Restaurant Revenue Management 101.

A restaurant sells seat-hours

A restaurant has a fixed number of seats and a fixed number of hours. What it sells is time at a seat. The measure is RevPASH: revenue per available seat-hour (Kimes).

Average check alone misleads, in the way ADR alone does. Four restaurants can each earn 7.20 a seat-hour: one at 40 percent seat occupancy and a check of 18, one at 60 percent and 12, one at 80 and 9, one at 90 and 8.

RevPASH = revenue ÷ (seats × hours) = seat occupancy × average check ÷ meal duration in hours.

Hot hours and cold hours

Measure RevPASH by the hour. An hour is hot at 70 percent seat occupancy or more, and cold below 40.

Never run one tool across the whole day. A discount that fills a cold hour gives money away in a hot one.

Duration

Measure each stage of the meal: greeting, order, first course, clearing, the bill. Shorten only the house’s waits. Never hurry the eating.

Guests accept a meal about 20 percent shorter without losing satisfaction, provided the minutes come out of the waiting before and after (Kimes, Wirtz and Noone).

Be modest about the gain. In Thompson’s simulations the real revenue gain from shorter meals was under a quarter of the simple arithmetic. Freed seats earn only when guests are waiting for them, so run the effort on nights that turn guests away.

Turns in a peak = length of the peak ÷ meal duration.

Table mix

Match the tables to the parties that arrive. One restaurant had 230 seats, almost all at tables for four, while two thirds of its parties were ones and twos. After the tables were rebalanced, seat occupancy in hot hours rose from 50 to 59 percent and revenue grew five points faster than at comparable outlets (Kimes; Thompson).

Tables that join are worth having only in rooms of about 100 seats where small parties dominate. In large rooms, dedicated tables win (Thompson).

Tables in use = parties per hour × share of that party size × minutes at table ÷ 60.

Reservations, no-shows and overbooking

Take a card guarantee first. In our worked example it brings no-shows down from 12 percent to 5, and it costs nothing. Overbook only on what remains.

Never overbook tables for six or more. Keep some tables for walk-ins as the buffer. If a party must wait, seat it within 15 minutes and bring a drink.

Pricing by demand

Guests accept different prices at lunch and dinner, by time of day, and through vouchers. They are neutral about weekday and weekend prices, and dislike paying for a better table (Kimes and Wirtz).

On nights that sell out, a prepaid cover charge that is fully redeemable against the bill secures the booking and costs the guest who arrives nothing. Use it for prime sittings only. Exempt hotel residents.

The menu

Each dish is judged on the money it leaves, and on how often it is chosen. That work has its own page: menu engineering, with a calculator.

Queues and service

A dining room planned to run completely full will queue. At 84 percent of tables in use, the average wait in one worked example is under five minutes. At 94 percent it is 27 minutes. Past 15 minutes guests leave.

Roster servers with a reserve. A server working at full stretch serves covers. One with a fifth of their time in hand serves guests.

Servers needed = guests ÷ (guests per server × (1 − reserve)).

Proving it

Take a baseline before changing anything: RevPASH by hour, seat occupancy, duration by stage. Change one thing. Measure again after eight to twelve weeks, against a comparable outlet where you have one. Without a baseline there is no proof.