Chapter 17: Hiring: Finding Cooks, Servers, Bartenders, and Hosts in an Industry with 75% Annual Turnover

"I don't have a labor cost problem. I have a hiring problem that shows up in the labor line." — constructed; a general manager, three months into a bad year

Overview

Here is a sentence that should bother you more than it does: in most American restaurants, the single largest recurring expense after food and wages is replacing the people who just left, and it appears on no line of the profit-and-loss statement.

Chapter 14 ran a Friday night at Bellwether where the grill cook did not come in. The night cost the restaurant \$116.70 in hard money — and, because the absent cook's unworked shift was worth more than the overtime it caused, the restaurant technically came out ahead on wages. That is what a labor report can see. What it cannot see is that the sous chef worked fifteen hours, eleven tables waited past the service standard, and a sauté cook stood alone for four hours. It also cannot see the several thousand dollars it will cost to replace that grill cook, because that money is scattered across a dozen accounts under names like "advertising," "training," and "management salary."

This chapter puts a number on it. Not an industry number — your number, for a specific position, in a specific restaurant. Because until turnover is a figure an operator can state out loud, nobody funds the thing that prevents it, and the entire discipline of hiring stays a chore squeezed between prep and service rather than the highest-return management activity in the building.

The industry runs annual turnover somewhere around 75%. Treat that figure with the skepticism this book has trained you to have — it moves enormously by position, market, and whether the count includes seasonal separations, and it is reported by parties who benefit from it sounding alarming. But no serious source puts restaurant turnover anywhere near normal for the wider economy, and any operator who has run a schedule knows why. What matters is not the national figure. What matters is that you can compute your own, price it, and decide how much preventing it is worth.

Then we build the process: descriptions that attract competence, sourcing that produces candidates, interviews that actually predict, a working interview that does not create legal exposure, and an onboarding plan for the first thirty days — which is when most of the damage is done.

In this chapter, you will learn to:

• Compute the fully loaded cost of one turnover, including the costs hidden in other people's salaries.
• Write a job description that a competent candidate would answer, and diagnose why most postings don't get one.
• Evaluate sourcing channels by cost per hire and by how long the resulting hires stay.
• Run a structured interview with scored questions, and explain why the unstructured version predicts so little.
• Run a stage or working interview fairly, safely, and without creating wage-and-hour exposure.
• Sequence the offer, Form I-9, and a thirty-day onboarding plan built around the moments people actually quit.

Learning Paths

🏗️ Opening — §17.8's Staffing Plan is a required section of your business plan, and §17.1's arithmetic is what justifies the money you will spend on retention. Read all of it. 📋 Managing — this is your chapter. If you run someone else's restaurant, hiring is the lever you control most directly and the one most likely to be judged badly. Weight §17.1, §17.4, and §17.8. 🍸 Beverage — bartenders have the longest tenure and the highest replacement cost of any hourly position in most restaurants; §17.1's position-by-position table shows why, and §17.4 covers the skills test. 🚚 Small Format — a truck or a pop-up may run on two or three people, which makes every single hire an enormous fraction of the operation. Your turnover cost per event is higher, not lower.

17.1 What turnover actually costs, computed

Start with the arithmetic, because everything else in this chapter is an argument about how much prevention is worth, and you cannot have that argument without a number.

Turnover rate is the number of separations in a period divided by the average number of positions in that period, usually annualized. A restaurant with 30 positions that sees 23 people leave in a year has a turnover rate of about 77%. That figure alone tells you very little — it does not distinguish a dishwasher who left in week two from a sous chef who left after four years — which is why the useful work is done by the cost of turnover, computed per position.

Building the number

Most operators, asked what a departure costs, name the recruiting fee and stop. The recruiting fee is the smallest component. Here is the full build for a line cook at Bellwether, at illustrative but realistic figures.

🧮 Run the Numbers

What one line cook actually costs to replace.

Component	How it arises	Cost
Advertising and job-board spend	one posting cycle, two boards	\$150
Management screening and interviewing	5 hours at \$32/hr fully loaded | \$160
The stage	4 paid hours at \$22/hr fully loaded | \$88
Orientation and paperwork	3 hrs new-hire at \$22 + 1 hr manager at \$32	\$98
Unproductive training wages	40 hrs at \$22, of which ~50% produces nothing yet | \$440
Trainer's diverted attention	20 hrs of a senior cook's time at \$26, half-lost | \$260
Reduced output, days 1–90	slower tickets, more re-fires, more waste	\$400
Learning-curve waste	over-portioning, burned proteins, wrong preps	\$250
Overtime covering the vacancy	3 weeks × 10 hrs × \$11 premium | \$330
Total		\$2,176

Call it \$2,180**. Now notice where it lives on the P&L: about \$150 of it is in "advertising." Roughly \$550 is inside management salary, which is a fixed cost nobody re-examines. Around \$1,100 is inside the hourly labor line, indistinguishable from productive wages. And \$650 — the waste and the reduced output — is inside food cost, where it will be blamed on portioning.

Not one dollar of it appears on a report called "turnover." That is the whole problem.

