Chapter 14 — Key Takeaways
The core claims
- A service is a manufacturing run with no inventory buffer and a customer watching. You cannot build ahead, ship late, or substitute. Everything that survives a service was built before four o'clock.
- Mise en place is a production system, not a virtue. A virtue depends on which cook is working; a production sheet with every par traced to a forecast survives a personnel change.
- Every kitchen has a binding constraint, and it is usually not the one on the floor plan. Bellwether's dining room allows about 151 covers. Its wood fire allows about 144. The fire wins.
- Capacity numbers are computed against an assumed arrival shape. 142 covers broke a kitchen rated for 144 because forty of them arrived as one ticket. Write the assumption next to the number.
- The pass is the coordination layer, and losing it is worse than losing a pair of hands. A station going down does not degrade a kitchen linearly, because covering it pulls the chef off the rail.
- Par cooking is what makes a small hot line possible. It converts eleven-to-fourteen-minute live cooks into three-minute finishes and moves the work into cheap, flexible prep hours — a prime-cost trade inside the same total.
- 86ing is a lever, not an accident. A capacity 86 is a priced decision made in advance. A stock-out 86 is a production-sheet failure disguised as one.
- Ticket time is a proxy, and proxies degrade when you push on them. Read it next to comps, remakes, and reviews, or do not read it.
- Resilience is about which decisions are still available and for how long. At 3:41 the chef held five options; at 7:20, one. Every system in this chapter exists to move discovery earlier.
The formulas and rules of thumb
| Quantity | How to compute it |
|---|---|
| Station throughput ceiling | working positions × (60 ÷ average dwell in minutes), less an allowance for tending |
| One fire slot | 60 ÷ hourly ceiling — at Bellwether, 2.14 minutes |
| Peak-hour station demand | (entrées × peak share × station share) + starters and sides routed to that station |
| Cover ceiling | solve peak-hour demand = hourly ceiling for covers |
| Overrun, in minutes | (demand − ceiling) × minutes per slot |
| Long-dwell item cost | an item dwelling 2× the average consumes 2 slots, not 1 |
| Par level | forecast × a stated coverage factor, written into the day of the week |
| Batch cadence | compare (labor per week + quality waste) across cadences — do not assume big batches are cheaper |
Bellwether's frozen kitchen numbers: hearth sustains 28 items an hour · peak fire utilization 67% on a 95-cover plan night (18.7 against 28) · entrée split 44% hearth / 34% plancha / 22% sauté-oven · entrée attachment 0.90 · plan week 62 / 78 / 92 / 120 / 123 = 475 covers · design ceiling ≈ 151 covers, hearth ceiling ≈ 144 · ticket-time standard 14 minutes for entrées, 22-minute not-to-exceed, ribeye tracked separately at 24 / 30.
The key terms
mise en place · prep list · production sheet · ticket time · expediting · 86ing · all day · batch cooking · pre-shift / stretch
What you should be able to do Monday morning
Walk your own line with a stopwatch and a notebook, count the working positions and time the average dwell at your slowest station, and compute the number of items an hour it actually sustains. Then take your busiest forecast night, work out the peak hour's demand on that station, and find out whether your restaurant's ceiling is the room or the equipment — and by how much. If the answer is the equipment, you now know something about your revenue forecast that nobody in the building knew on Friday.
Then do the cheap half: write the par on one production-sheet line as a stated multiple of a stated forecast, and put a physical count at three o'clock on the one item you cannot replace mid-service.