Most restaurant operators know they waste food and overschedule staff. The problem is measurement, not awareness. This article covers the forecasting methods, models, and tools that connect historical sales data to purchasing and labor decisions, with real operator results showing what changes when forecasting improves.
As of October 2026, most restaurant operators already know they waste food and overschedule staff. The problem is measurement, not awareness. Without a reliable forecast, managers order based on gut feel, build schedules from last week’s template, and hope the numbers work out at period end. They usually don’t. A good forecast connects historical sales data to future purchasing and labor decisions, so operators buy what they will sell and staff for the volume they will actually serve.
This article covers the methods, models, and tools that make that connection practical, with real operator results showing what changes when forecasting improves. Our customers have reported an average 5% reduction in food costs, up to $100K in reduced labor costs, and a 15% improvement in forecasting accuracy across their locations.
Restaurant forecasting is the process of using historical sales data, seasonal patterns, local events, and current trends to predict future demand. That prediction then drives purchasing, prep, and staffing decisions.
The definition sounds simple. The execution is where most operators struggle. A forecast is different from a budget (a budget is a financial target you set; a forecast is a prediction of what will actually happen). A forecast is also a calculation, not a guess, built on data your POS already collects: covers, ticket averages, item mix, daypart volume, and sales by hour.
When done well, forecasting answers three questions every shift: How much food should we prep? How many people do we need on the floor and in the kitchen? And what sales volume should we expect so we can spot problems in real time?
The gap between operators who forecast and those who don’t shows up directly on the P&L. Our budgeting and forecasting tools connect POS data to financial planning so that the forecast is not a spreadsheet exercise disconnected from actual operations. We connect accounting, inventory, and scheduling to forecasts.
Based on 2026 industry data, the restaurant industry in 2026 faces a mixed picture. Input costs remain elevated, consumer spending patterns have shifted since the pandemic, and labor markets stay tight in most metro areas. Operators who relied on broad revenue growth to cover inefficiencies are finding that growth alone no longer works.
For detailed projections on where the industry is heading, our 2026 State of the Restaurant Industry Report covers sales trends, cost pressures, and what top-performing operators are doing differently. The short version: margin management matters more than top-line growth, and forecasting is one of the few levers operators can pull without raising prices or cutting menu quality.
Five common approaches show up in restaurant operations, ranging from simple to data-intensive:
Historical averaging: take the same week or day from prior years and average the results. Simple to calculate, but it ignores trends and one-time events. A Tuesday in March 2024 may not predict a Tuesday in March 2026 if you have added a patio, changed your menu, or a new competitor opened nearby.
Year-over-year comparison: compare the same period from last year, adjusted for known changes (price increases, new locations, menu shifts). This works well for established restaurants with stable operations.
Moving averages: use the most recent 4, 8, or 12 weeks of data to smooth out anomalies. Moving averages react to trends faster than annual comparisons but can lag behind sudden shifts like a local event or weather change.
Regression analysis: model the relationship between sales and external variables (weather, local events, holidays, marketing spend). More accurate than simple averages but requires more data and some statistical comfort.
AI and machine-learning models: algorithms analyze large datasets, identify non-obvious patterns, and adjust predictions automatically. These models improve over time as they ingest more data.
We support multiple projection models within our weekly forecasting feature, including Smart Forecast (an algorithmic approach), Last Week Actual, Last Year Actual, Customized, and Imported models. Operators can choose the model that fits their data maturity and adjust as they grow.
For a deeper look at which numbers to use when building your forecast, this guide on restaurant sales forecasting walks through the practical choices.
The 30/30/30/10 rule is a budgeting framework that allocates revenue into four buckets:
| Category | Target % of revenue |
|---|---|
| Food and beverage costs | 30% |
| Labor costs | 30% |
| Overhead (rent, utilities, insurance, etc.) | 30% |
| Profit | 10% |
This is a starting point, not a law. Plenty of successful restaurants run food costs at 25% or labor at 35%. The rule’s value is as a diagnostic: if your food costs sit at 38%, you know where to look first.
Forecasting makes the rule actionable. Without a forecast, you only discover you overspent on food or labor after the period closes. With a forecast, you can adjust purchasing and scheduling before the money is spent. Our budgeting and forecasting blog explains how to build forecasts that map to these cost categories so managers can make daily decisions against a financial framework, not just an operational one.
Some operators will disagree with this: the 30/30/30/10 rule is too generous on food cost for most full-service restaurants in 2026. Operators who treat 30% as acceptable rather than as a ceiling are leaving money on the table. Felipe’s Mexican Taqueria drove their CoGS from 28% down to 23%, well below the “rule,” by using Restaurant365 to give store managers real-time visibility into food and liquor budgets.
