Chapter 22: Front of House Operations: Service Flow, Table Management, Reservation Systems, and Guest Experience

"The kitchen decides how good dinner is. The host stand decides how many dinners there are, and whether any of them arrive on time." — constructed; the sentence a general manager eventually says to a new owner

Overview

Here is the decision that sits on the host stand at 6:35 on a Friday, and it is worth more money than any decision the chef will make all night.

Two people are standing in front of you with no reservation. You have a two-top open — table 14, just cleared, wiped, reset. The book shows the seven o'clock block filling in behind them. You can seat these two right now, in eight seconds, and they will be delighted with you.

Or you can look at the ticket rail through the pass window, see nineteen open checks where there should be fourteen, do the arithmetic that this chapter is going to teach you, and understand that every cover you put in a chair between now and 7:45 lands on a wood-fired hearth that clears twenty-nine covers an hour and is already being asked for thirty-six. Seat them, and their entrées arrive at 7:52 instead of 7:22, and so do the entrées of the four tables that got there before them. Hold them fifteen minutes with a drink at the bar, and nobody's food is late.

That is the entire chapter. The dining room is a scheduling problem with feelings. The room has a capacity, the kitchen has a different and smaller capacity, guests arrive in a shape nobody chose, and somebody standing at a podium with a tablet has to reconcile all three in real time while being warm about it. Chapter 7 drew the floor plan and found the room could physically hold about 151 covers. Chapter 14 measured the fire and found the hearth sustains 28 items an hour, which caps the kitchen at about 144 covers a night — and then handed us the sentence this chapter exists to answer: an hour is the wrong unit for a station that behaves like a queue.

So we are going to find the real number. Not the floor plan's number and not the kitchen's number — the number the room actually produces when demand arrives the way demand actually arrives, in a lump between seven and eight. That number turns out to be about 132 covers, it is reachable by three completely independent routes, and it changes what Bellwether's business plan should say.

Along the way we will do the unglamorous work: the twelve steps of a service and the standard time for each, four sections cut out of seventeen tables, the rotation that keeps one server from earning double another for the same shift, the quote, the waitlist, the fire, the check, and the manager's walk. None of it is complicated. All of it is the difference between 95 covers and 132.

In this chapter, you will learn to:

• Walk a table through a full sequence of service with a clock on every step, and name what each step is actually for.
• Cut a floor plan into balanced stations, rotate them, and see the income inequality that an unbalanced floor creates before your staff does.
• Run a host stand as production scheduling — quote honestly, cap covers per quarter hour, and refuse the seat that will cost you the room.
• Choose a reservation-to-walk-in mix and justify the inventory you hold back.
• Break turn time into its segments, and separate the eleven minutes you may take from the guest from the eighty-eight you may not.
• Read a queue: why splitting a large party into waves creates exactly zero capacity, proven in arithmetic rather than asserted.
• Defend or revise a seat-turn assumption in a business plan, with the numbers to back either answer.

Learning Paths

🏗️ Opening — this is one of the three or four chapters that decides whether your forecast is real. Do the arithmetic in §22.3 and §22.5 by hand against your own floor plan, and read the Business Plan checkpoint twice; it revises an assumption you have been carrying since Chapter 1. 📋 Managing — §22.1, §22.2, and §22.8 are your job description. If you already run a room, go straight to §22.5 and compute your own table cycle by segment. Almost nobody has. 🍸 Beverage — §22.4 makes the bar the most valuable twelve seats in the building for reasons that have nothing to do with pour cost: the bar is your waitlist's holding pen and your only unreservable inventory. §22.1's drink-order standard is the first promise the room makes. 🚚 Small Format — you have no host stand and no reservation book, which means the queue is literally visible on the sidewalk. §22.3's pacing arithmetic and §22.6's queue math transfer directly to a truck window; §22.2 and §22.7 mostly do not.

22.1 The sequence of service, step by step, and the judgment in each step

The sequence of service is the ordered set of steps a guest passes through from the door to the sidewalk, with a standard attached to each one. It is the most basic document in front-of-house operations and it is the one most independents never write down, which is why their service is inconsistent in ways they cannot diagnose.

Writing it down does three things. It makes training possible (Chapter 18 builds the program on top of this document). It makes a bad night diagnosable — when a table has a poor experience you can say which step failed rather than concluding that the server "had an off night." And, less obviously, it turns the guest's visit into a series of timed events, which means the visit has a duration you can measure and manage — which is §22.5, and which is where the money is.

Here is Bellwether's, with the standard for a two-top. Times are measured from the moment the guest is seated.

FIGURE 22.1 — The sequence of service, two-top standard        [the Bellwether plan — modeled]

  STEP  WHAT HAPPENS                    WHO           STANDARD   CLOCK
  ────────────────────────────────────────────────────────────────────
   0    Greeted at the door             host          ≤30 sec    −3:00
   1    Walked and seated               host                      0:00
   2    Water down, server greet        server/busser ≤2 min      0:02
   3    Beverage order taken            server        ≤4 min      0:04
   4    Beverages on the table          bar/server    ≤8 min      0:08
   5    Food order taken                server        ≤14 min     0:14
   6    Ticket sent; first course fires server/POS    ≤15 min     0:15
   7    First course on the table       runner        ≤26 min     0:26
   8    Entrées FIRED (server's call)   server        ≤30 min     0:30
   9    Entrées on the table            runner        ≤52 min     0:52
  10    Table touch                     manager        +3 min     0:55
  11    Entrée plates cleared           busser        ≤78 min     1:18
  12    Dessert / coffee decision       server        ≤80 min     1:20
  13    Check dropped                   server        ≤83 min     1:23
  14    Payment settled, guests up      server        ≤88 min     1:28
  15    Table reset and available       busser/host   ≤96 min     1:36
  ────────────────────────────────────────────────────────────────────
  DWELL (steps 1–14) = 88 minutes.  RESET = 8 minutes.
  TABLE CYCLE = 96 minutes.  This single number drives all of §22.5.

Two things in that table are load-bearing and easy to skim past.

Step 8 is the only step a computer cannot do. Everything else is a task with a clock on it. Firing the entrées is a judgment: the server looks at how much food is left on the appetizer plates, how the table is talking, whether the wine has arrived, and how far behind the kitchen is, and decides when the hearth should start cooking. Fire too early and the entrées sit under the heat lamp or land while the guests are still eating their first course. Fire too late and you have added dead minutes to the cycle for no reason. The gap between step 8 and step 9 is 22 minutes because that is roughly what the hearth chicken takes — so the server must fire while the appetizer plates are still on the table, not when they are cleared. Servers who wait for the clear are systematically running twenty-minute gaps between courses and do not know it.

Steps 13, 14, and 15 total nineteen minutes and contain zero hospitality. Hold that thought; §22.5 comes back for it with a calculator.

The judgment inside each step

A standard tells you when. It does not tell you what good looks like. Here is the judgment layer, step by step, compressed.

The greet (step 0). Thirty seconds is not about speed; it is about acknowledgment. A guest who is seen within thirty seconds will wait five minutes patiently. A guest who stands unacknowledged for ninety seconds has already decided something about you. If the host is on the phone, the rule is: make eye contact, hold up one finger, and finish the call fast. If there are three people on the phone and two parties at the door, the door wins — always — and the phone gets voicemail. The phone can be called back. The couple at the door cannot.

The seat (step 1). The walk to the table is a thirty-second commercial for the restaurant. The host walks at the guest's pace, not their own, and arrives at the table before the guest so they can pull the chair or gesture to the banquette. Menus go down open, not stacked. And the host says one sentence that hands the table off: "Your server tonight is right behind me." That sentence starts the clock on step 2 in the guest's head, and it is why a two-minute greet feels prompt instead of slow.

The greet at the table (step 2). Water and eye contact within two minutes. This is the single highest-leverage step in the whole sequence, because it is the one that makes every subsequent delay survivable. A table that has been greeted is a table that is being taken care of, and a table that believes it is being taken care of will forgive a twelve-minute wait for a cocktail. A table that has not been greeted at six minutes has begun to look around the room for someone to blame.

The beverage order (steps 3–4). Eight minutes from seat to drinks on the table is the standard, and it is the promise the bar has to keep. It also has a revenue dimension: a table that has a drink in hand at 0:08 orders a second one; a table that gets its first drink at 0:19 orders one. Chapter 15 priced what that costs the beverage line. The FOH mechanism is unglamorous — the bar prints service tickets separately from the kitchen, the well drinks and the wine go first, and the two-ingredient cocktails are built before the four-ingredient ones so the table is never waiting on one drink.

The order (step 5). Fourteen minutes is deliberately not eight. A table that is rushed to order has been told that the restaurant's convenience outranks theirs, and it produces the single most common complaint in casual full service. But fourteen is not thirty, either: the server who "gives them a few more minutes" three times has quietly added twelve minutes to the cycle and pushed the entrée fire into a worse window for the kitchen. The technique is to come back at 0:11 with a question rather than a demand — "Can I answer anything about the menu?" — which gets the table to a decision without pressure.

👨‍🍳 On the Line

What actually goes wrong in the first fifteen minutes.

Almost every service failure I have seen in a casual full-service room traces to one of four things, all of them in the first quarter of the cycle.

The double-seat. Two tables land in one section within ninety seconds. The server cannot be in two places, so table A gets greeted at 0:02 and table B at 0:05 — and then the server, trying to catch up, takes both beverage orders at once and rings them in together, which delays both. Both tables are now four to six minutes behind standard, permanently, because nothing later in the sequence recovers time. This is a host failure, not a server failure, and §22.3 gives it a rule.

The stranded drink. A four-top orders three cocktails and one glass of wine. The bartender builds the three cocktails and the wine is poured last, so all four drinks land at 0:11 instead of the wine landing at 0:05. Fix: wine and beer go down as they are ready. Nobody at a table minds the wine arriving first. Everybody minds waiting.

The un-fired ticket. The server takes the order at 0:14, gets pulled to another table, and rings it at 0:19. Five minutes vanished and no one can see them, because the POS timestamp starts when the ticket is sent, not when the order was taken. This is why ticket-time reports (Chapter 14) chronically understate the guest's real wait — the kitchen's clock and the guest's clock start at different moments, and the gap between them is invisible on every report either side of the pass will ever see.

The silent fire. The server fires the entrées and tells no one. The runner does not know a four-top is coming up in twenty minutes, the busser does not know to pre-bus, and the manager does not know to be near that table at 0:55. Firing is a broadcast, not a keystroke.

The entrée (steps 8–9). Covered above and in §22.6.