The restaurant-wide figure

One position is not the business. Bellwether will employ about thirty-one people across eleven positions, and they do not turn over at the same rate or cost the same to replace. Dishwashers turn fastest and cost least; a sous chef turns rarely and costs a fortune.

FIGURE 17.1 — BELLWETHER'S ANNUAL TURNOVER LEDGER        [constructed teaching example]

  position          heads   turnover   separations   cost each      annual cost
  ─────────────────────────────────────────────────────────────────────────────────
  Line cook            4       100%         4.0       $2,180         $ 8,720
  Server               8        75%         6.0       $1,500         $ 9,000
  Busser / runner      4       125%         5.0       $  700         $ 3,500
  Dishwasher           3       150%         4.5       $  700         $ 3,150
  Bartender            3        50%         1.5       $2,400         $ 3,600
  Host                 3       100%         3.0       $  900         $ 2,700
  Prep cook            2       100%         2.0       $1,400         $ 2,800
  Sous chef            1        50%         0.5       $5,200         $ 2,600
  Assistant manager    1        33%         0.33      $6,000         $ 2,000
  ─────────────────────────────────────────────────────────────────────────────────
  TOTAL               29                   26.8                      $38,070

  (The two owner-partners are excluded; they do not turn over, though Chapter 39
   discusses what happens when one of them wants to.)

  $38,070 is 2.5% of the plan's $1,550,000 of revenue
             7.6% of the plan's $500,000 labor line
             and more than half the year's entire debt service.

Read the total again: \$38,070 a year, on a business whose plan projects about \$261,000 of operating profit before debt service. Turnover is quietly consuming roughly fifteen cents of every dollar of that profit, and it does so invisibly, in nine places at once.

That is the argument for everything else in this chapter and for all of Chapter 21. It is also the argument a manager needs when asking for money to raise a starting wage, buy a training program, or schedule an extra hour of overlap. Not "people are unhappy." "We are spending thirty-eight thousand dollars a year on this, and here is the build."

⚠️ Where the Money Leaks

The two costs nobody counts, and one of them is larger than everything in the table.

The cost of vacancy. Between the separation and the replacement's first productive shift, the work still happens — it is absorbed by overtime, by managers stepping onto the line, and by a team running short. At Bellwether, a line-cook vacancy averaging three weeks costs about \$330 in overtime premium alone. But Chapter 14 showed what happens when the kitchen runs a body short on a busy Friday: eleven tables past the service standard. Some of those guests do not come back, and Chapter 23 will show that a second visit is where profitability actually lives.

The cost of hiring the wrong person and keeping them. This one has no line at all. A cook who is adequate but not good costs you nothing measurable and perhaps two points of food cost in waste, a half-minute per ticket, and the slow attrition of the good cooks who are tired of covering. Operators tolerate it because firing feels expensive and hiring feels harder. The tolerance is usually the more expensive choice, and §17.4 exists to make it a less frequent one.

17.2 Writing a job description someone competent would answer

Most restaurant job postings are written in ten minutes by a tired manager, and they read like it. They are also the first thing a candidate learns about how the restaurant is run, which makes them a hiring document and a brand document — Chapter 3's argument about consistency applies to a job board as much as to a menu.

A job description is the document that states what a position does, what it requires, what it pays, and who it reports to. Written well, it screens candidates before they apply, which is the cheapest screening available.

What a bad posting looks like, and why it fails

The generic restaurant posting says: Line cook wanted. Must have 2 years experience. Fast-paced environment. Team player. Competitive pay. Apply within.

Every clause in that posting does work against the restaurant:

• "2 years experience" is a proxy for a skill you have not named. Two years at a chain doing assembly is not two years on a wood fire. Say the skill.
• "Fast-paced environment" means nothing; every restaurant is fast-paced. To a good cook it reads as a warning that the kitchen is disorganized.
• "Team player" is a request for compliance, and it is what people write when they cannot describe the team.
• "Competitive pay" means we will not tell you, which costs you every candidate who has other options and no time to waste. It also, in a growing number of jurisdictions, is becoming unlawful — several states and cities now require a pay range in the posting.
• "Apply within" requires the candidate to show up during service, when nobody can talk to them.

⚠️ Where the Money Leaks

Refusing to post a wage is the single most expensive line in a job ad.

A posting without a number is answered disproportionately by people with no better options — which is precisely the opposite of the selection you want. The candidates you most want are employed right now, are not desperate, and will not spend an evening applying to find out whether the job pays less than the one they have.

Post the range. If the range embarrasses you, that is information about the wage, not about the posting.

What a good one contains

A description worth writing has six parts: the restaurant in two sentences (what kind of food, what kind of room, what kind of service); the actual work (which station, what techniques, what volume — "you will run the sauté and pasta station on a five-night dinner service averaging 95 covers"); the schedule, honestly stated (five shifts, Tuesday through Saturday, 2:00 p.m. to close, weekends required); the pay range and the benefits that exist; what is genuinely required versus what can be taught; and how the process works and how long it takes.

