The Roster is a Labor-Cost Forecast: Building a Schedule that Protects Margin
Most restaurants build the roster from availability. The kitchens that hold margin build it from forecast. This is the exact process for turning next week's covers into next week's shifts.
Every roster you build is a labor cost forecast. Every shift you post is a bet on what next week's covers will look like. Most operators do not think of it this way. That is why most operators lose 2 to 4 points of margin every quarter to rostering slop.
The kitchens that hold labor at 24 to 28 percent of revenue year over year all follow a variant of the same forecast-first process.
The four inputs to a good roster
Covers forecast. Last four same-day-of-week covers, adjusted for known events. Public holidays. Big local sporting fixtures. A booked large table. A supplier no-show that pushed a menu item out. Every operator's forecast is only as good as its adjustments.
Sales mix forecast. Not just covers. What mix. If your covers are flat but your average check is climbing because a new tasting menu launched, front of house needs more experienced servers on the floor. Back of house needs one more line cook on the plating station.
Skill map. Every staff member's certified station list. Grill, sauce, pastry, cold larder, plating, bar, floor. A shift is not built by filling a slot. It is built by ensuring every station is covered by someone signed off to run it.
Wage rates. By role, by station, by weekday versus weekend. Overtime rules. Public holiday multipliers. Break requirements. The wage table is the constraint that makes rostering an optimisation problem instead of a wishlist.
The Sunday build cadence
The rosters we respect are always built on the same day of the week, before the next week starts. Never mid-week. Never the day of. The sequence is:
-
Load next week's covers forecast. By day, by daypart. A Friday dinner service is not a Friday lunch is not a Sunday brunch.
-
Overlay the sales mix forecast. Any daypart where the mix requires an extra station gets flagged.
-
Populate the required shift skeleton. For every daypart: how many grill, how many sauce, how many pastry, how many floor, how many bar. This is the demand curve.
-
Match to the skill map and availability calendar. Any staff member available and certified for a required station is placed. Everyone else is placed against their next-best station.
-
Compute the wage total for the drafted week. Compare to the target labor percent (usually 26 percent). Anything above 28 percent gets pruned before publish.
-
Publish 8 days ahead. Give staff time to plan. A roster that lands the night before is a roster that will fall over.
The two rules that make it work
Rule 1: Every station has a primary and a secondary. If the primary is out sick, the secondary steps up. If both are out, the roster fails safely and the shift needs a substitute. No station goes uncovered.
Rule 2: Overtime is priced, not accepted. Overtime is 1.5x wages. A shift that would tip a staff member into overtime gets flagged. The manager either splits the shift, calls in a substitute, or accepts the cost with a documented reason.
What good rostering looks like on the P&L
- Total labor at 24 to 28 percent of revenue, consistent quarter over quarter.
- Overtime under 3 percent of total hours worked.
- Zero shifts published later than the roster publish rule allows.
- Every station coverage failure logged with cause, resolution, and corrective action.
A roster is not a scheduling task. It is a weekly bet on next week's business. The kitchens that treat it that way keep their labor line flat. The kitchens that do not, watch it drift up two points per quarter until the business is unsustainable.