Labour is usually the largest controllable line in a site P&L, and it is usually the least examined. Rotas get copied forward, adjusted for holidays, and rarely tested against what demand actually did.
The result is a persistent mismatch: over-resourced quiet periods that no one notices, and under-resourced peaks that cost sales and service quality at the same time.
Start with the curve, not the headcount
Plot transactions or covers in fifteen-minute increments across a full trading week. The shape is almost always sharper than managers expect, and the peaks are narrower.
Deploy against the shape. In practice this means shorter, better-targeted shifts around peak and honest resourcing of the trough — not a blanket headcount reduction.
Protect the standard while you do it
Labour reduction that damages service is not a saving; it is a deferred cost. The controls that make it safe are simple: a defined minimum staffing floor, a named bottleneck station, and a weekly review of the gap between planned and actual.
Done properly, a labour rebuild returns single-digit percentage points of cost while improving peak-hour experience. Done as a cost-cutting exercise, it returns the cost with interest.