The table touch (step 10). A table touch is a brief, deliberate visit by a manager or the person in charge of the floor, made after the guests have had two or three bites of the entrée, for the express purpose of finding out whether the food and the experience are what the table wanted. It is not the server checking back — that is step 10a and it is the server's job. The manager's touch is different because the manager can fix things: comp, re-fire, move a table, buy a round, get the chef out. Chapter 23 owns what you do with what you learn. Here is the mechanic:

• Timing. Two to three minutes after the entrées land. Sooner and they haven't tasted it; later and they are halfway through and will say it's fine rather than send it back.
• Position. Beside the table, not looming over it, and not crouching, which reads as theatrical.
• The question. "How is everything?" produces "fine" one hundred percent of the time and is therefore worthless. Ask something answerable: "First time with the hearth chicken?" or "How's the temperature on that?" or, best of all, "Anything I can get better for you?" — which gives explicit permission to complain and is the only version that surfaces a problem while there is still time.
• Coverage. You cannot touch every table on a Friday. Touch every table on a Tuesday, every four-top and above on a Friday, and every table that the host flagged (first-timers, celebrations, a party the book shows has been here six times).

🤝 Hospitality

The first thirty seconds and the last ninety are the ones they'll describe to someone else.

There is a well-observed pattern in how people remember an experience: the beginning and the end carry disproportionate weight, and the middle compresses. Every operator who has read a review knows this empirically — the review talks about the greeting, the one thing that went wrong, and the goodbye, and rarely mentions the twenty-six minutes of the entrée that were the actual meal.

The operational consequence is not that the middle doesn't matter. It is that the cheapest minutes to improve are at the two ends, and almost every restaurant staffs them worst. The door is covered by the newest, lowest-paid person in the building. The goodbye is covered by nobody, because the server has already moved on to the next table and the host is on the phone.

At Bellwether the rule is that somebody says goodbye to every single table, by name where the book gives us one, and it is the manager's job when the host is buried. It costs nothing. It has no line on the P&L. And it is the last thing the guest experiences before deciding, without ever putting it into words, whether they are the kind of person who goes to Bellwether.

Chapter 23 makes the commercial case for the second visit properly. This chapter's contribution is narrower and more mechanical: put the goodbye in the sequence of service, give it an owner, and audit it. An unowned step does not happen.

22.2 The floor plan in motion: stations, rotation, and balancing a server's night

Chapter 7 drew a room. This section puts people in it.

Table management is the practice of deciding which party sits at which table, at which time, in whose station — and it is the single largest lever the front of house has over both revenue and the kitchen's workload. Everything in the rest of this chapter is a special case of it.

Start with the room as the plan actually specifies it: 17 tables and 56 dining seats, plus a 12-seat bar, 68 total. Eight two-tops, seven four-tops, two six-tops. 890 square feet of dining floor, which is 15.9 square feet per dining seat — comfortable for the category and a deliberate Chapter 7 choice, since density is a brand decision before it is a revenue decision.

FIGURE 22.2 — Bellwether's dining room: 17 tables in four sections   [the Bellwether plan]

        RIVERMILL AVENUE  ·  window wall
   ┌────────────────────────────────────────────────────────────────┐
   │   [T1]2   [T2]2   [T3]2   [T4]2                     ENTRY ▼    │
   │   ─────── SECTION 1 · 4 tables / 8 seats ───────   HOST STAND  │
   │                                                                │
   │   ╔══════ BANQUETTE RUN ══════╗        ┌───────────────────┐   │
   │   ║ [T5]2 [T6]2 [T7]4 [T8]4   ║        │  BAR · 12 seats   │   │
   │   ║ ── SECTION 2 · 4 / 12 ──  ║        │  ████████████     │   │
   │   ╚═══════════════════════════╝        └───────────────────┘   │
   │                                                                │
   │   [T9]4   [T10]4   [T11]4   [T12]4          SERVICE STATION    │
   │   ─────── SECTION 3 · 4 tables / 16 seats ───────              │
   │                                                                │
   │   [T13]4  [T14]2  [T15]2   [T16]6   [T17]6                     │
   │   ─────── SECTION 4 · 5 tables / 20 seats ───────              │
   │                                              ▲ THE PASS        │
   └────────────────────────────────────────────┬──────────┬────────┘
                                                │  HEARTH  │
                                                └──────────┘

   [Tn]s  =  table number and seat count.   Dining 56 + bar 12 = 68 seats.
   The 16-seat seasonal patio opens off the window wall and is not shown.
   T5–T13 push into a single 40-seat run for a private party — see §22.4.
   Not to scale; schematic only.

On the 40-top. Nine tables — T5 through T13 — push into one continuous run along the banquette. Individually those nine tables seat 32. As a single run they seat 40, because a continuous banquette seats at roughly two feet a person and you recover the seats that individual tables waste at their ends and joins. This is a genuine asset: it is the only way a 68-seat restaurant can sell a 40-person private party at all. It is also, as §22.4 will show in unpleasant detail, the most expensive thing in the building to sell badly, because deploying the run takes 32 of the 56 dining seats out of à la carte service for the entire prime window.

Cutting the sections

A station (Chapter 7 defined the term for the kitchen; on the floor it is used the same way) is the set of tables one server is responsible for. Station rotation is the practice of moving servers through different stations on a published cycle so that the good stations and the bad ones are shared.

Cut the room four ways and you get the sections in Figure 22.2. Notice what happens.

Four sections, and the largest is two and a half times the seats of the smallest. Every table has the same number of steps in the sequence of service, so a server's labor scales roughly with parties, but their income scales with covers and check average. This is not a hypothetical grievance. It is arithmetic, and the staff will do it in their heads by week three.

🧮 Run the Numbers

What an unrotated floor pays each server on a Friday.

A Friday on plan is 120 covers — 96 in the dining room and 24 at the bar (Bellwether models the dining room at about 80% of covers). Assume every section turns at the room's average of 2.18 seatings per table and that parties fit their tables the way the plan's party mix implies.

Section	Tables	Parties	Avg party	Covers	Sales @ \$46	Tips @ 18%
1 — window	4	8.3	2.0	16.5	\$759 | \$137
2 — banquette	4	8.3	2.5	20.7	\$952 | \$171
3 — center	4	8.3	3.0	24.8	\$1,141 | \$205
4 — hearth side	5	10.4	3.3	34.0	\$1,564 | \$282
Total	17	35.3	2.6	96	\$4,416** | **\$795

(Covers check: 96 × \$46 = \$4,416. Tip rate illustrative and constructed; Chapter 20 owns tip law and Chapter 19 owns wages.)

Section 4's server takes home more than double Section 1's, for the same five hours, on the same night, doing the same job. \$282 against \$137 is \$145 a shift. Over a hundred Friday-and-Saturday shifts a year that is roughly \$14,500 — which is a raise, or a resignation, depending on which side of the room you were standing on.

There is no version of this you can fix by cutting the sections differently. Sections cannot be made equal, because tables are not equal and the room is not symmetrical. The only fix is rotation.

Bellwether's rotation rule, which belongs in the training manual:

1. Sections rotate on a published cycle. Servers move one section per shift in a fixed order: 1 → 2 → 3 → 4 → 1. Over four shifts everybody has worked every station. The cycle is posted with the schedule so nobody has to ask, and nobody can trade their way into permanent ownership of Section 4.
2. Seniority buys shifts, not sections. The senior server gets the Friday; they do not get the hearth side every Friday. This distinction is the whole ballgame, and losing it is the most common way a floor's culture rots (Chapter 21).
3. The rotation survives a cut. When the room drops from four sections to three at 8:30, the absorbed tables go to the adjacent section, and the rotation resumes where it left off next shift.

Seating the rotation: covers, not tables

There is a second rotation, and it is the host's. Somebody has to decide which section gets the next party, and the naive answer — go around the room in order — is wrong.

Round-robin by table over-seats Section 1, which has four small tables and eight seats, and under-seats Section 4, which has five tables and twenty. By 8:00 the Section 1 server has been seated nine times and made \$700; the Section 4 server has been seated seven times and made \$1,400. Round-robin equalizes work and exaggerates income.

Rotate on covers. The host stand keeps a running cover count by section and seats the section with the fewest covers, subject to the table actually fitting the party. In practice:

   THE SEATING PAD, 7:00 p.m. Friday                [the Bellwether plan — modeled]

     SECTION      1      2      3      4
     covers      12     15     19     22     ← running count, seated tonight
     open       T2     —      T11    T15
     ────────────────────────────────────────────────────────────────
     next party: 4-top    → Section 1 is lowest, but has no four-top.
                            Seat T11 (Section 3). LOG THE SKIP.
     next party: 2-top    → Section 1. And the one after that, too,
                            until the cover count evens out.

Three rules govern the pad, and they are worth memorizing because they are the ones a host breaks under pressure:

• Fit first, balance second, log the skip. Never put a two-top at a six-top to keep the rotation clean. Seat the fit, record that Section 1 was skipped, and catch it up on the next two seatings.
• No section gets two tables inside four minutes. This is the double-seat rule from §22.1, and it is the most valuable single line in the host's training. Four minutes is roughly the time it takes a server to greet a table, take a beverage order, and ring it in. A host who honors it protects both the guest experience and the kitchen, because two tables greeted properly order at 0:14 and fire in sequence, while two tables greeted at once order at 0:20 and fire together — which drops two simultaneous entrée tickets on the hearth instead of two staggered ones.
• The section with the six-tops gets a lighter table count. Section 4 has five tables and Section 1 has four, and that looks backwards until you realize a six-top is not one and a half two-tops of work — it is closer to three. Chapter 19 builds the labor guide; the floor mechanic is simply that large tables buy their server relief elsewhere in the section.

How many sections, and when

Sections are not fixed; the room opens and closes them by volume. Bellwether's shape of the week gives five very different rooms.

(Turns = covers ÷ 68 seats. 62 + 78 + 92 + 120 + 123 = 475; 475 ÷ 5 = 95 covers a night; 95 ÷ 68 = 1.40. Headcount shown is the floor, not the payroll — Chapter 19 owns the hours and the dollars.)

Two operating notes on that table.

The "4 → 5" on Friday and Saturday means the room opens with four sections and splits Section 4 at 6:30, when the second wave of servers clocks in. The fifth section is carved out of the two six-tops and T13, which are the tables that hurt a server most when the room is full. Opening five sections at 5:00 would put a server on the floor for ninety minutes to serve nine covers, which Chapter 19 will tell you costs about a labor hour per cover.

Closing a section takes forty-five minutes. This is the mechanic managers get wrong. If you want a server gone at 9:00, you stop seating their section at 8:15, because their last table needs a full cycle. A manager who decides at 8:55 to cut somebody has just told a server to abandon a table that has not seen dessert, or has just kept a server on the clock until 10:20. The cut is a seating decision made forty-five minutes earlier, at the host stand, not a payroll decision made at the terminal.

