Chapter 10 — Key Takeaways

The core claims

1. A menu is five documents at once — a sales instrument, a production schedule, a purchasing specification, a labor model, and a brand statement. Jobs three and four together are prime cost, which means the menu sets both halves of the number that predicts survival before you have hired anyone or bought anything.

2. You do not buy ingredients for dishes. You buy deliveries and decide what they become. That reframe is the whole of §10.3, and it is why the order guide is written from the cross-utilization map rather than from the menu.

3. Build the item list by subtraction, through three filters in order: concept fit, then skill fit, then equipment fit. Skill fit — can the cooks I can actually hire produce this on the four hundredth repetition? — kills more good dishes than the other two combined, and first-time operators skip it because the chef-owner is standing right there being able to do it.

4. Every additional item costs about \$3,020 a year before you buy an ingredient for it — prep labor, spoilage, ordering and counting, training — or \$6.45 a plate on an item selling nine a week. None of it appears on a P&L as its own line. It appears as cost drift.

5. Item count is a throughput variable. The menu decides whether four people on a line can do 95 covers. Bellwether moved two of eight entrées off the hearth, which is worth about twenty fire actions a night and the difference between 68% peak utilization and a line that cannot survive.

6. A standardized recipe carries no money. Quantities in weights, method, hold and reheat, portion, ticket-time target, and a plain description of what correct looks like. The cost card is Chapter 11's document — and Chapter 11 cannot cost "a handful."

7. Most menu psychology is more confidently repeated than it is evidenced. Primacy, recency, and the fact that people scan rather than read are well supported. The golden triangle is not established; the dollar-sign finding is weak; and some of the most-quoted research on descriptive menu language has been corrected or retracted. Design so that you win either way.

8. A price ladder must reconcile to the check. Bellwether's twenty-two prices produce \$33.12 of food per dinner cover — not by assertion but by arithmetic that resolves in three columns, and that is falsifiable from day one of point-of-sale data.

9. Cross-utilization has limits and they are real: it concentrates supply risk, it can make a menu taste the same, it can buy food cost with labor at a bad exchange rate, and it fails at very small volume.

The arithmetic

$$\text{Food per cover} = \sum_{\text{categories}} (\text{attachment rate} \times \text{weighted average price})$$

$$\text{Weighted average price} = \sum (\text{item price} \times \text{share of category orders})$$

$$\text{Hidden annual cost of an item} = \text{prep labor} + \text{spoilage} + \text{ordering} + \text{training}$$

$$\text{Peak fire utilization} = \frac{\text{fire actions per hour at peak}}{\text{positions} \div \text{average dwell} \times 60 \times \text{packing efficiency}}$$

Bellwether's menu, in one block

Key terms

menu (as a document) · menu categories · cross-utilization · item count · standardized recipe · menu psychology · price anchoring · menu panel · sweet spot · price ladder

What you should be able to do Monday morning

Take any restaurant's menu, count the items, and for each one name the station it fires from and one other item that shares a component with it. The items where you cannot do either are your candidates for removal — and you will find more of them than you expect.

Then do the arithmetic in the other direction: multiply your category attachment rates by your weighted average prices and see whether the number you get is the average check you have been telling people you run. If it is not, one of the two is wrong, and it is worth an hour of a Tuesday to find out which.