Understaffing a peak cost 41 stores 6.15% of sales
Mani, Kesavan and Swaminathan took hourly traffic, sales and labor data from 41 stores of a large retail chain and compared staffing against demand hour by hour. Their work was published in Production and Operations Management in 2015.
All 41 stores were systematically understaffed during a three-hour peak. Not most of them. All of them.
Matching staffing to the traffic pattern was worth 6.15% in lost sales and 5.74% in profitability.
Those are not small numbers for a business running on single-digit margins, and no new hires were needed to capture them. The reason no hiring was needed is the part most operators get wrong.
The 41 stores had enough hours, spread across the wrong times
The 41 stores were not short of people. They had enough hours on the payroll. The hours were spread evenly across the trading day while the customers were not.
So the same headcount that left a line at the counter at lunchtime was more than the store needed at ten in the morning. Both conditions cost money. The quiet hour costs you wages. The busy hour costs you sales, and the sales are the bigger number.
Your labor line appears on a report every week. Lost sales appear on no report at all, which is why this survives in businesses that are otherwise well run.
Forecasting error and scheduling constraint need different fixes
The paper separates the two causes, and the distinction matters because the responses are not the same.
Forecasting error. The store did not see the peak coming. Traffic was higher than the plan assumed, or the plan was built on an average rather than an hourly pattern.
Scheduling constraint. The store saw the peak coming and could not respond to it. Shift lengths, availability, break rules or changeover times meant the schedule could not bend to the traffic even though the traffic was predictable.
Most operators assume they have the first problem. Most of them have the second. Either way, the question underneath both is how many people an hour actually justifies.
An extra hour earns more than it costs at the peak, and less mid-morning
What you are looking for is the point where an extra hour still earns more than it costs, and that point is different at different times of day. At the peak it sits well above where most stores are scheduled. Mid morning it sits below.
Zeynep Ton at MIT studied the retailers who operate this way deliberately. Costco, Trader Joe’s, QuikTrip and Mercadona all run with more staff than their competitors, and all of them make more money per store. Her explanation is about conformance rather than service: a thin crew cannot complete the work properly, so shelves stay unstocked and the operation degrades in ways customers notice.
Her point about why the practice persists is the useful one. Pressure to minimize payroll and to meet a monthly target pushes a store manager to cut hours, and the cost of doing so appears later, somewhere else, attached to nothing in particular.
Four operating decisions make a well-staffed store profitable
Extra hours on their own change nothing. Ton’s work across The Good Jobs Strategy and The Case for Good Jobs identifies four operating decisions that turn a fuller crew into a better result.
- Focus and simplify. Fewer products, fewer promotions, fewer one-off tasks. Every item you remove is time returned to the floor.
- Standardise and empower. Fix the routine work so it runs the same way every time, then let the crew decide the rest without asking.
- Cross-train. One absence has three possible fills rather than one, and quiet hours stop being idle.
- Run with slack. Staff above the minimum so people have time to serve, restock and fix problems in the moment.
Later implementations at Quest Diagnostics, Sam’s Club and Mud Bay report lower turnover, higher sales and better productivity, so the approach is not limited to the four retailers it was first observed in.
Ton’s four decisions are also the answer to an owner who treats good conditions as something you afford after a good year.
Staffing reaches profit through the customer, and 7,939 business units show the link
The chain from staffing to money is not direct, which is why the cost stays invisible on a report.
Heskett, Sasser and Schlesinger set out the service-profit chain in Harvard Business Review in 1994, from five years of research across American Express, Southwest Airlines, Taco Bell and Ritz-Carlton among others. The chain runs from internal service quality to employee satisfaction, then to customer loyalty, then to profit.
Harter, Schmidt and Hayes tested the middle of that chain at scale. Their meta-analysis in the Journal of Applied Psychology covered 7,939 business units across 36 companies and found consistent business-unit relationships between employee satisfaction and five outcomes: customer satisfaction, productivity, profit, turnover and accidents. The strongest relationship was with turnover, then customer satisfaction.
For an operator the reading is straightforward. Understaffing the peak does not subtract sales the way a discount does. Understaffing degrades the experience, and the lost sales arrive later, attached to nothing you can point at.
Two numbers you already have will find your own gap
None of that helps until you know where your own gap sits, and finding it takes two numbers you already have. Plot hourly sales against hourly labor for one store across one full week.
- Pull hourly sales for one store, for one full week. Every point of sale system will export this.
- Pull hourly labor for the same store and the same week. Scheduled hours are fine. Actual clocked hours are better.
- Plot both on one chart, with the hours of the day along the bottom.
Four things to look for on the chart
| What to look for | What it means |
|---|---|
| Sales climbing while the labor line stays flat | Your peak, and where the lost money sits |
| Labor high and sales low, usually mid morning and mid afternoon | The hours you can transfer |
| The same peak at the same time every day | A permanent schedule change, not a weekly judgment call |
| Shift changeovers sitting inside the busiest hour | The cheapest fix on this list |
The changeover row catches more operators than it should. A changeover during the peak takes two people partly out of service at the exact moment you need both.
Transfer hours before you add them
The first pass costs nothing. In most operations the total stays the same and the shape changes.
Set shift start times against the traffic, not against the clock. A store whose peak begins at 11:30 does not want a shift starting at noon because noon is tidy.
Take changeovers out of the peak. Fifteen minutes either side is usually enough.
Check the pattern by day, not only by week. Saturday’s shape is rarely Tuesday’s shape, and a single weekly template guarantees you are wrong on at least one of them.
Re-run the chart in a month. The point of the exercise is a habit rather than a one-off audit.
Labor percentage tells you nothing about where your hours sit
Labor percentage is a real constraint and nobody is suggesting you ignore it. What the research changes is where you look when the number is under pressure.
Cutting hours is the fastest way to shift the percentage and the slowest way to find out what it cost you. Rearranging hours does not touch the percentage at all, and in the 41 stores it was worth more than 6% of sales.
If your labor budget is tight, the chart is a better first step than the scissors.
FAQ
- How many weeks of data do I need before the pattern is real? One week shows you the shape. A month tells you which parts of that shape repeat. Run one week first, because the peak is usually obvious enough to act on immediately.
- My point of sale system will not export hourly data. What now? Count transactions by hand for three days at the hours you suspect. Rough counts find a three-hour peak perfectly well, and the exercise is about the shape rather than the decimal place.
- Does any of this apply to a business with flat demand? Retail and food operations are rarely flat once you look at them hourly. If your chart turns out flat, the exercise has cost you an afternoon and ruled out a common problem.
- Should I raise total hours or only rearrange them? Rearrange first, then measure. Most operators find the peak gap closes without a bigger wage bill, and you will have a much better case for extra hours if you can show what the rearrangement returned.
- What if my managers push back because their bonus is on labor percentage? Then the metric is the problem. A manager paid on a ratio will protect the ratio. Measure them on sales per labor hour instead, which improves when hours are placed well.
References
- Mani, Kesavan and Swaminathan, Production and Operations Management, 2015, 24(2).
- Ton, “The Effect of Labor on Profitability: The Role of Quality,” Harvard Business School working paper, 2009.
- Ton, The Good Jobs Strategy (2014) and The Case for Good Jobs (2023).
- Heskett, Sasser and Schlesinger, the service-profit chain, Harvard Business Review, 1994.
- Harter, Schmidt and Hayes, Journal of Applied Psychology, 2002, 87(2).