⚠️ Where the Money Leaks

The dead section.

Here is a leak nobody ever writes down. On a Thursday at 92 covers you open four sections. Demand arrives unevenly — it always does — and by 7:30 Sections 2, 3, and 4 are full while Section 1 has one table. The host has been seating "in rotation," which they interpret as in order, and Section 1's two-tops don't fit the four-tops walking in.

Section 1 does 11 covers on the night instead of 20. That is 9 covers × \$46 = **\$414 of sales that did not happen in a room that had the seats for them.** The server who worked it made \$85 in tips. They will remember that number for a long time, and Chapter 21 will tell you what it costs when they leave.

The countermeasure is a table-mix discipline, not a rotation discipline. If four-tops are the constraint, the host stand should be combining — pushing T1 and T2 together into a four-top and seating it — rather than letting a section die. Bellwether's window two-tops are specified with matching heights precisely so they can be married. It takes forty seconds and it is the cheapest revenue in the building.

Chapter 24 prices the underlying problem — the cost of seating two people at a four-top — as a table mix question. The floor-level fix is: a host who cannot combine tables is leaving money on them.

22.3 The host stand: quoting, seating, pacing, and protecting the kitchen

Now the central claim of the chapter, stated plainly: the host stand controls more revenue than any other position in the building, and in a restaurant with a constrained cooking platform it is a production-scheduling function that happens to greet people.

That is not a compliment paid to hosts. It is a description of what the job does. The host decides:

• Who sits where — which determines whether a party of two consumes two seats or four, and therefore how many of the room's 56 dining seats are actually earning.
• When they sit — which determines the shape of the demand curve arriving at a hearth that clears a fixed number of covers an hour.
• Whether a walk-in stays or leaves — which determines a chunk of Friday's cover count.
• What the kitchen is asked to do in any given fifteen minutes — which determines ticket times, which determines whether the room's best night is also its worst reviewed.

Chapter 14 handed us the constraint. Let us translate it into the unit the floor actually uses.

Translating the kitchen's ceiling into covers

Chapter 14 measured the hearth at 28 items an hour, sustained. "Items" is the kitchen's unit. The floor thinks in covers. Convert:

At Bellwether the entrée attachment is 0.90 — nine entrées for every ten covers — and essentially every hot entrée on the menu touches the fire, which is why the hood upgrade in Chapters 6 and 7 cost what it did. The hearth additionally fires the roasted shares and the flatbreads that guests order without an entrée. Call it 0.97 hearth items per cover.

$$\text{Covers per hour} = \frac{28 \text{ items/hr}}{0.97 \text{ items/cover}} = 28.9 \approx 29 \text{ covers per hour}$$

The fire is under real load from about 5:30 p.m. to 10:30 p.m. — half an hour after the doors open until half an hour after the last seating — which is five hours.

$$29 \text{ covers/hr} \times 5 \text{ hr} = 145 \approx \textbf{144 covers a night}$$

That reconciles exactly with Chapter 14's figure, and it is worth pausing on what the reconciliation means: 144 covers consumes the entire fire window at one hundred percent duty cycle with nothing left over. It is not a target. It is the number you would get if demand arrived perfectly flat — 29 covers an hour, every hour, for five hours, never 30 and never 28.

Demand does not arrive flat. Demand arrives in a lump between seven and eight.

The arrival shape

The arrival shape is the distribution of seatings across the service window. It is the thing the host stand actually controls, and almost nobody measures it.

Here is a Friday on plan — 120 covers, well under the 144 ceiling, a night that on paper the kitchen should stroll through.

FIGURE 22.3 — Friday's arrival shape against the fire's rate   [the Bellwether plan — modeled]

  SEATINGS BY HALF HOUR (covers)
   5:00  ████████ 8
   5:30  ██████████ 10
   6:00  ██████████████ 14
   6:30  ████████████████ 16
   7:00  ██████████████████ 18
   7:30  ██████████████████ 18
   8:00  ████████████████ 16
   8:30  ████████████ 12
   9:00  ████████ 8                                    total = 120

  FIRE LOAD BY HOUR — covers whose entrées land in each window.
  The fire's peak trails the door's peak by about half an hour.

   40 ┤
   36 ┤                              ▓▓▓▓▓▓▓▓▓
   32 ┤                              ▓▓▓▓▓▓▓▓▓
   29 ┤═══════════════▓▓▓▓▓▓▓▓▓══════▓▓▓▓▓▓▓▓▓══════════════ CAPACITY 29
   28 ┤               ▓▓▓▓▓▓▓▓▓      ▓▓▓▓▓▓▓▓▓   ▓▓▓▓▓▓▓▓▓
   20 ┤   ▓▓▓▓▓▓▓▓▓   ▓▓▓▓▓▓▓▓▓      ▓▓▓▓▓▓▓▓▓   ▓▓▓▓▓▓▓▓▓
   12 ┤   ▓▓▓▓▓▓▓▓▓   ▓▓▓▓▓▓▓▓▓      ▓▓▓▓▓▓▓▓▓   ▓▓▓▓▓▓▓▓▓
    8 ┤   ▓▓▓▓▓▓▓▓▓   ▓▓▓▓▓▓▓▓▓      ▓▓▓▓▓▓▓▓▓   ▓▓▓▓▓▓▓▓▓  ▓▓▓▓▓
    0 ┼───┴───────────┴──────────────┴───────────┴──────────┴────────
       5:30–6:30    6:30–7:30      7:30–8:30   8:30–9:30  9:30–10:30
          18           30             36          28          8   = 120
       −11 spare    +1 OVER       +7 OVER      −1 spare    clears

Read the bottom line. On a 120-cover Friday — twenty-four covers below the kitchen's ceiling — the 7:30 fire hour is asked for 36 covers against a capacity of 29.

Seven covers over. The hearth's minute is $60 \div 29 = 2.07$ minutes per cover, so seven covers of overflow is about fourteen minutes of backlog sitting at the pass at 8:30. Every table seated between 7:00 and 7:45 gets its entrées fourteen minutes late. A 22-minute fire becomes a 36-minute fire, and the guest, who cannot see the hearth, concludes the kitchen is slow.

And here is the part that a rate calculation hides: the 8:30 hour has one cover of slack. One. The backlog does not clear during service at any meaningful rate. It rides all the way to 9:30, when arrivals collapse and the fire finally gets ahead. A queue does not catch up during a rush. It catches up after one. That is what Chapter 14 meant by handing us the sentence about the hour being the wrong unit, and it is the single most useful thing on this page.

The peak-to-average ratio, and the real ceiling

Generalize it. Define the peak-to-average ratio as the busiest fire hour divided by the night's average fire hour. The Friday above runs 36 ÷ 24 = 1.50. That is a normal unmanaged neighborhood dinner curve. A well-managed book — one where the host stand actively steers early and late — can get it to about 1.33, and no further, because you cannot flatten demand you do not have. Nobody wants a 5:15 reservation on a Friday just because it would help your hearth.

(Backlog minutes = covers over capacity × 2.07. All figures modeled from the Bellwether plan.)

Four things fall out of that table, and they are the chapter's core findings.

1. A 95-cover night — the plan's average — never queues at all, under either shape. The plan's average is genuinely comfortable. That is worth saying, because most of this chapter sounds alarming.
2. A 120-cover Friday necessarily queues. There is no arrival shape available to a real restaurant that fits 120 covers inside a 29-cover-an-hour fire without a peak-hour overload. The only question is whether the overload is six minutes or fourteen.
3. The host stand is worth roughly eight to eleven minutes of peak ticket time, every Friday and Saturday, on identical covers and identical revenue. That is the value of the position, expressed in the only unit the guest experiences.
4. 144 is not reachable. At the kitchen's own ceiling, even a well-managed book runs nineteen minutes behind at the peak. The honest managed ceiling is about 132 covers — twelve minutes of peak backlog, which is survivable — and §22.5 is going to arrive at 132 again by a completely different route.

