When labour runs high, the instinct is to cut hours. It is the fastest lever and the most visible, and it is usually the wrong one to pull first — because high labour cost is far more often a distribution problem than a volume problem.
You can have the right number of hours in the week and still be paying too much for them, because they are in the wrong places.
Start by separating the two questions
There are only two reasons labour cost is too high:
- You are buying more hours than the sales support
- You are buying the right number of hours at the wrong times, or at premium rates
These need opposite responses. Cutting hours when the real problem is distribution means you end up understaffed at peak and still overstaffed at open — worse service and the same cost. So establish which one you have before acting.
The quickest diagnostic: pull four weeks of sales by hour alongside labour hours by hour. If the two curves have roughly the same shape and labour is simply too tall, it is volume. If the shapes do not match — labour flat while sales peak and trough — it is distribution, and that is the more common finding.
Overtime is a structural signal, not an exception
Occasional overtime is normal; someone is sick, a party books late. Overtime that appears every week is not an exception, it is part of your schedule, and it is the most expensive hour you buy.
Work the arithmetic on your own numbers. Take one extra overtime hour per shift, five shifts a week, at time-and-a-half on your actual wage. Multiply by 52. For most independent operations that single line is larger than any menu change you were considering, and unlike a menu change it requires no guest to behave differently.
Recurring overtime almost always traces to one of three things: a schedule built before you knew your real sales pattern, a role only one person can cover, or a close that routinely runs past its planned time. All three are fixable without touching headcount.
Build the schedule from sales, not from last week
Most schedules are copied forward. Last week's schedule becomes this week's, adjusted for who asked for time off. That means the schedule encodes whatever the business looked like whenever someone last built it from scratch — often years ago, at a different volume.
The alternative is not complicated:
- Pull sales by day-part for the last 8–12 weeks, so seasonality is visible
- Set a target labour percentage per day-part, not just an overall target — lunch and dinner rarely support the same ratio
- Schedule to the pattern, then set a guardrail: who leaves first if sales come in under, and who is called if they come in over
- Compare scheduled to actual every week and ask why they differ
That last step is the one that makes the rest hold. A schedule nobody reviews against actuals drifts straight back to habit within a month.
Cross-training is a labour strategy
If one person is the only one who can run a station, you are not staffing that station — you are staffing that person. It forces overtime when they are needed, idle cost when they are not, and it means their holiday is an operational event.
Cross-training reads like a training initiative and functions as a cost control. Two people who can each cover two stations give you scheduling flexibility that no amount of careful rota-building can replicate. This is one of the practical returns of having written procedures and role scorecards: you cannot cross-train quickly against knowledge that only exists in someone's head.
What not to do
Do not cut the shift that protects the guest experience. The hours that look most cuttable on a spreadsheet are often the ones absorbing variability — the floater, the early prep, the extra closer. Removing them does not remove the work; it moves it onto people who are already busy, and it surfaces as slower tickets and inconsistency.
Do not chase a benchmark percentage you did not set. A target labour percentage is only meaningful against your own format, service style and price point. A full-service restaurant and a counter-service operation have no business sharing a number.
Do not fix labour in isolation. Labour, menu mix and prep planning are one system. Trimming hours while leaving a menu that requires heavy prep just relocates the problem to quality.
Where to start this week
Pull four weeks of labour reports and find every overtime hour. Group them by person, role and day. Nearly always, a pattern appears — one role, one shift, one recurring cause. Fixing that single pattern is usually worth more than a general instruction to everyone to watch their hours, and it costs nobody their shift.
If labour is one of several things that feel off at once, the busy-but-not-profitable problem covers how to work out which to address first.