That last item is worth dwelling on. Telling a candidate "we respond within three days, then a conversation, then a paid four-hour stage, then a decision within a week" does two things: it filters out people unwilling to do a stage, and it signals that this restaurant runs on systems. In an industry where a great many applications vanish into silence, a stated process is a competitive advantage that costs nothing.

The description is an operational document, not an advertisement

Here is the part most operators miss, and it is why this material sits in a chapter about money rather than in an appendix about forms.

A job description is not a human-resources formality you complete because a template told you to. It is an operational document, and it keeps working long after the posting comes down. It does at least three jobs after the hire, and each of them costs real money on the day the document does not exist.

It is the standard, written down before anybody had a reason to argue about it. Six weeks in, a cook is leaving the station dirty and taking twenty minutes off the top of every shift. If the description said "you break down and clean your own station" and "on at 2:30", the conversation is a reminder. If it did not, the conversation is a new rule invented by a chef who is annoyed — which is how a correction becomes a grievance.

It is the schedule expectation, agreed in advance. The single most common early separation in this industry is a person who took a job believing it was four shifts and discovering it is five, or believing weekends were negotiable. That is not a bad hire. That is a document failure, and §17.9's log will show you exactly what it costs.

And it is the record behind a classification decision. Whether a position is exempt from overtime, whether someone is an employee or a contractor, and what a job's actual duties are — these are questions with legal answers that turn on what the work really consists of, and the description is the contemporaneous record of what you said the work was. Chapter 20 covers the framework and the tests; do not try to reason it out from this chapter. What matters here is that the document you write in ten minutes on a Friday is the document somebody may read carefully two years later.

Five things a description has to name, and most postings name none of them: the station, the shift, the physical demands, the reporting line, and the standard.

SAMPLE POSTING — LINE COOK, SAUTÉ / PASTA STATION           [constructed teaching example]

  POSITION       Line cook — sauté and pasta station.
  REPORTS TO     The sous chef. The chef expedites Wednesday through Saturday.
  SCHEDULE       Five shifts, Tuesday through Saturday. On at 2:30 p.m., out between
                 10:30 and 11:15 p.m. Both weekend nights are required. A weekend
                 brunch shift may be offered and is not required.
  WAGE           $19.50–$21.50 an hour depending on experience, reviewed at 90 days.
                 Shift meal every shift worked. Paid time off accrues from day one.
                 We pay for your food-handler course.

  THE RESTAURANT Sixty-eight seats, wood-fired hearth, a short seasonal American menu
                 that changes about six times a year. Dinner Tuesday through Saturday
                 plus weekend brunch. Ninety to one hundred twenty-five covers on a
                 weekend night. Everything on your station is made here.

  THE WORK       You run sauté and pasta alone from 5:00 p.m. Four to six items live on
                 the station at a time and two of them change every six weeks. You do
                 your own mise from 2:30, you break down and clean your own station, and
                 you leave a prep list for the next day before you go. On a Saturday you
                 will fire roughly 60 covers off two burners and a plancha.

  PHYSICAL       Standing eight to nine hours. Lifting to 50 lb — stock pots, produce
                 cases, a bag of flour — with help available. Sustained work within four
                 feet of an open hearth; the station runs hot. Frequent reaching,
                 bending, knives, and hot pans.

  REQUIRED       Two years on a hot line in a scratch kitchen, OR one year plus a
                 culinary program. A food-handler card before your first paid shift, or
                 within your first week if you don't have one. Legal authorization to
                 work in the United States, verified on day one as we do for everyone.

  WE WILL TEACH  The hearth. Our pasta. Our specs and our plating. Pars and ordering.
                 You do not need to arrive knowing any of it.

  THE STANDARD   Tickets in the window inside twelve minutes on a weekend night. Station
                 set by 4:45, broken down clean by close. If a plate isn't right it does
                 not leave the pass — we would rather re-fire it and tell the table.

  THE PROCESS    Apply by email with anything that shows where you have cooked; a résumé
                 is welcome and not required. We reply to everyone within three business
                 days. Then a 40-minute conversation, then a PAID four-hour stage on a
                 Tuesday or Wednesday, then a decision within one week.

Read the physical-demands block again, because operators leave it out for the wrong reason. They leave it out because it sounds discouraging. It is discouraging, accurately — and a candidate who cannot stand nine feet from a hearth for nine hours needs to know that before they take the job, not on day four. Stating the demands honestly is also the starting point for any conversation about accommodation, which has a framework of its own; Chapter 20 owns it. What belongs here is the operational rule: write down what the body actually has to do, before anyone applies, and write the same thing for every candidate.

And notice what the posting does not do. It does not promise a promotion. It does not say "family." It does not say "opportunity for growth." It names a station, a schedule, a wage range, a standard, and a process — and the candidate who answers it has already agreed to all five.

👨‍🍳 On the Line

How hiring actually happens, versus how it should.

The honest version: a cook quits on a Thursday. The chef, who is now working that station, posts something on Friday between prep and service. Over the next ten days they screen candidates on a phone in the walk-in, interview two of them standing up at 4:15 p.m., and hire whoever seems least likely to disappear. The whole process is conducted by a person who is exhausted and short-staffed, which is precisely the state in which humans make their worst decisions.