🧾 Read the Numbers

```text FIGURE 22.4 — "The book at four o'clock, Friday" [the Bellwether plan — modeled, pre-opening] THE ARTIFACT The reservation book as the host reads it at 4:00 p.m., before pre-shift. Friday dinner, second week of October. Reserved covers by quarter hour, walk-in hold, and the resulting fire load. THE CONTEXT Bellwether, 68 seats, 17 dining tables. Modeled on-plan Friday: 120 covers forecast, 78 reserved (65%), 42 held for walk-ins. The hearth clears 29 covers an hour. Kitchen is fully staffed; no 86s; weather clear.

  TIME     RESERVED   HELD    SEATED    RUNNING   FIRE WINDOW   LOAD  vs 29
            COVERS   WALK-IN   TOTAL     TOTAL
  5:00         4        4        8          8
  5:30         6        4       10         18     5:30–6:30      18   −11
  6:00         9        5       14         32
  6:30        11        5       16         48     6:30–7:30      30   +1
  7:00        13        5       18         66
  7:30        13        5       18         84     7:30–8:30      36   +7  ◄──
  8:00        11        5       16        100
  8:30         7        5       12        112     8:30–9:30      28   −1
  9:00         4        4        8        120     9:30–10:30      8   clears
  ─────────────────────────────────────────────────────────────────────────
  TOTAL       78       42      120        120                   120

WHAT IT SHOWS A book that looks healthy and is not. Covers total 120 against a kitchen ceiling of 144, so at a glance there is 24 covers of headroom. There isn't. Thirty-six covers of fire land in the 7:30 hour against 29 of capacity — seven over, about 14 minutes of backlog at the pass by 8:30, and only one cover an hour of slack afterward to work it off. The tables seated between 7:00 and 7:45 will wait roughly 36 minutes for entrées instead of 22. WHAT IT DOESN'T It does not show party sizes, so it cannot tell you whether the 7:30 block is nine two-tops or four four-tops and a six — which changes the table availability completely. It does not show which tables are held, so it cannot tell you whether the 42 walk-in covers can physically be seated. It does not show how long anybody actually stays. And it says nothing about no-shows, which Chapter 24 policies. THE DECISION Before service: move six reserved covers out of the 7:00 and 7:30 rows — call them today and offer 6:15 or 8:45 with a glass of wine on arrival — and cap walk-in seating at 8 covers per quarter hour from 6:45 to 8:00. Same 120 covers, peak fire hour drops from 36 to about 32, backlog from 14 minutes to 6. THE LESSON A reservation book is not a list of guests. It is a production schedule for a station with a fixed hourly rate, and the total at the bottom is the least informative number on it. ```

Quoting

Quoting is telling a walk-in party how long they will wait, and it is a forecast, not a courtesy. The honest method has three inputs:

$$\text{Quote} = (\text{time until a fitting table clears}) + (\text{reset}) - (\text{time already elapsed on that table})$$

In practice the host does not run that formula; they read the floor. A two-top seated at 7:10 whose entrées went down at 7:45 will settle and stand up around 8:35 — because the sequence of service in §22.1 says entrée-to-departure is about 36 minutes. Add eight minutes of reset. So at 7:55, a two-top walking in gets quoted 35 to 45 minutes, and you say the top of the range.

That last clause is the whole craft. Quote long, seat early.

• A party quoted 45 minutes and seated in 35 is delighted. You beat your promise. They start dinner in a good mood, which is worth more than the ten minutes cost you.
• A party quoted 30 minutes and seated in 40 is angry, and the anger compounds, because they spent ten minutes past the promise watching the door and rehearsing a grievance.
• The asymmetry is enormous and it runs one direction only. There is no such thing as being punished for beating a quote.

Three failure modes to train against:

Quoting from the book instead of the floor. The book says a four-top is reserved for 8:00, so the host quotes 8:00. But the four-top currently sitting at T10 ordered dessert at 7:40 and is not leaving until 8:20. The book is a plan; the floor is a fact. Quote from the floor.

Quote drift. A soft-hearted host quotes short because they do not want people to leave. The result is not fewer walkaways — it is the same number of walkaways forty minutes later, plus twenty people standing in a 890-square-foot dining room's entry being visibly unhappy in front of every table by the window. Twenty angry people in your foyer is a worse outcome than ten people who went somewhere else.

Not logging it. If you do not record quoted-versus-actual for every wait, your quoting never improves, because nobody knows whether it is any good. Bellwether's standard: 85% of parties seated at or before the quoted time, and the median actual within five minutes under the quote. That is checkable on a Monday, it costs one column on the waitlist screen, and it is the only quality measure the host stand has.

🧮 Run the Numbers

What the host stand is worth, in dollars, over a year.

Take the two levers this chapter establishes and apply them to the two nights that matter.

Lever one — flatten the arrival shape. Move roughly six covers out of the peak hour on Friday and Saturday, by offering earlier and later times when the book is being built and by capping the walk-in seating rate during the peak. Peak-to-average goes from 1.50 to about 1.33. Peak backlog falls from 14 minutes to 6.

Lever two — take eleven minutes off the back of the table cycle (this is §22.5's finding; take it on credit for a moment). Table capacity rises by about four parties a night.

Together, and only if the demand exists to fill the shape you have created:

	Plan	Managed	Δ
Friday covers	120	132	+12
Saturday covers	123	135	+12
Week (Tue–Sat)	475	499	+24
Seat turns	1.40	1.47	+0.07
Annual dinner covers	24,700	25,948	+1,248
Annual dinner sales @ \$46 | \$1,136,200	\$1,193,608 | **+\$57,408**

Now the honesty. That \$57,408 is conditional and the condition is large. It exists only if there are twelve more parties on a Friday who wanted a table and could not get one. On a Tuesday at 62 covers, both levers are worth exactly zero — you cannot turn a table faster than the next guest arrives, and there is no next guest.

So state it correctly: when demand exceeds the book, the host stand converts into covers. When demand equals the book, the same work converts into ticket time — which is not revenue this week but is the mechanism of the second visit, and Chapter 23 will tell you what that is worth. Both are real. Neither should be double-counted.

⚠️ Where the Money Leaks

The wall of tickets, and the manager who caused it.

The most expensive twenty minutes in a restaurant's week is the twenty minutes when a manager, seeing people waiting and open tables, seats four tables in six minutes.

Follow it through. Four tables — say eleven covers — all greeted between 7:32 and 7:38. All order between 7:46 and 7:52. All fire between 8:00 and 8:06. Eleven covers of hearth demand land inside six minutes on a station that clears one cover every two minutes and seven seconds. The eleventh cover is twenty-three minutes behind the first, and every ticket already in the queue sits behind all eleven.

Those four tables produced 11 × \$46 = **\$506** of sales. They cost roughly twenty minutes of ticket time across every open table in the room — call it fifteen tables and forty covers whose dinner just got slower. The manager was trying to be efficient. They spent forty guests' experience to buy eleven guests' revenue, and the trade is almost never worth it.

The countermeasure is a number, posted at the host stand: eight covers per quarter hour. At Bellwether that is 32 covers an hour, which is exactly the managed peak the arithmetic supports. Three tables per quarter hour, no more than one party of five or more per quarter hour, and no section double-seated inside four minutes. A host who is holding a table with people standing in front of them is not being unhelpful. They are doing the most technically demanding part of the job.

🔍 Check Your Understanding

1. Bellwether's hearth clears 29 covers an hour. A Saturday is booked at 123 covers with an unmanaged peak-to-average ratio of 1.50 over a five-hour fire window. What is the peak fire hour, how many covers over capacity is it, and roughly how many minutes of backlog does that produce?
2. Why does the fire's peak hour trail the door's peak by about half an hour, and what does that imply about when a host's seating decision becomes visible at the pass?
3. A host quotes a walk-in 25 minutes and seats them in 32. A second host quotes 40 and seats them in
4. Both parties waited the same amount of time. Explain, in commercial terms, why the outcomes are not the same.

(1: average fire hour = 123 ÷ 5 = 24.6; peak = 24.6 × 1.50 = 36.9, call it 37; that is 8 over 29; 8 × 2.07 = about 17 minutes of backlog. 2: because a table is seated, greeted, and ordered before anything fires — roughly 30 minutes in the sequence of service — so a seating decision made at 6:35 lands on the hearth at about 7:05. 3: the first party's expectation was broken by seven minutes and they spent those seven minutes watching the door; the second party's expectation was beaten by eight minutes and they started dinner feeling well treated. Identical cost to the restaurant, opposite effect on the second visit.)

22.4 Reservations, walk-ins, waitlists, and the mix that fits your concept

The reservation book — whether it is a leather-bound page or a screen — is the restaurant's record of committed future demand: who is coming, when, how many, and anything you know about them. It is also, as §22.3 established, a production schedule.

The question every independent has to answer is what fraction of the room to sell in advance.

The three postures

All-reservation. Every seat is booked. You know your covers by 4:00 p.m., you can order and staff precisely, and your arrival shape is entirely under your control. The costs are real: you are exposed to no-shows on inventory you can no longer sell, you cannot serve the neighborhood's Tuesday spontaneity, and you have made your restaurant into a place people plan for rather than a place they go to. For a destination tasting-menu room that is exactly right. For a neighborhood restaurant it is usually a slow strangulation.

All walk-in. No book at all. This is the posture of a lot of very good restaurants, and its advantages are underrated: no no-shows, no platform fees, no phone, and a line outside that is itself marketing. Its cost is that you have surrendered the arrival shape. You cannot flatten a curve you cannot see, so on a busy night you get a 7:15 wall and there is nothing the host can do about it except quote long and hold the door.

Hybrid. Book part of the room, hold part for walk-ins. Almost every successful neighborhood restaurant lands here, and the interesting question is the ratio.

Bellwether's answer:

(267 + 208 = 475. Modeled from the Bellwether plan.)

The logic behind the split, which you should be able to argue for your own concept:

Book more on the nights you need the shape. Friday and Saturday are the nights the fire is constrained, and reservations are the only instrument that lets a host stand move demand before it arrives. On Tuesday, with 62 covers spread over five hours, shape is irrelevant — the fire is never asked for more than about 15 covers an hour — so the book's job is smaller.

Hold enough that the neighborhood can still get in. This is the part operators talk themselves out of. A restaurant that is fully booked three weeks out on a Friday has stopped being a neighborhood restaurant. The people who live four blocks away and decide at 6:40 that they don't want to cook can never get a table — and those people are the second, fifth, and twentieth visits that Chapter 23 will show you are where profitability actually lives. Selling every seat in advance is a way of trading your most valuable guests for your least loyal ones.

Walk-ins fill the shoulders the book cannot sell. Almost nobody calls to reserve 5:00 or 8:45. Walk-ins take both, cheerfully, because they are choosing between your bar at 8:45 and going home. Chapter 24 works the shoulder-hour economics properly; the FOH observation is simply that the 5:00 and 9:00 rows in Figure 22.4 are 8 covers each and both are more than half walk-in.

The bar never takes reservations. Twelve seats, first-come, always. This is a deliberate structural choice and it does three jobs at once, which is why it earns its own callout below.

The hold-back, and how to release it

Holding inventory for walk-ins only works if the hold is a rule rather than a hope. Bellwether's:

• Two-tops T1–T4 and T14–T15 are never released to the book more than 24 hours out. These are the six tables a walk-in couple is most likely to fit, and they are the ones the book will otherwise consume first.
• The 5:00–5:45 and 8:45–9:00 rows are open to the book at any time, because the book almost never wants them and a reserved shoulder cover is a gift.
• The six-tops T16 and T17 take reservations only. A six-top walk-in on a Friday is a fifty-minute quote you will not beat, and holding two six-tops for a walk-in that may not come is expensive inventory.
• The bar is never held and never released. It is simply not in the book.

👨‍🍳 On the Line

Working a waitlist, and why the bar is worth more than its pour cost.

The waitlist is the ordered queue of walk-in parties, with the quote attached. It is a working document, not a list, and there are four fields that matter: name, party size, quote given, and time quoted. If the fourth one is missing you cannot audit anything.

Now the twelve bar seats, which do three jobs.

One: they are the holding pen. A party quoted 35 minutes who is standing in the entry will leave at minute 20. The same party sitting at the bar with a drink will wait 45 and be pleasant about it. This is not a trick; it is the difference between waiting and being somewhere.

Two: the wait produces revenue. On a Friday, of the 42 walk-in covers, perhaps 25 wait an average of 22 minutes. If a bit over half of them buy a drink at around \$13, that is roughly 25 × 0.55 × \$13 ≈ **\$179 a night — call it \$18,000 a year across five nights, and every dollar of it is at bar margin. (Conversion and drink price are illustrative and constructed; measure your own.)

Three, and largest: they are inventory that never no-shows. Twelve seats that cannot be reserved cannot be reserved-and-abandoned. On the plan's numbers the bar carries 95 covers a week — about 20% of dinner — at 1.58 turns, and it does it with no book, no deposit, and no phone call.

The failure mode to train against: a bar full of waiting parties is a bar that cannot seat drinkers. On a Saturday at 8:00 the host must know which bar seats are waitlist and which are guests, because giving a two-top a bar seat and then asking them to move to a table in fifteen minutes is a good outcome, and letting a party of four camp on four bar stools for an hour waiting for T11 is not. The rule: waitlist parties get told, out loud, "we'll move you the moment your table is ready." They almost always say yes, and now the seat is honest.

The honest wait

One more technique, and it is the most useful sentence in this section.

There will be nights when you have an open table and a backed-up kitchen. The wall of tickets from §22.3 is twenty-five minutes deep and T7 just cleared. Every instinct says seat it.

Tell the truth instead:

"I can seat you right now, but I want to be straight with you — the kitchen is running about twenty-five minutes behind, so your entrées would be slow. Or I can sit you at 8:15 and you'll eat on time. Which would you rather?"

Roughly half will take the table anyway and will be fine, because they were told. The other half will take 8:15 and be grateful. What you have done is convert an unmanaged twenty-five-minute problem into a chosen twenty-five-minute wait, and a chosen wait is a completely different experience from an imposed one. You have also just moved covers out of the peak fire hour without anyone feeling refused.

This is the honest version of what the industry sometimes does dishonestly — quoting a wait with tables visibly open in order to pace the kitchen. Pacing the kitchen that way is correct. Doing it silently, while a guest stares at an empty table, is how a restaurant earns a reputation for arrogance. Say the reason out loud and the same decision becomes hospitality.

The 40-top, and what it actually costs the room

Now the hard case, and it is the frozen Friday from Chapter 14: 142 covers on the books including a 40-top private party at 6:30.

Start with the physical fact from Figure 22.2. The 40-top deploys the banquette run — T5 through T13, nine tables, 32 of the room's 56 dining seats. For the entire prime window, à la carte service has:

• Eight tables — T1, T2, T3, T4, T14, T15 (six two-tops) and T16, T17 (two six-tops)
• 24 dining seats, plus the 12-seat bar

That is the room. Eight tables and 36 seats to serve 102 à la carte covers, on the busiest night of the week, alongside a party of 40.

🧮 Run the Numbers

The frozen Friday, seat by seat.

What the room can physically do that night.

The eight remaining tables will turn about 2.5 times each in a four-hour seating window at Bellwether's cycle times, which is 20 parties. At an average party of 2.6, that is 52 covers. The bar at 2.0 turns is 24 covers. The party is 40.

$$52 + 24 + 40 = 116 \text{ covers}$$

The book says 142. The night was oversold by 26 covers before anyone walked in the door.

What saves it. The party is contracted to a set menu and a stated end time, so it eats faster than à la carte: seated 6:15, done by 8:30. The run breaks back down into tables and is available again at 9:00, which buys one late seating — 9 parties, about 23 covers.

$$52 + 24 + 40 + 23 = 139 \approx \textbf{138 covers}$$

Which is exactly where Chapter 14's frozen Friday landed. The night did not "come in a little under the book." The book was never seatable, and the 9:00 reseat of the banquette run is the only reason it got as close as it did.

Which makes the party's departure time the most valuable number in the reservation book. If the party lingers to 9:45 — and parties linger unless the event order says otherwise — you lose all 23 of those late covers:

$$23 \times \$46 = \textbf{\$1,058}$$

One line in a banquet event order, worth a thousand dollars. Chapter 29 writes the event order and prices the party; this chapter's contribution is telling you which line to fight for.

And that is only the seats. The fire is worse, and §22.6 takes it apart.

22.5 Turn time: measuring it, influencing it, and the line between efficient and rushed

Turn time is how long a table is occupied by one party, from seated to departed. Its close cousin, the table cycle, adds the reset — the minutes between one party leaving and the next being seatable. The cycle, not the dwell, is what determines how many parties a table produces in a night.

Cover count is simply the number of guests served in a period, and every metric in this section is denominated in it.

Almost nobody measures any of this properly, which is why almost nobody can answer the only question that matters: given my room, how many covers can I actually produce?

Measuring it by segment

An average turn time is nearly useless. "Our tables turn in about an hour forty" tells you nothing you can act on. Break it into segments and the picture changes completely.

FIGURE 22.5 — Where the 96 minutes go, two-top    [the Bellwether plan — modeled]

  SEGMENT                          STANDARD   MODELED ACTUAL   VARIANCE
  ──────────────────────────────────────────────────────────────────────
  A  Seated → beverage order          4 min         5 min        +1
  B  Beverage order → food order     10 min        11 min        +1
  C  Food order → first course       12 min        12 min         —
  D  First course → entrée           26 min        27 min        +1
  E  Entrée → plates cleared         26 min        26 min         —
  F  Cleared → dessert decision       2 min         3 min        +1
  G  Dessert decision → check drop    3 min         5 min        +2
  H  Check dropped → settled          5 min        13 min        +8   ◄──
  ──────────────────────────────────────────────────────────────────────
     DWELL (A–H)                     88 min       102 min       +14
  I  Departed → table reset           8 min        11 min        +3   ◄──
  ──────────────────────────────────────────────────────────────────────
     TABLE CYCLE                     96 min       113 min       +17

  Segments A–F (82 min) are DINNER. The guest is eating, drinking, talking.
  Segments G–I (14 min standard / 19 modeled) are LOGISTICS. Nobody is
  enjoying them, and eleven of those minutes are recoverable.

That figure is the most important thing in the section. Eleven of the seventeen minutes of variance sit in segments H and I — the check settle and the reset — and not one of them is dinner.

This is the line between efficient and rushed, and it is a bright one:

• Compressing segments A through F is rushing. Taking the order at 0:08 instead of 0:14, clearing plates before everyone is finished, dropping the check with the entrée, hovering. Every one of these produces the same complaint — they wanted us out — and every one of them costs you the second visit that Chapter 23 will show is the entire business model.
• Compressing segments G, H, and I is service. A guest who has asked for the check and is waiting thirteen minutes for someone to run a card is not enjoying an extended dinner. They are being ignored, in the most memorable ninety seconds of the visit. Cutting that to five minutes makes their night better and yours longer.

🧾 Read the Numbers