Building a sales forecast does not require a data science degree. It requires consistent data, a repeatable process, and the discipline to update weekly.
Pull at least 8 weeks of POS sales data, broken out by day and daypart. If you have 52 weeks of data, even better.
Identify your forecast metric. Most operators forecast net sales or gross sales. Some, like Taziki’s Mediterranean Café, use entrées per labor hour as a staffing benchmark alongside dollar forecasts.
Choose a projection model. For newer restaurants with limited history, a moving average of recent weeks works. For established locations, year-over-year comparison adjusted for known changes (price increases, menu changes, construction nearby) is more reliable.
Adjust for known events. Holidays, local sports games, school schedules, weather forecasts, and marketing promotions all shift demand. Build these into your weekly forecast as manual adjustments.
Convert the sales forecast into purchasing and labor plans. This is where most operators stop too early. A forecast that does not connect to a prep list and a schedule is just a number on a screen.
Compare actuals to forecast daily. Track variance and refine your model each week.
We automate much of this process by pulling POS data directly into the forecasting module, which generates projections by week, day, hour, and even 15-30 minute intervals. For operators who want to start with a manual approach, this blog on conducting a sales forecast with POS data provides a practical walkthrough.
The 15-minute increment detail matters more than it sounds. Paul Potvin’s team, featured on a Restaurant365 podcast, builds labor schedules in 15-minute increments tied to sales forecasts so managers know exactly how many people to have on the floor at any given hour. Their operations goal: run between 98% and 102% of theoretical labor hours based on actual versus forecasted sales.
Forecasting is a set of related predictions that feed different operational decisions.
Demand forecasting predicts covers, order volume, or transactions by daypart, and every other forecast builds on this foundation. If you expect 220 covers on a Friday night instead of 180, your prep list, your line staffing, and your server sections all change.
Inventory forecasting translates demand into purchasing quantities. The goal is ordering enough to avoid 86’ing menu items without buying so much that product expires or quality degrades. Maru Hospitality Group used our actual versus theoretical (AvT) reporting to identify that hand-cut ribeye generated more than 1,000 pounds of waste between 2018 and 2019, roughly $10,000 in lost sales. They switched to a vacuum-sealed product and kept that money.
Lorenzo Castillo, accounting business partner and interim general manager at Maru Hospitality Group, described the broader impact: “Overall, we reduced our menu options by about 20% so that we could focus on our more profitable and more popular items… We saw sales on items jump more than 20% on the very next menu, just by where we put the item and how we changed it.”
For operators looking to tighten inventory management, our inventory management software connects purchasing to recipes to actual usage.
Food-cost forecasting predicts what your ingredients should cost based on your projected sales mix. The gap between theoretical food cost (what you should spend given perfect portioning and zero waste) and actual food cost (what you really spent) is where profit hides.
Kabob House reduced food costs from 32% to 26% of sales after implementing Restaurant365. Owner Mr. Baqer put it plainly: “If you don’t have the numbers, you might think you’re making money when you’re not.”
Our food cost management tools help operators track this gap in real time rather than discovering it weeks later.
Labor forecasting connects your sales prediction to a staffing plan that answers a more specific question than “how many people do I need?”: how many people do you need at 11:45 a.m. on a Wednesday versus 7:30 p.m. on a Saturday?
Derek Belanger, Director of Operations at Barrio, described the shift: “Before, I would feel like we’re overscheduled, but it was only a feeling, not data. Today, we can forecast more accurately, we can schedule more accurately, and our operations are getting tighter.”
Barrio achieved a 21% food cost in recent quarters despite industry-wide inflationary pressures.
Eli’s Restaurant Group estimated they were losing approximately $1.5 million in potential profit annually by 2019 from inefficiencies in food, labor, and supply cost tracking. After adopting Restaurant365’s advanced features, they closed at least 75% of that gap, saving $600,000 in labor and another $500,000 in food and LBW costs. Restaurant365 replaced their static “set schedules” with dynamic, data-driven planning tied to sales forecasts.
Sales forecasting is the umbrella that ties the others together. It answers: what will we sell, when, and in what mix? Every other forecast depends on this number being close to right.
Dan Simpson, CEO of Taziki’s Mediterranean Café, described the operational change: “We get better day-to-day reporting now, and the next day we can see exactly where we are with our food and our sales.” His team found that insights from actual versus theoretical food cost reporting revealed inefficiencies that percentage-based metrics had previously concealed.