Everything in this chapter is an attempt to move decisions out of that moment. A description written in advance, a scoring sheet prepared in advance, a stage structured in advance, and a bench of past applicants kept warm — all of it exists so that the tired chef on Friday is executing a system rather than improvising one.

This is why hiring is a slow-season activity. The best time to build the pipeline is the week you do not need it.

17.3 Sourcing: referrals, schools, boards, walk-ins, and poaching ethics

Sourcing channels are the routes by which candidates reach you. They differ enormously in cost, in volume, and — most importantly and least often measured — in how long the resulting hires stay.

The pattern is consistent and worth internalizing: the cheapest channels produce the best retention, and the most expensive channel produces the worst. Job boards are where you go when the pipeline has failed, and you pay for that failure twice — once in posting fees and once in the shorter tenure of the people you hire from them.

Which means the highest-return sourcing activity available to a restaurant is a referral program that people actually use. A \$300 referral bonus, paid half at hire and half at ninety days, is roughly one-seventh of a line cook's replacement cost — and it recruits from a pool that has been pre-screened by someone whose own standing depends on the recommendation.

Channel by channel: what each one is good for, and what it asks of you

The table above compares channels. This is the part that tells you how to actually work them, because a sourcing plan is not a preference — it is five habits, each of which has to be built before the week you need it.

Referrals: the highest-yield channel, and the highest-risk one. A referral arrives pre-screened by someone who has worked your Fridays and knows exactly what the job is. That is why it retains best, and it is most of the argument for the bonus.

It carries three risks that operators discover late. First, referrals reproduce the room you already have. They come from the same schools, the same neighborhoods, and the same networks, and a restaurant that sources exclusively from its staff's address books will look, in three years, like one person's address book. That has ethical weight and, in some circumstances, legal weight — Chapter 20 is where the framework lives. The operational answer is simple: run referrals as one channel among five, never the only one. Second, a failed referral costs you two people's morale, not one — the referrer feels responsible, and often behaves as though they are. Third, and least anticipated: people who arrive together frequently leave together. A pair of friends on the same station is a wonderful eleven months and a very bad Tuesday.

Industry boards and groups: cheap, targeted, and only available to people already in the room. The regional restaurant-industry job pages, the local chef and bartender groups, the trade association's board — these reach people who are currently employed, which is precisely the pool a general job board misses. The cost is not money. The cost is that you have to already be a member: to have answered other people's posts, sent someone a cook you could not use, and shown up. You cannot join one of these in the week you are short.

Walk-ins and the stage that follows them: still real, and still badly handled. Walk-ins remain a meaningful channel for hosts, bussers, and servers, and they are usually wasted because they arrive at 6:40 p.m. and are handed a form by a bartender. Fix it with one line on the door and the posting: applications taken Tuesday through Thursday, 2:00 to 4:00. That converts an interruption into a scheduled channel, and it means the person who walks in gets a conversation instead of a clipboard.

Culinary and hospitality programs: a real pipeline that costs mentorship rather than money. Programs are excellent for prep, garde manger, brunch line, and pastry, and they are a genuine bargain if — and only if — you have the senior person and the hours to teach. A program graduate has cooked for an instructor, on a schedule, with a rubric. They have not cooked for a Saturday. The relationship is built the same way as the industry groups: a demo, an externship slot, a chef who takes the call. It takes a year to build and then it produces candidates every May and December, forever.

The rehire list: the cheapest and best channel in the building, and almost nobody keeps one. Every person who leaves on decent terms goes on a list with three fields — the date, the reason, and one line about what would bring them back. You call the list quarterly. It works because a returning cook already knows the menu, the point-of-sale (POS) system, the walk-in, the closing checklist, and where the fire extinguisher is. Look at what that does to §17.1's build: the advertising line disappears, the screening line collapses, the stage is unnecessary, and the learning curve is a week rather than a quarter. What does not disappear is the ramp — you still pay wages while they get their speed back — but you are paying it for days instead of months.

🧮 Run the Numbers

What it costs to put one line cook on the schedule, by channel. (constructed teaching example)

This is acquisition only — getting a person to their first shift. It is a subset of the \$2,180 in §17.1, not a competitor to it; the training, the ramp, and the vacancy overtime sit on top of every row identically. Management time is priced at the same \$32 an hour fully loaded that §17.1 used.

Channel	Cash out	Manager hours to one hire	Hours at \$32	Total to fill one opening
Rehire list	\$0 | 1.0 | \$32	\$32
Walk-in	\$0 | 2.5 | \$80	\$80
Culinary program	\$0 | 3.5 | \$112	\$112
Industry board or group	\$60 | 4.0 | \$128	\$188
General job board (§17.1's base case)	\$150 | 5.0 | \$160	\$310
Staff referral	\$300 | 1.5 | \$48	\$348

Two things in that table are worth more than the totals.