```text FIGURE 22.6 — "The turn-time report, a modeled week" [the Bellwether plan — modeled, pre-opening] THE ARTIFACT Table-dwell report, dinner service, one full week (Tue–Sat), taken from the reservation platform's seat/clear timestamps joined to POS check open/close. 17 dining tables, 146 seatings, 380 dining-room covers. THE CONTEXT Bellwether on plan. 475 dinner covers for the week; the dining room carries about 80% of them and the bar the other 95. Party mix as forecast: an average dining party of 2.6.

  TABLE TYPE      TABLES  SEATINGS  AVG PARTY  COVERS   DWELL  RESET  CYCLE
  Two-top            8       78        2.0       156    102     11     113
  Four-top           7       53        3.0       159    113     12     125
  Six-top            2       15        4.7        70    133     14     147
  ───────────────────────────────────────────────────────────────────────
  DINING ROOM       17      146        2.6       385           weighted 121
  Bar (12 seats)     —       —          —         95            —       —

  TURNS         seat turns, dining room   380 ÷ 56  =  1.36 per night
                seat turns, bar            95 ÷ 12  =  1.58 per night
                seat turns, blended       475 ÷ 68  =  1.40 per night
                TABLE turns, dining room  146 ÷ 85  =  1.72 per table-night

WHAT IT SHOWS Three different "turns" for the same restaurant, and the plan quotes the least useful one. Blended seat turns of 1.40 is the business plan's number. The dining room alone runs 1.36. The tables run 1.72 — which is the figure a manager can act on, because tables are what you seat. It also shows the room's real cycle: 113 minutes for a two-top against a 96-minute standard. WHAT IT DOESN'T It does not distinguish a table that sat empty because nobody wanted it from one that sat empty because it wasn't reset — the two look identical in the data and have completely different fixes. It does not capture the minutes between a guest ordering and the server ringing it in. It says nothing about party-to-table fit: 146 parties occupied 456 seat-slots to seat 385 covers, which is 84% utilization even when every table is full. And it cannot see the covers that walked out during a quote. THE DECISION Attack segments H and I only. Put payment on a handheld at the table so a settle is five minutes rather than thirteen, and give the reset an owner and a clock — a busser assigned to reset, not "whoever's closest." Do not touch segments A through F. Re-run the report in four weeks against a 96-minute two-top standard. THE LESSON Average turn time is a number you can quote. Turn time by segment is a number you can manage. The eleven minutes worth taking are all after the guest has stopped eating. ```

(Covers reconcile: 156 + 159 + 70 = 385 against 380 forecast — the two-tenths of rounding in average party size. Seat-slots: 78 × 2 + 53 × 4 + 15 × 6 = 156 + 212 + 90 = 458; 385 ÷ 458 = 84%.)

What the eleven minutes are worth

Now compute it, because a recommendation without a dollar figure is a preference.

🧮 Run the Numbers

Eleven minutes, priced.

The seating window is 5:00 to 9:00 — 240 minutes. Seatings per table = 240 ÷ cycle.

Table type	Tables	Cycle now	Seatings	Cycle at standard	Seatings	Δ parties	Avg party	Δ covers
Two-top	8	113	2.12	96	2.50	+3.0	2.0	+6.0
Four-top	7	125	1.92	111	2.16	+1.7	3.0	+5.1
Six-top	2	147	1.63	133	1.80	+0.3	4.7	+1.4
	17		35.0		40.0	+5.0		+12.5

About twelve covers a night — and every one of them comes from the eleven minutes after the guest asked for the check, not from a single minute of their dinner.

At \$46: **\$575 a night.** But only on nights where someone is waiting. Applied to Friday and Saturday only, 52 weeks:

$$\$575 \times 2 \times 52 = \textbf{\$59,800 a year}$$

Round it down to \$45,000–\$55,000 to allow for the nights when the twelfth cover does not materialize, and note the enormous assumption buried in it: there is a party waiting for every one of those tables. On a Tuesday, this entire calculation is worth zero dollars.

Is 1.40 turns defensible? — the three routes to 132

The business plan has carried 1.40 seat turns since Chapter 1. Ninety-five covers a night on 68 seats. This is the chapter that either defends it or revises it, so let us test it properly.

Route one: the queue. §22.3 established that the hearth clears 29 covers an hour and that a realistically well-managed book runs a peak-to-average ratio of about 1.33. At twelve minutes of peak backlog — the most a room can absorb without the guest noticing — the arithmetic supports about 132 covers.

Route two: the floor plan, corrected for fit. Chapter 7's ceiling is about 151 covers. Decompose it: the 12-seat bar takes singles and pairs and fits almost perfectly, contributing perhaps 28 covers at 2.3 turns; the 56 dining seats supply the other 123 at about 2.2 turns. But Chapter 7's figure assumes perfect party-to-table fit, and Figure 22.6 measured the plan's own fit at 84%. Apply it where it belongs — to the dining room, not the bar:

$$(123 \times 0.84) + 28 = 103 + 28 = \textbf{131 covers}$$

Route three: the table cycle. At the 96-minute standard, the 17 tables produce 40 parties a night (the table above), which at a 2.6 average party is 104 dining covers. The bar at 2.5 turns adds 30. Total: 134 covers.

