Labor variance: how to find the four points you are losing

That last sentence is the part most owners get wrong. Labor feels like a general condition, something the whole restaurant is doing to you. It is almost never general. It is a Monday lunch, a Sunday close, and one overlapping section on Thursday.

In the restaurants I ran, the labor problem was a scheduling problem long before it was a wage problem. Nobody was overpaid. People were scheduled for volume that did not arrive, and nothing in the week was measured at a level fine enough to notice.

How big is the gap between a profitable restaurant and a losing one?

Roughly nine points of labor, and that is not a guess.

The National Restaurant Association's Restaurant Operations Data Abstract, 2025 edition, collected financial and operating data from more than 900 restaurant operators nationwide. Salaries and wages including benefits, as a median share of sales in 2024, came in at 36.5% for full-service restaurants. Split the same population by result and the medians separate hard: operators who reported a pre-tax profit ran 34.2%, and operators who reported a loss ran 42.9%.

Limited-service shows the same shape at a lower level. All respondents 31.7%, profitable operators 30.0%, loss-making operators 34.1%.

The distance between a profitable full-service restaurant and a losing one is 8.7 points of labor. Not a philosophy of management, not a concept score. Points on a line you already receive every month.

Two things about that number before you use it. The Association states plainly that the Abstract data is "not intended to represent standards or goals for individual restaurants," and is meant as a management tool for gauging performance. And the figure is salaries and wages including benefits. If you are comparing your wages-only number against 36.5%, you are comparing two different quantities and your restaurant looks better than it is.

Where does labor variance actually come from?

Four places, in roughly the order they matter.

Hours that were scheduled out of habit. This week's schedule was built from last week's, which was built from the week before, back to a season that no longer exists. Nobody ever re-justified the second server on Tuesday lunch. They are on the schedule because they have been on the schedule.

Punches that do not match the schedule. Ten minutes early at the start, twenty minutes late at the end, a break that got skipped because the room got busy. Per person it is nothing. Across a full crew, six days a week, it is a line item.

Overtime nobody decided on. Overtime almost never gets approved. It gets discovered. Someone covered a call out on Friday and crossed the threshold on Saturday, and the manager who made that call did not know they were making it.

Volume that did not show. The schedule was correct for the sales you forecast. The forecast was wrong. That is a forecasting failure being paid for out of the labor line.

Notice that only the last one is about the market. The other three are yours.

There is a reason habit is the most expensive of the four right now. The Association estimates that average hourly earnings of restaurant employees are up 41% since February 2020. A staffing template built before that and copied forward since has been quietly re-priced every year without anyone re-deciding it.

Why does the payroll number surprise you when the schedule looked fine?

Because a schedule is a plan and payroll is a record, and nothing in most restaurants forces the two to be compared while there is still time to act.

The schedule gets posted. The week happens. Fourteen days later a number lands on a statement, aggregated across every shift, every station, and both weeks. By then you cannot tell a Monday from a Saturday.

A monthly number cannot diagnose a daily problem. It can only confirm that one exists.

If nobody ever trained you to read the statement itself, start there, because labor variance is invisible until you can see the line it lands on and what sits next to it. Reading a restaurant P&L without a finance background is a shorter lesson than people expect.

How do you find the shifts that are causing it?

Pull three columns for every shift in the last four weeks: scheduled hours, actual hours, and net sales.

Then compute sales per labor hour for each shift and sort worst to best. You are not looking for a trend. You are looking for the handful of shifts that own most of the number.

What tends to surface:

• A daypart that carries a full floor for a fraction of the covers, usually a weekday lunch that stopped working two years ago
• A closing crew scheduled to leave at the same time, so the last hour is four people doing a one person job
• An opening overlap where the morning prep cook and the opening line cook are both paid to be there for the same forty five minutes
• A manager on the clock during a shift a keyholder could run

None of that is exotic. It is just invisible at the monthly level and obvious at the shift level.

Do this before you touch a single person's hours. The instinct is to cut across the board, which punishes the shifts that were already tight and does nothing to the ones that were not.

What does one bad shift actually cost over a year?

Here is the arithmetic, with numbers you can swap for your own.

Take a full-service restaurant at $1.2M in annual net sales. Apply the Association's medians to that volume and the range is easy to see:

The distance from the all-respondent median to the profitable median is 2.3 points, which on $1.2M is $27,600 a year. The distance from losing to profitable is 8.7 points, or $104,400.

Now work it from the other end, one shift at a time.