The general-board row is §17.1's build. The \$150 of advertising and the five hours of screening in the replacement-cost table are the job-board case — which means the \$2,180 you have been quoting all chapter is quietly the most expensive way you can hire. Every other row makes it smaller.

The referral is the most expensive line and the best purchase on the table. It costs \$348 against the board's \$310 — and the \$300 of it is paid in halves, the second half only if the person is still there at ninety days, which means you are buying a hire you have already partly verified. Against a channel that produces the shortest tenure in the business, spending \$38 more is not a close call.

And the cheapest row, at \$32, is the one you cannot buy at all.

The chicken and the egg, named rather than solved

Look at that list honestly and a pattern falls out that most hiring advice steps around.

Every channel that produces good hires requires you to already be a good employer. The rehire list requires people who left on terms good enough to come back. Referrals require staff willing to attach their name to you. The culinary program requires a chef there who thinks their best student would be safe with you. The industry groups require a reputation among people who have all worked with someone who worked for you.

The general job board requires none of that. It is the one channel available to a restaurant nobody wants to work at — and it produces the shortest tenures, which makes the restaurant harder to work at, which sends it back to the board. That is a loop, it is real, and it is the actual mechanism behind a great many restaurants that describe themselves as being in a labor market problem when they are in a reputation problem.

There is no clever way out of it, and you should be suspicious of anyone who offers one. The honest way out is small, slow, and made entirely of things in this chapter: put the wage in the posting, answer every applicant, pay the stage, keep the schedule you promised, and let one person leave on good terms and say so to somebody. It is a two-year project. There is no version of it that helps you this week.

Which is exactly the hiring angle on the second Friday in October. At 3:40 p.m. the grill cook does not arrive, and every question in the building is about tonight — call someone in, run short, cut the menu, comp the party. Chapter 19 prices those choices. But the decision that produced the crisis was made in June, and it was a decision about a list.

A restaurant with four names it could call at 3:40 — a bench candidate who scored a 16 in April, a prep cook who left for school and said to call, a culinary student who staged in May, a bartender's brother who cooks at a hotel — has an expensive and unpleasant Friday. A restaurant with no list has an emergency, and on Monday it will post to a general board, spend \$150 and five hours, hire whoever answers, and do the entire thing again in eleven weeks. The pipeline is not what you use when you are hiring. It is what determines whether a no-show is a problem or a crisis.

⚖️ Code and Compliance

Poaching, and the agreements that try to stop it.

Approaching a good cook who works somewhere else is normal, legal in most circumstances, and how a great deal of restaurant hiring happens. It is also a relationship decision: the restaurant industry in any given city is small, and an operator known for stripping a neighbor's line will find their own line stripped.

Two legal notes, both jurisdiction-dependent and both worth real advice:

• Non-compete agreements for hourly restaurant workers have come under significant regulatory and legislative pressure in recent years, and their enforceability against low-wage employees varies widely by state — in several, they are sharply restricted or void. Do not assume one you signed, or one you ask someone to sign, is enforceable.
• No-poach agreements between employers — including between franchisees of the same brand — have drawn antitrust scrutiny. An informal understanding among neighboring operators not to hire each other's staff is not obviously lawful.

Verify locally with an attorney before relying on any of this. The safe operating posture is to compete for people on wage, schedule, and culture rather than on contract.

17.4 Screening and the structured interview

Here is a finding that should change how you hire, and that most operators have never encountered: the unstructured interview — the conversation, the gut read, the "I can tell in five minutes" — is among the weakest predictors of job performance in the entire body of personnel-selection research. It is not useless. It is close to it, and it is reliably worse than the people conducting it believe.

The reason is not that interviewers are stupid. It is that an unstructured conversation measures rapport, confidence, and similarity to the interviewer far more efficiently than it measures whether someone can hold a station on a Saturday. Restaurant hiring is unusually exposed to this, because the industry selects for people who are personable and quick, and those are exactly the traits an unstructured interview over-rewards.

What a structured interview is

A structured interview asks every candidate for a position the same predetermined questions, in the same order, scored against a defined rubric. That is the whole idea, and it is dull, and it works substantially better.

Three practical rules make it work:

1. Ask about behavior, not hypotheticals. "What would you do if you got slammed?" invites a performance. "Tell me about the worst night you've worked. What actually happened, and what did you do?" invites a report, and reports are checkable.
2. Score immediately, before discussing. Two managers who confer before scoring produce one opinion held by two people.
3. Ask what the job requires, and nothing else. Every question should map to a duty in the description from §17.2.

The questions, and what the answers actually tell you

Structure tells you how to ask. It does not tell you what to ask, and that is where most scorecards quietly fail — a rigorously administered set of useless questions is still useless, and now it has a total at the bottom that makes it look like measurement.

Sort your questions into two piles before you write any of them.

Questions that test experience establish what a person has done: which stations, what volume, what kind of kitchen, how many covers on a Saturday, whether they have ever worked live fire. These are the easy half and they are checkable — a reference call can confirm them. They are also the half most likely to be inflated, so ask them in a form that is hard to inflate. Not "have you worked a busy station," which has only one answer, but "what were your Saturday covers, and how many of them came off your station?" A number is much harder to improvise than an adjective.