Three routes — a queue, a floor plan, and a stopwatch — none of which knows about the others, and they land at 131, 132, and 134.

FIGURE 22.7 — Three ceilings and the one that's real        [the Bellwether plan — modeled]

   151  ├──────────────────────────────────────────────  Ch. 7 floor plan
        │                                                (perfect fit, rectangular demand)
   144  ├─────────────────────────────────────────       Ch. 14 kitchen
        │                                                (28 items/hr × 5 hr, flat)
        │
   132  ├═══════════════════════════════════════         THE OPERATING CEILING
        │                                                queue 132 · fit 131 · cycle 134
   123  ├────────────────────────────────                plan Saturday   (1.81 turns)
   120  ├───────────────────────────────                 plan Friday     (1.76 turns)
        │
    95  ├───────────────────────                         PLAN AVERAGE    (1.40 turns)
        │
    62  ├─────────────                                   plan Tuesday    (0.91 turns)
        └──────────────────────────────────────────────

   The two published ceilings are both true and neither is reachable, because
   both assume demand arrives in a shape no dining room has ever seen.

So: is 1.40 defensible?

Yes — and it is conservative. The average is 95 covers against an operating ceiling of about 132. Nothing in the room or the kitchen prevents 1.40; the constraint on an average night is demand, not capacity, and the plan is right not to bet year one on a green team hitting a managed peak.

But it is the wrong number to manage with, for two specific reasons.

First, 1.40 hides where the covers are. Friday and Saturday are 120 + 123 = 243 covers — 51% of the week on two of the five nights. On those two nights the room runs 1.76 and 1.81 turns, and the hearth's peak hour is already asked for 36 and 37 covers against a sustained 29. The plan's average is comfortable and the plan's Friday is not, and no average will ever tell you that.

Second, 1.40 hides that Friday is simultaneously at two different limits. The 120-cover Friday needs 96 dining covers, which needs 37 parties. At the cycle times the restaurant will actually run in year one — 113 minutes for a two-top — 17 tables produce about 35 parties. Friday as planned is two parties short of physically possible, and seven covers an hour over the fire's rate. The eleven minutes in segments H and I is not an optimization. On the plan's own numbers it is the difference between hitting Friday and missing it.

The revision, then, is not to the number. It is to the unit:

Keep 1.40 as the forecast. Replace it as an operating target with a peak-hour cover cap of 32 and a quarter-hour cap of 8, and measure the book against those. A restaurant that hits 1.40 while running 36-cover fire hours has hit its forecast badly. A restaurant that runs 32-cover fire hours will find 1.47 waiting for it.

🤝 Hospitality

The line, stated so a nineteen-year-old server can hold it.

Everything in this section is a revenue argument, and revenue arguments are exactly how restaurants talk themselves into rushing people. So here is the rule, in the form it has to take to survive contact with a Saturday night:

You may take minutes from the restaurant. You may not take minutes from the guest.

The reset, the check run, the card, the pre-bus, the table that sat dirty for eleven minutes because nobody owned it — those are ours, and every one we recover is a minute the guest never wanted. The second glass of wine, the dessert conversation, the ten minutes after the plates are gone when they are just talking — those are theirs. They paid for them. In a business whose product is ninety minutes of somebody's evening, the minutes are the product.

There is one honest exception, and it belongs in the training manual too: when a table is holding a room that has people waiting, you may ask — once, warmly, and only through a manager. "Can I move you to the bar for another round? I'd love to give this table to a couple who've been waiting." Offered with a drink and offered once, most people say yes and feel like insiders. Offered twice, or offered by a server who has been told to turn the table, it is the thing they will tell their friends about, and it will not be a compliment.

22.6 Ticket flow: firing, coursing, and the FOH/BOH handshake

The ticket is the only document the front and the back of the house both read. Everything else — the book, the seating pad, the prep list, the all-day board — belongs to one side. How tickets are written, when they are sent, and when they are fired is the handshake, and it is where most FOH/BOH conflict actually originates (Chapter 21 handles the conflict; this section handles the cause).

The three verbs

SEND. The server rings the order. The ticket prints or lands on the kitchen display, and the kitchen's clock starts. Everything ordered is now visible to the line and counts on the all-day board.

FIRE. The entrées go on. At Bellwether, appetizers fire on send and entrées fire on the server's call — the server hits "fire" at the terminal, and the expediter calls it. This is the coursing model for a room where a table's pace is set by the table, not by the kitchen.

PICK UP. Plates leave the pass. The runner takes them. This is the number every ticket-time report measures, and per §22.1 it starts too late to describe the guest's actual wait.

The alternative — fire on send, where the entire ticket goes on at once — is what quick service and some high-volume casual rooms do, and it is genuinely faster. It also means the table's pace is the kitchen's pace, and a table that lingers over appetizers gets entrées under a heat lamp. For an 890-square-foot dining room with a 22-minute hearth entrée, fire-on-call is correct. Know which model you are running and make sure both sides of the pass know it is the same one.

Coursing, and the judgment in the fire

Coursing is deciding what arrives when. Three rules cover most of it:

• Fire against the cook time, not against the plates. The hearth chicken is 22 minutes. If the server waits for the appetizer plates to be cleared, the table sits for twenty-two silent minutes. The entrées fire when the appetizers are most of the way eaten — roughly when the table is two-thirds through — so the entrées land two or three minutes after the clear.
• Course the table, not the ticket. A four-top where two people ordered appetizers and two did not is a coursing problem, not an ordering problem. Either the two appetizers become a shared first course for the table, or all four entrées fire on send. The unacceptable answer is two people eating while two people watch.
• Write the exceptions on the ticket, not in your head. "Fire 2 & 4 together," "hold the fish, allergy ticket," "birthday, 8:30." A verbal instruction to a line cook at 7:40 on a Friday has a survival rate approaching zero, and Chapter 25 will tell you what an allergy communicated verbally is worth in liability.

👨‍🍳 On the Line

The pass at 7:40, and the two most expensive habits on the floor.

Habit one: sitting on the fire. A server has a table that is talking, not eating, and does not want the entrées to land early — so they hold the fire ten minutes. Entirely reasonable, table by table. But five servers doing it simultaneously moves fifteen covers of hearth demand out of the 7:30 window and into the 8:00 window. Sometimes that helps enormously. Sometimes it takes an 8:00 hour that was at 28 and puts it at 43. Nobody is deciding this. Five people are each making a local decision and the sum is being discovered at the pass twenty minutes later.

The fix is not to forbid it. The fix is that the expediter can see pending fires — the tickets that are sent but not fired — and the manager's walk (§22.8) includes a glance at that count. If nine tables are sitting on fires at 7:35, the room is about to hand the hearth a wall, and the host stand needs to stop seating for ten minutes now, not at 8:00 when the ticket times show it.

Habit two: the on-the-fly. A table has been forgotten, a plate came back wrong, a re-fire is needed. "On the fly" means jump the queue. And on a night with twenty-six covers of backlog, an on-the-fly does not create a plate — it takes one from whoever was next. Every on-the-fly is a decision that some other table waits two more minutes.

That is often the right call. A re-fire on a wrong temperature absolutely jumps the line; the guest is already eating nothing while their party eats. But it should be called, out loud, by a manager or the expediter — never inserted quietly by a server at a terminal, because the person who inserts it cannot see the queue it is jumping.

The queue, proven: why wave-splitting creates nothing

Now back to the frozen Friday, because it is the cleanest possible demonstration of what a queue is.

142 covers on the books, a 40-top at 6:30, 138 covers landed. Chapter 14 found from the pass that splitting the party into waves creates zero capacity and merely moves the delay onto the à la carte queue. Here is the same fact from the floor, in arithmetic.

The party is coursed and its entrées fire as a block. Priority is not in question: the party is contracted and the à la carte tables are not, so the party fires first in every window. That assumption is the whole result.

FIGURE 22.8 — The frozen Friday as a queue         [the Bellwether plan — modeled]

  Hearth capacity 29 covers/hour.  Party fires first in every window.

  SCENARIO A — party fired as one block at 7:00
  FIRE WINDOW    NEW À LA   NEW    CAP   FIRED             À LA CARTE
                   CARTE   PARTY                            BACKLOG
  5:30–6:30         14        —     29   14                     0
  6:30–7:30         20       40     29   29 party              20   (+11 party)
  7:30–8:30         24        —     29   11 party + 18 àlc     26   ◄ peak
  8:30–9:30         17        —     29   29 àlc                14
  9:30–10:30        23        —     29   29 àlc                 8
  10:30–11:00        —        —     14    8 àlc                 0
  ─────────────────────────────────────────────────────────────────
  totals            98       40          138 covers fired

  SCENARIO B — party split into two waves of 20 (7:00 and 7:40)
  5:30–6:30         14        —     29   14                     0
  6:30–7:30         20       20     29   20 party +  9 àlc     11
  7:30–8:30         24       20     29   20 party +  9 àlc     26   ◄ peak
  8:30–9:30         17        —     29   29 àlc                14
  9:30–10:30        23        —     29   29 àlc                 8
  10:30–11:00        —        —     14    8 àlc                 0
  ─────────────────────────────────────────────────────────────────
  totals            98       40          138 covers fired

  PEAK À LA CARTE BACKLOG:  Scenario A = 26 covers.  Scenario B = 26 covers.
  LAST PLATE OFF THE PASS:  Scenario A ≈ 10:47.      Scenario B ≈ 10:47.
  26 covers × 2.07 min = 54 MINUTES, either way.

Identical. Not similar — identical, to the cover. Splitting the party changed which minute each guest's food appeared and changed nothing whatsoever about the queue's length, because the fire's rate did not change and the number of items did not change. You cannot reorder your way out of a capacity problem. Chapter 14 said it from the pass; here it is with the arithmetic underneath it.

Note also the sting in the tail: the à la carte backlog is still 14 covers deep at 9:30, which means the 23 guests who got the banquette run's late reseat are waiting behind a party that ate at seven. A queue does not respect when you arrived.

What actually creates capacity

Exactly three things, and only one of them belongs to the front of house.

1. Take items off the hearth — a party menu whose entrée comes out of the oven rather than off the fire. This is Chapter 14's lever and Chapter 29's menu, and it is the single most effective one: 40 covers moved to a different station removes the entire problem.
2. Move the party's fire out of the peak — seat and fire at 6:00 rather than 7:00, when à la carte demand is 14 covers instead of 20. Worth something; not worth much, because 40 covers in one block overruns a 29-cover hour no matter which hour you choose.
3. Take the covers off the à la carte book — the host stand's lever, and the one this chapter owns.

