What Hospitality Can Learn from Retail About Demand-Based Scheduling

Large retailers have spent decades and millions of dollars perfecting demand-based scheduling. Their labor models incorporate sales forecasts, foot traffic patterns, weather data, local events, and dozens of other variables to predict exactly how many people they’ll need, hour by hour.

Hospitality has been slower to adopt these practices. Many restaurants and hotels still schedule based on “what we did last year” or manager intuition. But the same principles that help retailers optimize labor costs work just as well—sometimes better—in hospitality.

The Core Concept: Match Labor to Demand

Demand-based scheduling starts with a simple principle: schedule more people when you’re busy and fewer when you’re slow. While this sounds obvious, traditional hospitality scheduling often misses the mark.

Think about how most restaurant schedules are built: the manager looks at the day of the week, maybe checks for obvious events (Valentine’s Day, Mother’s Day), and schedules roughly the same number of people they always schedule for that day. This ignores enormous variability in actual demand.

Retailers think differently. A Target store doesn’t schedule the same number of cashiers every Tuesday. They look at projected sales for this specific Tuesday—accounting for school calendars, weather, paycheck cycles, competitor promotions, and dozens of other factors.

What Data Should You Be Using?

Hospitality businesses have access to more demand signals than they typically use:

Reservations and bookings: Hotels know occupancy weeks in advance. Restaurants with reservation systems know covers by hour. This is your most reliable demand signal—use it for more than just table assignments.

Historical patterns: Your POS data contains treasure. What were actual covers last Tuesday? What about the Tuesday before that? What about the same Tuesday last year? Look at 4-8 week patterns minimum, and weight recent weeks more heavily.

Weather: Rain suppresses walk-in traffic for most restaurants. Sunny weekends boost hotel pool bar sales. Snow days empty downtown lunch spots but fill suburban family restaurants. Weather forecasts are available two weeks out—incorporate them.

Local events: A convention at the nearby center, a concert at the arena, a home game for the local team. These events drive or suppress traffic depending on your location and concept. Track what happens during these events and use that data for future scheduling.

Marketing activity: If your restaurant is running a promotion next week, or a positive review just hit the local paper, expect demand increases. Build this into your forecast.

Seasonal patterns: Q4 holidays, summer tourism, spring break, graduation season. These broad patterns should inform your baseline expectations by month and week.

Building a Simple Demand Model

You don’t need a data science team to improve your forecasting. Start with a basic model:

Step 1: Calculate your baseline. Average covers or revenue by day of week for the past 8 weeks. This is your starting point for any given day.

Step 2: Add a trend adjustment. Is business up or down compared to 8 weeks ago? Apply a simple percentage adjustment.

Step 3: Add known adjustments. Reservations on the books? Weather forecast? Local event? Adjust your baseline up or down accordingly.

Step 4: Convert to labor hours. If you know you need 1 server per 20 covers on average, multiply your expected covers by that ratio to get your needed server hours.

This won’t be perfect, but it will be far better than “we always schedule 4 servers on Tuesday.”

Applying Retail’s Granularity

Retailers don’t just forecast daily—they forecast hourly. A big-box store might have 20 cashiers at 5 PM on a Saturday and 3 at 9 AM on a Monday. They staff to the curve of demand, not to a daily average.

Hospitality can do the same. You know your peak hours. You probably already adjust somewhat—more servers for dinner than lunch. But how precise are you?

If your dinner rush runs 6-8 PM and you’re slammed during that window but overstaffed at 5 and 9, you’re losing money on both ends. Consider staggered start times: some servers at 4:30, more at 5:30, the last batch at 6:00. Similarly, staggered cutoffs to manage the post-rush wind-down.

What Retail Gets Wrong (And How to Avoid It)

Retail demand scheduling has earned criticism for practices hospitality should avoid:

Just-in-time scheduling: Some retailers post schedules with minimal notice, giving employees just days to plan their lives. This optimizes for the business but destroys work-life balance and drives turnover. Hospitality should forecast ahead enough to give reasonable notice.

Cutting hours mid-shift: “We’re slow, go home early” saves labor costs but frustrates employees who counted on those hours. Build reasonable buffers rather than sending people home constantly.

Split shifts: Working 11-2, going home, and returning 5-9 might match demand perfectly but is brutal on employees. Limit split shifts or compensate them appropriately.

Ignoring the human element: Algorithms can optimize for cost while destroying culture. The goal is informed scheduling, not purely algorithmic scheduling. Managers should use demand data as input, not as dictator.

Starting Small

You don’t have to implement a full demand forecasting system overnight. Start with one improvement:

If you don’t track covers by hour, start.
If you don’t look at weather before scheduling, start.
If you don’t check local events, start.
If you don’t compare your scheduled hours to actual demand afterward, start.

Each data point you add makes your scheduling more accurate. Each post-mortem comparing forecast to actual teaches you something. Over time, you’ll develop the same demand-responsive scheduling that makes retail so efficient—while avoiding the practices that make retail scheduling infamous.