Questions that test judgment establish what a person does when the plan fails — which is the only condition a restaurant reliably produces. These are the ones that predict, and they are almost always behavioral: they ask for a thing that happened, not a thing that would.

Then there is a third pile, which is most of what gets asked in restaurant interviews.

The questions that predict nothing. "Where do you see yourself in five years?" measures how recently someone read an article about interviews. Every competent candidate has an answer and it is the same answer, and the honest reply — cooking, probably, somewhere with a garden — scores worse than the rehearsed one. "What's your greatest weakness?" invites a rehearsed lie and penalizes the candidate who tells you something true. "Are you a team player?" has never once received a no. "Tell me about yourself" is a perfectly good way to let someone's shoulders come down, and it should be scored zero, because that is what it is for. "Why do you want to work here?" is worth asking and worth about thirty seconds; it tells you whether they looked at the menu, and nothing else.

The questions that predict something. Four, and you can run an entire kitchen hiring process on them:

1. "Tell me about the last time you were properly in the weeds. What happened?"
2. "Describe a service that went badly. What did you do, and what did it cost?"
3. "You've plated something you're not proud of and the expo is standing there waiting. Walk me through what you do."
4. "Tell me about the chef you learned the most from — and one thing you disagreed with them about."

Now the useful part, which is not the questions. It is what the answers actually tell you.

The pattern across that table is one thing, and it is the thing to write on the scorecard: score the behavior described, not the fluency of the description. A cook who tells a flat, awkward story containing three specific details — the ticket time, what they did with the second sauté pan, what the sous said at ten o'clock — is telling you far more than one who tells a smooth story containing none. Restaurant interviewers systematically get this backwards, because the industry runs on people who are quick and warm in conversation, and the interviewer is one of them.

One question to close with, every time: "What would you need from us to do this job well?" It is not a soft question. The answers sort candidates immediately — a clear prep list, someone to tell me the specs once instead of five times, knowing what time I actually leave — and every one of them is a piece of information about what you would have to be for this person to still be here in a year.