🧮 Run the Numbers

What twenty covers you refuse to sell are actually worth.

Scenario A (as booked): 98 à la carte covers alongside the party. Peak à la carte backlog 26 covers = 54 minutes. The tables seated between 6:15 and 7:30 wait roughly fifty-five minutes for entrées on the busiest night of the week.

Scenario C (the host protects the window): at the moment the party's deposit is taken, twenty à la carte covers come off the 6:00–7:30 rows of the book. Total covers 118 rather than 138.

Fire window	New àlc	New party	Cap	Fired	À la carte backlog
5:30–6:30	14	—	29	14	0
6:30–7:30	8	40	29	29 party	8 (+11 party)
7:30–8:30	12	—	29	11 party + 18 àlc	2
8:30–9:30	21	—	29	23 àlc	0
9:30–10:30	23	—	29	23 àlc	0

Peak à la carte backlog: 8 covers = 17 minutes, against 54.

The trade, stated exactly: twenty covers × \$46 = **\$920 of revenue deliberately not sold, in exchange for removing thirty-seven minutes** of the worst ticket time of the week from roughly forty guests, and getting the room out clean by 10:00 instead of 10:47.

I am not going to give you a number for how many of those forty guests would not have come back — nobody has one worth quoting, and inventing one would be exactly the kind of false precision this book refuses. What I will tell you is that Chapter 23 is about to demonstrate that the second visit is where the profit lives, and that \$920 is a cheap premium against forty simultaneous bad nights.

And the rule that falls out of it, which is the most useful sentence in this chapter:

When you sell a 40-top at 6:30, you are also selling forty covers of pacing inventory out of the 6:30–7:45 window. Take them off the à la carte book at the moment you take the deposit, or the kitchen will take them off for you at the pass.

Forty covers at eight per quarter hour is seventy-five minutes of the book. Chapter 29 prices the party and will tell you it earns more per head than à la carte — that is why you take it. This chapter's job is to make sure the seventy-five minutes appears in the decision.

The handshake, scheduled

Three fixed moments where the two houses talk. Put them on the clock or they will happen only when something has already gone wrong.

The 7:45 conversation is the one that gets skipped, and it is the one that matters. It is thirty seconds long, it happens at the pass, and it consists of exactly two questions: "Where are you?" and "What do you want me to do with the door?" A kitchen that is asked that question at 7:45 will tell you the truth. A kitchen that is not asked will simply fall further behind while the host keeps seating.

22.7 The check, payment, and the last ninety seconds

The check is the itemized bill presented to a guest, and it is the last document your restaurant hands to somebody who is about to decide whether to come back. It is treated, almost universally, as an afterthought.

Accuracy first

An incorrect check undoes an entire evening. It is the one error the guest is certain of — they know what they ordered — and it converts a warm room into an argument at exactly the moment memory is being formed. The mechanics that prevent it are dull and non-negotiable:

• The server reads the check before it goes down. Every time. Items, seat numbers, modifiers, anything comped or voided, and the tax line.
• Comps and voids are on the check as comps and voids, not silently deleted, because Chapter 34 is going to need that audit trail and Chapter 31 is going to need it to explain your discount line.
• Split checks are decided at the beginning, not the end. "Is this together or separate?" is a question for step 5 in the sequence of service, and asking it then converts a four-minute end-of-meal operation into a zero-minute one. Every POS handles seat-level ordering; almost nobody trains it.

Timing the drop

Too early reads as eviction; too late reads as neglect, and both are common. Bellwether's rule:

The check goes down on the second decline. The first "no dessert, thank you" is not a signal — people change their minds about coffee. The second decline, or a phone coming out, or coats being gathered, is the signal. Then the check goes down within three minutes, presented with a sentence that removes the pressure: "No rush at all — whenever you're ready."

That sentence matters more than it looks. A check on the table with "no rush" attached is permission. A check on the table in silence is a message.

The settle, and where eight minutes live

Figure 22.5 found that check-dropped-to-settled runs thirteen minutes against a five-minute standard — the single largest recoverable segment in the entire table cycle. It is thirteen minutes because of a specific, fixable mechanic: the server takes the folder to a fixed terminal, waits behind another server, processes, walks back, waits for a signature, and returns. Four trips.

Payment at the table on a handheld collapses it to one. That is a technology decision — Chapter 26 owns the stack, the contracts, and the processing fees — but the FOH consequence is the twelve covers a night computed in §22.5, and it belongs in the specification when the POS is chosen, not discovered afterward.

Two things not to do while chasing those minutes:

• Do not pre-authorize, pre-drop, or hover. A server standing three feet away holding a card reader is worse than thirteen minutes of waiting.
• Do not make the tip prompt the guest's problem. Whatever your tipping model, the screen or the slip should be legible, the default should not be adversarial, and any service charge must be disclosed clearly and in advance. Chapter 20 owns the law here, and it varies enormously.

⚖️ Code and Compliance

Three front-of-house compliance points that live at the check and the door.

Payment disclosure. If you add a service charge, an automatic gratuity on large parties, a credit card surcharge, or a kitchen-appreciation fee, the rules on how and where it must be disclosed vary by state and sometimes by city, and several jurisdictions have specific menu-disclosure requirements. Card-network rules also constrain surcharging independently of state law. Disclose it on the menu, on the check, and out loud for large parties, and have an attorney look at the wording before it goes to print. Chapter 20 covers the gratuity-versus-service-charge distinction, which has large wage-law consequences.

Card data. The front of house handles payment instruments all night. Do not write full card numbers on paper, do not store them in the reservation record for anything but a policy the guest agreed to, and do not let a phone photograph of a card become a normal way of taking a deposit. Your processor's requirements are contractual, not optional.

The inspector at the host stand. Chapter 13 and Chapter 25 walk the surprise inspection from the kitchen's side — the walk-in at 46°F, the raw chicken over the greens, the sanitizer bucket at 50 ppm. But the inspector comes in through your door, and the host is who meets them. Train it explicitly:

• Greet them, ask for identification, and get a manager immediately — do not walk them anywhere, do not answer operational questions, and do not "give the kitchen a heads up," which in some jurisdictions can itself become a finding and in all of them looks exactly like what it is.
• The front of house has its own inspection surface and it is routinely forgotten: the bar's hand sink (accessible, stocked, not used as a dump sink), ice bins and scoops (scoop stored outside the ice, never a glass), the glass washer's sanitizer or final rinse temperature, bar garnish handled with tongs or gloves as ready-to-eat food, the server station's reach-in holding at or below 41°F, and sanitizer buckets on the floor at the correct concentration — typically 200–400 ppm for quat, verified with test strips, not by eye.
• All of this varies by jurisdiction and changes. Verify locally, keep the logs, and treat the front of house as part of the food-safety system rather than as the room where the food is delivered.

The last ninety seconds

The guests stand up. What happens next is free, and almost nobody owns it.

• Somebody says goodbye, by name where the book gives you one. Assigned: the manager if the host is buried, and the server always says something even if the manager gets the door.
• Someone gets the door if their hands are full. This is a ten-second act that people remember for years, which tells you something about how rarely it is done.
• Nothing is being cleared while they are still standing there. A busser stripping the table while a party puts their coats on is the single clearest possible statement that the restaurant is finished with them. Reset begins when they are past the host stand, not before.

Chapter 23 makes the argument for why this is worth doing. The operational point here is narrower: put it in the sequence of service as step 14a, give it a name and an owner, and audit it on the manager's walk. Steps with owners happen. Steps without owners are aspirations.

22.8 Running a shift: the manager's walk, the touch, and the save

Everything above is a system. This section is about the person holding it together for five hours.

Before the doors

The front-of-house pre-shift is not the same meeting as the kitchen's, though at a 68-seat restaurant they may happen back to back. Chapter 21 owns the pre-shift as a cultural instrument; the FOH content is a checklist:

1. The number. Tonight's forecast covers and, more importantly, the shape — "one-twenty tonight, and the seven-o'clock hour is heavy, so we're capping at eight covers a quarter."
2. The book. Large parties and their times. Anything flagged: first visits, celebrations, regulars, allergies, a guest who had a bad night last month.
3. The 86s and the counts. What is gone and what is limited, in numbers — "nine chickens left" is actionable; "we're low on chicken" is not.
4. The sections and the rotation. Who is where, when the fifth section opens, and the projected cut order.
5. One thing. A single service point being worked on this week — the four-minute rule, the second decline, the goodbye. One. A pre-shift that lists nine priorities has none.

The manager's walk

The most valuable habit a floor manager can build is a circuit on a clock. Not wandering — a fixed route, roughly every twenty minutes, looking at specific things in a specific order.

FIGURE 22.9 — The manager's walk, every 20 minutes         [the Bellwether plan]

   ┌─► HOST STAND ──► BAR ──► DINING ROOM ──► THE PASS ──► BACK ─┐
   │                                                             │
   └─────────────────────────────────────────────────────────────┘

   HOST STAND    Covers seated this quarter hour vs the cap of 8.
                 Waitlist length and the oldest quote outstanding.
                 The next 30 minutes of the book — what's coming.

   BAR           Drinks on the rail waiting to be run. Waitlist parties
                 seated and whether they know they'll be moved. Bartender's
                 hands — if they're washing glassware, the bar is behind.

   DINING ROOM   Walk the sections in rotation order. Look for: water
                 levels, un-bussed plates, a table with no drinks, a table
                 whose entrées have been down 3 minutes (that's a touch),
                 and any table where nobody is talking. Silence is data.

   THE PASS      Current ticket time. PENDING FIRES — how many tickets are
                 sent but not fired. Anything about to 86. The expediter's
                 face, which is a faster instrument than any report.

   BACK          Restrooms, and then the one question: "what do you want me
                 to do with the door for the next thirty minutes?"

Five stops, three or four minutes, twelve to fifteen times a service. What makes it work is that it is the same route every time, which means deviations become visible. A manager who walks randomly sees whatever is loudest. A manager who walks a circuit notices that the bar rail has had drinks on it at three consecutive stops, which is a runner problem forty minutes before it becomes a complaint.

The save

A save is catching a table that is going badly while there is still time to change the outcome. Chapter 23 owns service recovery — what you do after something has gone wrong, and what a comp is actually worth. This chapter owns the part before that: detection.