Take Tuesday lunch. Five people scheduled, six hours each, thirty labor hours. Use a loaded hourly cost of $22, meaning wages plus payroll taxes plus whatever benefits you carry. Substitute your own figure here, because this is the input that varies most between restaurants. Thirty hours at $22 is $660 of labor.

That shift does $1,800 in net sales. Sales per labor hour: $1,800 divided by 30, or $60.

Now find the actual waste inside it. Two of those five start forty five minutes before there is anything for them to do, and the fifth person is scheduled through a two hour tail with four tables in the room. That is roughly six hours of labor that bought nothing. Six hours at $22 is $132.

$132 on one shift, once a week, is $6,864 a year.

Four shifts like that, and you are at roughly $27,000. That is the entire gap between the median full-service restaurant and the profitable one, and it was four shifts. Not a wage decision, not a menu decision, and nothing a guest would have noticed.

Run the same arithmetic on your worst six shifts before you do anything else. Most operators find three or four real ones and two that turn out to be correctly staffed and simply slow, which is a different problem.

Why does the fix stop working by week three?

Because cutting hours is a decision and holding the cut is a system, and most operators only make the decision.

You trim the schedule. Week one it holds. Week two someone calls out and the manager covers with an extra body, correctly, because the alternative was a bad service. Week three that extra body is back on the template. By week five the schedule looks like it did before, and nobody made a choice to put it there.

The thing that holds a labor number is not willpower. It is a small number of rules that survive contact with a full dining room, which is what an operating control actually is and why most written policies are not one.

Labor and food sit on the same statement and get hunted the same way, but they do not hide in the same places. Food cost variance shows up somewhere else entirely, and looking for it in the schedule will waste a month.

Where does this advice break down?

In three places worth naming.

When the wage floor is set for you. California's AB 1228 set a $20 an hour minimum for fast food workers at chains with more than 60 locations nationwide, effective 1 April 2024. If that applies to you, part of your labor line is legislation, not scheduling, and the lever moves to hours and throughput rather than rate.

Be careful about what you conclude from that law, because the evidence is genuinely split. Research summarised by the Cato Institute and circulated as NBER working papers found California fast-food employment down 2.7% against the rest of the country between September 2023 and September 2024, or 3.2% after adjusting for pre-existing trends. Other work using a synthetic difference-in-differences approach produced an employment elasticity of −0.04, with specifications ranging from −0.29 to +0.26, which is statistically indistinguishable from zero, and early work from the UC Berkeley Labor Center found higher pay and modestly higher prices without a measurable employment drop. Both camps do agree on one finding: the separation rate fell. Anyone quoting you a confident job-loss number from that law is taking a position, not reporting a result.

When the comparison is not like for like. Service model, tip structure, whether managers sit in labor or in G&A, and whether benefits are inside the number all move it by points. The Association's own caveat exists for exactly this reason. Your eight-week trend against yourself is a better instrument than any median.

When you are already at 34.2%. If your labor is running at the profitable median and you are still not making money, the problem is somewhere else on the statement. The Association's pre-pandemic picture of a typical independent was food around 33 cents of every sales dollar, labor around 33 cents, everything else around 29%, and a pre-tax margin near 5%. Squeezing a line that is already at benchmark to protect a 5% margin is how operators burn a quarter of management attention for nothing.

What to do this week

Pull scheduled versus actual hours for the last four weeks, by shift. Most scheduling software exports this in two clicks. If yours does not, do it by hand for two weeks. It takes an hour.

Add net sales to each row and calculate sales per labor hour. Sort it. Look at the bottom six shifts.

For each of those six, write one sentence about what happened. Not a plan, just a cause. "We schedule three servers and do thirty covers" is a cause. "Labor was high" is not.

Then run the $132 arithmetic above on each one, using your own loaded hourly cost. You are converting a feeling into an annual number, and the annual number is what makes the conversation with your manager concrete.

Then change one thing on one shift and watch it for two weeks. One change is measurable. Five changes at once tell you nothing about which one worked.

Questions about your labor line?

Email me at eric@seod.com with your scheduled hours and actual hours for one bad week, by shift. No sales figures needed if you would rather not send them. I will tell you which shifts are carrying the variance and what I would look at first.

I ran multi unit restaurants for sixteen years and managed more than $54M in annual P&L, so this is a familiar spreadsheet. The same habit of finding the number that moved and then finding the cause applies to reading a monthly marketing report critically, which is the other place operators get handed a total with no diagnosis attached.

Otherwise keep reading more on running the operation.