🧾 Read the Numbers

```text FIGURE 17.2 — "Line cook, structured scorecard" [constructed teaching example] THE ARTIFACT A five-question scored interview sheet, used identically for all four candidates for the sauté/pasta position. Scored 1-5 immediately after each interview, before any discussion between the two managers. THE CONTEXT Bellwether, week 3 of hiring-up before opening. Two interviewers: the chef and the sous. Four candidates, one opening.

               QUESTION                                    C1   C2   C3   C4
               1. Worst night worked: what happened,
                  what did you do?                          4    5    2    3
               2. Walk me through your mise for a
                  five-item station.                        5    3    3    4
               3. A ticket comes back. Describe the
                  last time that happened to you.           3    4    2    4
               4. How do you know when you're behind,
                  and what do you do first?                 4    4    2    2
               5. Describe a time you disagreed with
                  a chef. What did you do?                  2    5    4    3
               ─────────────────────────────────────────────────────────────
               TOTAL (of 25)                               18   21   13   16

WHAT IT SHOWS Candidate 2 scores highest overall and notably on questions 1, 3, and 5 — the three questions about behavior under pressure and conflict, which are the ones the Friday-night scenario in Chapter 14 actually tests. WHAT IT DOESN'T It does not show whether any of them can cook. Interview scores measure what people SAY about their work. Question 2 gets closest and is still a description of mise en place, not mise en place. Only the stage in §17.5 tests the hands. It also cannot see fit with this specific team, or whether Candidate 2's strong answer on question 5 reflects good judgment or a habit of conflict. THE DECISION Stage candidates 2 and 1. Do not stage 3. Keep 4 warm — a 16 with a strong mise answer is a bench candidate worth calling in eight weeks. THE LESSON Score before you discuss, and let the scorecard narrow the field rather than confirm the impression you formed in the first thirty seconds. ```

Notice the last line of the decision: keep candidate 4 warm. Most restaurants discard every candidate but the hire, and then start from zero eleven weeks later when someone quits. A bench of three previously-interviewed people is worth more than any job board.

What you may not ask, and where the risk actually is

§17.6 sets out the categories and Chapter 20 covers the law properly. What belongs here is the operational point, because the danger in an interview is almost never the question list.

Nobody writes do you have children on a scorecard. It arrives at 4:20 p.m., in the ninety seconds of small talk before the structured part begins, in the friendliest possible register — do you live nearby? oh, is that a long drive with the kids? which church is that behind? — asked by a manager who is trying to be kind and has just walked through three protected categories without noticing.

Two habits fix it, and neither costs anything:

• Treat the whole encounter as the interview. It starts at hello and ends at the door. If a question would be wrong on the sheet, it is wrong in the parking lot.
• Convert every curiosity into a job requirement. You do not need to know whether someone has children; you need to know whether they can work Friday and Saturday nights, which is a question about the job and which you ask every candidate identically. You do not need to know someone's age; you need to know whether they are old enough to serve alcohol where you operate, which is also a question about the job. Nearly every improper question has a proper version that gets you the information you actually needed.

And if a candidate volunteers protected information — they will, because people talk, and because a person trying to be liked will explain their whole life — do not follow it, do not write it down, and do not let it move the question order. Go back to the sheet. The consistency is the defense, and it is also, incidentally, the thing that makes the scorecard predictive in the first place.

👨‍🍳 On the Line

The interview runs in both directions, and most operators are losing the direction they aren't watching.

An interview at 4:15 p.m. feels like an examination with one examiner. It is not. The person across the table is deciding whether to work here, and in a tight labor market they very often have somewhere else to be on Tuesday.

Which makes the last three minutes — "do you have any questions for me?" — the most informative part of the conversation, and the part managers treat as a formality while they are already thinking about the delivery.

A candidate who asks nothing has told you something. Usually one of three things: they have decided they do not want the job and are being polite; they have never been in a kitchen where asking was safe, which is worth knowing and is fixable; or they are not going to ask questions on the line either, which is not fixable and is expensive on a Friday.

The questions that should reassure you are the operational ones. What does the schedule actually look like? Who orders for this station? How often does the menu change? What time does the kitchen really leave? What happened to the last person in this job? Those are the questions of somebody planning to still be here in a year.

The one that should give you pause is a candidate who asks only about pay and only about time off, with nothing at all about the work — though hold that lightly. A cook who leads with wage and schedule has often simply been burned by an operator who was vague about both, and this industry produces those in quantity.

Have a true answer ready for "what happened to the last person in this job?" You will be asked it, and the industry in any city is small enough that a false answer is a short-lived one. If the honest answer is "the last three lasted under six weeks," the candidate is going to find that out — and so should you. §17.10 is about what that answer is actually telling you.

🔍 Check Your Understanding

1. Why does an unstructured interview systematically over-reward candidates who resemble the interviewer?
2. A manager says the scorecard "takes the humanity out of it." What is the strongest version of that objection, and what is the answer?
3. Bellwether interviews four candidates and stages two. What does staging only the top scorers risk missing?

(1: Rapport is easier to establish with people who share your background, references, and speech patterns, and rapport is what an unstructured conversation actually measures. 2: The strong version is that restaurants are relational and a scorecard cannot measure whether someone will fit a specific team — true, which is why the scorecard narrows rather than decides, and the stage and the reference check follow it. 3: Interview scores measure self-description; a candidate who interviews poorly may cook superbly, and the correlation between the two is weaker than most operators assume. Where the position is hands-first, stage more people and interview less.)

17.5 The stage and the working interview

A stage — from the French stagiaire — is a trial shift in which a candidate works in the kitchen so both sides can assess fit. In the front of house the equivalent is usually called a working interview. It is the single most predictive step in restaurant hiring, because it tests the hands rather than the résumé.

It is also the step most likely to create legal exposure, and a great many restaurants run it in a way that is difficult to defend.

⚖️ Code and Compliance

The unpaid stage is a wage-and-hour problem waiting to happen.

The tradition of the unpaid trial shift runs deep in professional kitchens, and it sits uneasily with wage-and-hour law. The general principle under the federal Fair Labor Standards Act — and Chapter 20 covers the framework properly — is that if the work primarily benefits the employer, it is work, and work is compensable. A candidate who spends four hours doing prep that you would otherwise have paid someone to do has performed work, whatever the shift was called.

There is a narrow space for a genuinely observational visit — a candidate watching service for an hour, touching nothing. The moment they pick up a knife and produce something you sell, the analysis changes.