The signals, in rough order of how early they appear:

The last row is the one that connects this section back to the whole chapter. A rising ticket time is not a kitchen problem to be solved in the kitchen. It is a door problem, and the door is the only place it can be solved. By the time the pass is behind, the kitchen has no lever at all — the fire clears 29 covers an hour whatever anybody does. The manager's job at that moment is to walk to the host stand and take covers out of the next thirty minutes.

🔍 Check Your Understanding

1. It is 7:50 on a Friday. Ticket times have moved from 21 minutes to 29 minutes in the last twenty minutes, and eleven tables are sitting on un-fired entrées. Name the two things you do in the next sixty seconds, and say which one is the kitchen's and which is yours.
2. Why does closing a server's section require a decision forty-five minutes before that server leaves? What goes wrong if the decision is made at the moment of the cut?
3. A manager touches every table on a Friday and the room's ticket times get worse. Explain a plausible mechanism.

(1: neither is the kitchen's — the kitchen has no lever. Cap or stop seating at the host stand for the next twenty minutes, and get the pending-fire count down by telling servers to fire in sequence rather than all at once. The kitchen's only contribution is telling you the true ticket time. 2: because the last table in that section needs a full 96-to-113-minute cycle; stopping seating at 8:15 is what makes a 9:00 cut possible. Deciding at 8:55 forces you either to abandon a table mid-meal or to keep the server until 10:20. 3: the manager is standing in the dining room instead of at the host stand and the pass, so nobody is managing the door — and the door is the only control on ticket times. On a Friday, touch the four-tops and above and the flagged tables, and spend the recovered time on the circuit.)

🍽️ The Business Plan

Checkpoint 22 of 40 — the Service & FOH Operations section.

This chapter contributes the operating specification for the room: how a guest is served, how the tables are divided, how the book is built, and what the plan's seat-turn assumption actually rests on.

1. The sequence of service. Fifteen steps with a standard time for each (Figure 22.1), producing an 88-minute dwell and a 96-minute table cycle for a two-top. The document is the training spine (Chapter 18) and the audit instrument for every service failure.

2. Table management and the floor in motion. Seventeen tables in four sections (Figure 22.2), a published rotation of 1 → 2 → 3 → 4 so that no server owns the twenty-seat station, and a host seating rotation that balances on covers, not tables — fit first, balance second, log the skip. Sections open and close by volume, from three on a Tuesday to five on a Friday, and a section is closed to seating forty-five minutes before its server is cut.

3. The host stand's operating rules. These are the plan's new numbers:

4. The book: 56% reserved, 44% walk-in. 267 reserved and 208 walk-in covers a week; 65/35 on Friday and Saturday, 45/55 early in the week. Six two-tops held out of the book beyond 24 hours, both six-tops reservation-only, the twelve bar seats never in the book at all.

5. Turn time, by segment. A 96-minute standard cycle against a modeled 113-minute actual, with the entire recoverable gap in the check settle and the reset — eleven minutes, none of them dinner, worth about twelve covers on a night when someone is waiting.

6. The 1.40-turn assumption — defended, and re-expressed. This is the checkpoint's real work.

1.40 seat turns is defensible and conservative as a forecast. Ninety-five covers a night sits well below an operating ceiling of about 132, and three independent methods — a queue analysis at the hearth's 29-cover hour, Chapter 7's floor plan corrected for the plan's own 84% party-to-table fit, and the table cycle at standard — converge on 131, 132, and 134. Nothing in the room or the kitchen prevents 1.40. The constraint on an average night is demand.

But 1.40 is the wrong unit to operate with, and the plan should say so. Two of the five nights carry 51% of the week's covers (120 + 123 = 243 of 475) and run 1.76 and 1.81 turns. On those two nights the hearth's peak hour is asked for 36 and 37 covers against a sustained 29, and the room needs 37 and 38 table seatings against a year-one capacity of about 35. Friday as planned is simultaneously over the fire's rate and at the table cycle's limit, and the average conceals both.

The plan therefore keeps 1.40 as the forecast and adds a peak-hour cover cap of 32 and a quarter-hour cap of 8 as the operating targets. Managed to those caps — and with the eleven minutes recovered from the check settle and the reset — the same room supports 132 on Friday and 135 on Saturday, a 499-cover week, 1.47 turns, and \$1,193,608 of annual dinner sales against the plan's \$1,136,200: an upside of \$57,408, conditional entirely on the demand existing to fill it.

Reconciling to the top line. Dinner at plan is 24,700 covers × \$46 = **\$1,136,200; brunch is 11,440 × \$24 = **\$274,560; together \$1,410,760** of the year-one forecast of **\$1,550,000. The remaining \$139,240 does not come from this room on these nights — it sits in the 16-seat patio in season and the private-event line. This chapter cannot produce it. Chapter 24 and Chapter 29 have to.

What this checkpoint does not settle:

• RevPASH, yield management, table mix, and the cost of seating two at a four-top. Figure 22.6 measured 84% party-to-table fit and left the money on the table deliberately. Chapter 24.
• No-show and deposit policy. This chapter models an unclaimed 7:00 four-top as \$184 of lost inventory and stops there. Chapter 24.
• Hospitality, service recovery, reviews, and what a comp is worth. §22.8 detects; it does not repair. Chapter 23.
• The reservation platform and payment hardware, and what the stack costs as a percentage of sales. The twelve covers a night in §22.5 assume payment at the table. Chapter 26.
• What the floor costs. Five servers, two bartenders, two hosts, and three bussers on a Friday is a labor structure, not a labor number. Chapter 19.
• The event itself — the 40-top's menu, per-head price, deposit, and end time. This chapter only insists that the end time is in the contract. Chapter 29.

Open questions carried forward:

1. Does Rivermill District demand actually support 132 covers on a Friday, or is the plan's 120 the ceiling of the neighborhood rather than the ceiling of the room? (Chapters 24, 27)
2. Can a new floor team hold a 96-minute cycle in year one, or is 113 minutes what a green room realistically runs for six months? (Chapters 18, 19)
3. What does the 84% party-to-table fit cost in dollars, and does the answer change the table mix? (Chapter 24)
4. If the hearth is the ceiling on two nights a week, is there a menu answer — a non-hearth entrée that sells at the peak — and what does it do to the mix? (Chapters 12, 24)
5. How many large parties a year can this room absorb before the à la carte guest notices? (Chapter 29)

Conclusion

The room is not a container. It is a production line with a fixed hourly rate at the hot end, a fixed number of workstations in the middle, and a demand curve arriving at the front that nobody designed and everybody has to absorb.

That reframing is what this chapter has been for. Chapter 7 gave us a floor plan that holds 151. Chapter 14 gave us a fire that clears 144. Both numbers are true and neither is reachable, because both assume demand arrives in a rectangle. Real demand arrives in a lump between seven and eight, and once you take the lump seriously — through a queue, through party-to-table fit, or through a stopwatch on the table cycle — the room's honest capacity is about 132 covers, and all three routes agree.

The plan's 1.40 turns survives, comfortably, as a forecast. What does not survive is the idea that an average tells you anything about a Friday. Half the week's covers land on two nights, and on those two nights the room is at the table cycle's limit and over the fire's rate simultaneously. The countermeasures are unglamorous and cheap: eight covers a quarter hour, no section double-seated inside four minutes, a quote you beat, eleven minutes taken from the check settle and the reset and not one minute taken from anybody's dinner, and a manager who walks the same five stops every twenty minutes and asks the pass what to do with the door.

None of that is hospitality. It is the scaffolding hospitality stands on — because a guest whose entrées arrive thirty-six minutes late has had a bad night no matter how warmly anyone spoke to them, and a room that hits its cover count by wrecking its ticket times has hit its forecast badly.

Which is exactly where Chapter 23 begins. We have made the room work. We have said nothing yet about how it feels, what to do when something goes wrong anyway, what a comp actually costs, what a review is worth, and why the second visit — not the first — is where a restaurant's profit has been hiding this whole time.

Key Terms

Table management — the practice of deciding which party sits at which table, at which time, in whose station; the front of house's largest single lever over both revenue and the kitchen's workload. (Ch. 22)

Station rotation — moving servers through different sections on a published cycle so that unequal stations produce equal income over time. Sections cannot be made equal; rotation is the only remedy. (Ch. 22)

Turn time — how long a table is occupied by one party, from seated to departed. Distinct from the table cycle, which adds the reset and is what actually determines how many parties a table produces. (Ch. 22)

Pacing — controlling the rate at which covers are seated so that demand on a constrained station stays inside its sustained hourly rate; expressed operationally as a cover cap per quarter hour. (Ch. 22)

The reservation book — the record of committed future demand: who is coming, when, how many, and what you know about them. Functionally a production schedule for the kitchen. (Ch. 22)

Waitlist and quoting — the ordered queue of walk-in parties, and the forecast of how long each will wait. The governing rule is quote long, seat early; the quality measure is quoted-versus-actual. (Ch. 22)

Cover count — the number of guests served in a period. The denominator of nearly every operating metric in the front of house. (Ch. 22)

The check — the itemized bill presented to the guest; the last document the restaurant hands to someone deciding whether to return, and the largest recoverable segment of the table cycle. (Ch. 22)

The sequence of service — the ordered set of steps a guest passes through from door to sidewalk, with a standard time attached to each; the training spine and the diagnostic instrument for service failure. (Ch. 22)

Table touch — a brief, deliberate visit by a manager two to three minutes after the entrées land, made with a question the guest can actually answer, for the purpose of finding a problem while there is still time to fix it. (Ch. 22)

Arrival shape (working term) — the distribution of seatings across the service window, and the thing a host stand actually controls; summarized by the peak-to-average ratio of fire load per hour. (Ch. 22)

Fire window (working term) — the hours during which a constrained cooking station is under real load; at Bellwether, 5:30 to 10:30, five hours, 29 covers an hour. (Ch. 22)

Spaced Review

1. Without looking back: the hearth clears 29 covers an hour and the fire window is five hours. Why is 144 covers not a target?
2. A restaurant reports 1.40 seat turns and a manager reports 1.72 table turns for the same nights. Both are correct. Explain the difference and say which one you would put in a business plan and which one you would put on the host stand.
3. From Chapter 14: the frozen Friday's grill cook no-shows at 3:40. Nothing in this chapter's arithmetic changes — the hearth still clears 29 covers an hour. What does change, and where in this chapter's rules would you make the adjustment?
4. From Chapter 1: prime cost is COGS plus labor. Name two decisions the host stand makes on a Friday night that move the labor half of prime cost, and one that moves the COGS half.
5. The recurring question: an operator wants to raise turns by taking twelve minutes out of the table cycle. Their consultant proposes clearing entrée plates as each guest finishes rather than when the table finishes. Using Figure 22.5, say which segment that touches, whether it is on the right side of the efficient/rushed line, and what you would do instead to find the same twelve minutes.