Forecasting methods fall on a spectrum from simple arithmetic to complex machine learning. The right choice depends on your data, your scale, and how much variance you can tolerate.
Simple methods work for single-location operators with stable demand patterns. A 4-week moving average of daily sales, adjusted manually for holidays and weather, is good enough to prevent the worst overstaffing and over-ordering mistakes.
Time-series models analyze sequences of data points over time to identify trends, seasonality, and cyclical patterns. These models (ARIMA, exponential smoothing, and their variants) require more data but capture patterns that simple averages miss. Snowflake’s documentation on time-series forecasting provides a technical overview of how these models work at scale. IBM’s forecasting resources cover the statistical foundations.
Machine-learning models go further by incorporating external variables (weather, events, promotions, competitor activity) and learning non-linear relationships in the data. These models improve with more data, which makes them particularly useful for multi-unit operators who can train models across dozens or hundreds of locations.
The practical difference: a simple model might tell you “Fridays average $12,000.” A time-series model might tell you “Fridays in March average $11,200, trending up 3% year over year.” An ML model might tell you “This Friday will do $13,400 because there is a concert at the venue two blocks away and the weather forecast is 72 and sunny.”
Paul Potvin’s team took this further by conducting a time and motion study that identified something like 563 tasks performed in a restaurant. They used those task-level standards to build labor hour models tied directly to sales forecasts. That level of granularity is unusual, and most operators don’t need it. But it shows what becomes possible when forecasting is treated as a core operational discipline rather than a finance exercise.
AI in restaurant forecasting is real, but the term covers a wide range of capabilities. Some vendors use “AI” to describe basic regression models. Others are running genuine machine-learning systems that improve over time.
Pattern recognition across large datasets. AI excels at finding relationships between sales and external factors (weather, events, day of week, seasonality) that humans miss or cannot process fast enough.
Automated model selection. Good AI systems test multiple models against your data and pick the one that performs best for each location, daypart, or menu category.
Continuous learning. Unlike a static spreadsheet model, AI forecasts get more accurate as they ingest more data. In October 2026, our customers have seen a 15% improvement in forecasting accuracy using the platform’s AI capabilities.
AI without clean data. If your POS data has gaps, your item counts are inconsistent, or your recipes are not entered accurately, no algorithm can compensate. Garbage in, garbage out.
Black-box predictions without operator context. An AI model might not know that your best server quit last Tuesday or that the street outside your restaurant is under construction. Human judgment still matters for adjusting AI-generated forecasts.
Point solutions that see only one slice of the business. Most AI tools for restaurants operate on either financial data or operational data, not both. R365 AI is built on the full restaurant P&L, connecting financial and operational data across accounting, inventory, workforce, and payroll in one system. That distinction matters because a labor forecast that does not account for food cost targets (or vice versa) optimizes one metric while potentially hurting another.
The Whataburger franchisee GVCS reduced time spent on financial reconciliation by over 50% and achieved more accurate forecasting after implementing Restaurant365, which helped maintain tighter profit margins. They are on track to triple their store count from 12 to 38 locations.
563 tasks. That is the number one operator identified in a single restaurant. Most restaurant management software treats scheduling as a shift-level problem. The operators getting the best results treat it as a task-level problem, matching specific labor to specific demand signals in 15-minute windows. That is where AI forecasting earns its value: telling you exactly how many hands you need at 11:45 a.m., not predicting yesterday’s average.
As of 2026, we serve 62,000+ restaurant locations and integrate with 100+ POS systems.
Blog Menu
Better forecasting reduces food waste and overstaffing because it replaces guessing with measurement. The operators in this article achieved their results by connecting POS data to purchasing and scheduling decisions through a single restaurant management system, then acting on what the numbers showed them.
Eli’s Restaurant Group saved $600,000 in labor costs and $500,000 in food and liquor, beer, and wine costs annually. Maru Hospitality Group reduced ribeye waste by more than 1,000 pounds. Kabob House cut prime costs from 72% to 61% of sales. Barrio hit a 21% food cost during an inflationary period. These are specific results from specific operators using Restaurant365.
Schedule a free demo of Restaurant365 to see how forecasting, scheduling, inventory, and accounting work together in one platform built for restaurants.
Last updated: October 2026.
Share this blog:
See why more than 50,000 restaurants use Restaurant365
Restaurant365 brings together accounting, operations, scheduling, and more in a flexible platform—empowering restaurants to choose the solutions they need and scale with confidence.
