for seasonal payroll fluctuations, schedule equipment purchases during high-revenue months, and maintain adequate cash reserves during slower periods.

Common Forecasting Mistakes to Avoid

1. Using stale data. A forecast based on data from three months ago is barely better than guessing. Update your inputs weekly at minimum.
2. Ignoring outliers without investigation. That anomalous $14,000 Saturday might have been caused by a private event, a nearby concert, or a competitor closing temporarily. Investigate before removing outliers from your dataset.
3. Forecasting revenue without forecasting mix. Knowing you will do $9,000 on Thursday is useful. Knowing it will be 40% from your highest-margin entrees versus 40% from lower-margin specials is critical for purchasing and profitability.
4. Not measuring forecast accuracy. Track your predictions against actual results every week. Calculate your Mean Absolute Percentage Error (MAPE). If it is above 15% consistently, your model needs recalibration.
5. Treating forecasts as fixed plans. Forecasts are educated predictions, not guarantees. Build flexibility into your staffing and purchasing to accommodate the 5-10% variance that even the best models produce. Our revenue forecasting model guide explains how to build in appropriate buffers.

Frequently Asked Questions