The safe practice, and the one this book recommends: pay for the stage. Four hours at the position's starting wage is roughly \$88 — about four percent of what it costs to replace a line cook, and the cheapest insurance available against both a bad hire and a wage claim. It also signals something true about how the restaurant treats people, to a candidate who is evaluating you.

Additional considerations, all jurisdiction-dependent: workers' compensation coverage for someone not yet on payroll, food-handler certification requirements before a person handles food, and minors' work restrictions. Verify locally.

Running one that tells you something

A stage that consists of "jump on garde manger and see how it goes" produces an impression, not information. A structured one has a plan:

• Give the same tasks to every candidate for the position. At Bellwether, the sauté stage is: break down and portion for the pasta station, execute four covers of the hand-cut pasta during a slow early service, then break down and clean the station.
• Watch three things specifically. Knife work and organization (is the mise ordered? does the board stay clean?), behavior under a small pressure (fire two tickets close together and watch), and how they take a correction — which is the single most informative moment in the shift, because you will be correcting this person for months.
• Watch the close. Anyone can be sharp for the first hour. How someone breaks down a station at the end of a shift tells you what they will be like in month six.
• Ask the team afterward. The cooks who worked next to the candidate know things the chef standing at the pass does not.

👨‍🍳 On the Line

The correction test.

Midway through a stage, correct the candidate on something small and specific — a portion, a knife angle, the way they wiped a plate. It does not much matter whether the correction is important.

What you are watching for is the next ninety seconds. Do they say "yes chef" and do it? Do they explain why they did it the other way? Do they do it correctly once and then drift back? Do they get quiet and stay quiet?

None of those responses is automatically disqualifying, and the point is not to find someone submissive — a cook who says "I did it that way because the pan was too hot" may be right, and the kitchens that punish that reply end up with cooks who do not mention that the walk-in is at 46°F. What you are learning is how information will travel between you and this person for the next two years, and you are learning it for the price of one small correction.

What the stage is not testing

Operators watch the wrong thing during a stage, and they watch it for an understandable reason: it is the most visible thing in the room. They watch the knife.

Knife work is close to the least informative signal available in four hours. A competent cook on an unfamiliar board, with their roll on the wrong side of the station and the chef standing eight feet away, is slower and rougher than they actually are. Meanwhile a weak cook who has spent two years doing nothing but brunoise for somebody's prep list will look superb for the entire shift. Raw speed is worse still, because from the pass, nerves and slowness are indistinguishable for the first two hours. Both of these things are teachable, and both arrive on their own by about week six.

The bullet list above pairs knife work and organization, and the pairing is doing all the work. It is the organization you are reading. The knife is scenery.

Rank what you can see by how much it actually tells you:

The fourth row deserves its own paragraph, because it is different from the correction test above. The correction test is the ninety seconds after you say something. This is the ninety seconds after nobody says anything. They overcook a piece of fish. They drop a pan. They plate an app with the wrong garnish and catch it themselves at the pass. The only person who noticed is them.

What happens next is the most honest thing you will see all night. Do they say it out loud — "chef, refiring the halibut, two minutes" — and reset? Do they hide it and hope? Do they get flustered and lose the next three tickets, which is the actual failure mode, because one mistake is free and the spiral is not? Do they stop and stand still?

A cook who names their own error before you find it is a cook whose station you will still be able to read a year from now, from the pass, without asking. That is worth more than every knife cut you will watch all night.

Running the stage for the candidate

Half of a stage is you evaluating them. The other half — routinely, almost universally wasted — is them evaluating you, and four hours inside your kitchen is an enormous quantity of information to hand somebody you have not hired.

Five things, none of which takes more than a minute:

• Tell the team before the candidate arrives. "We have a sauté stage tonight, three to seven." A station that was not told somebody is coming will treat them as an obstacle, and the candidate will read that correctly.
• Introduce them to the people they will actually work next to — the station on either side, the dish pit, the expo. Nobody should have to introduce themselves twice while carrying a hotel pan.
• Give them the same brief you would give a new hire: where things are, what the specs are, what happens if something goes wrong, and who to ask.
• Feed them. Family meal, or a plate after. It costs about two dollars and it is the single clearest statement your restaurant makes about how it treats people who work in it.
• Tell them when they will hear, and then actually call on that day — including the no. A candidate who is left in silence for two weeks will tell other cooks, and other cooks are the channel §17.3 said retains best.

A candidate who spends four hours being ignored by a station nobody told they were coming has learned exactly what §17.2's posting was trying to say about how this restaurant runs, and has learned the opposite of it.

⚖️ Code and Compliance

Paying the stage is a payroll action, not a gesture.

The callout above says to pay for the stage. This is the part operators get wrong immediately afterward: "we'll pay you for the trail if we hire you" is not a fix, and neither is cash out of the drawer at the end of the night.

If you pay somebody for a shift, you have employed them for a shift, and the machinery that attaches to employment attaches to that shift:

• They go on payroll before the shift, not after. Which means the new-hire paperwork in §17.7 — including the employment-eligibility verification — starts before they touch a knife, rather than a week later once you have decided you like them.
• The hours belong on a time record, at a stated rate, and they count as hours worked in that workweek for overtime purposes like any other hours.
• Workers' compensation coverage and any required food-handler certification are questions to settle before the shift, not to discover after one.
• Paying cash off the books to avoid all of the above does not simplify the problem. It converts a hiring practice into a payroll problem, and it is the worst version of this available.

That is more administrative friction than a four-hour trial shift feels like it deserves — which is precisely why the unpaid stage persists. The shortcut is genuinely more convenient, and convenience is how this industry acquires most of its inherited bad customs. Build the two-minute version once — a stage-day packet, a rate, a time entry — and it costs nothing forever afterward.

All of this varies by jurisdiction. Verify locally, and have an employment attorney look at your standing practice once before you rely on it. Chapter 20 treats the framework properly.

17.6 References, background checks, and what you may not ask

Reference checks are the step every operator skips and every operator wishes they had not. They take twenty minutes, and the useful part is a single question: "Would you hire this person again?" The answer, and more informatively the pause before it, tells you most of what a reference can legally tell you — many employers restrict themselves to confirming dates and title.

Call the reference the candidate did not offer, when you can do so without jeopardizing their current job. Ask permission first, always: a candidate who is currently employed and has not told their chef they are looking is entitled to have that respected, and a restaurant that burns a candidate that way will be known for it within a month.

Making a reference call produce something

Most reference calls fail inside the first fifteen seconds, and they fail for a structural reason rather than a personal one. Understanding the structure is what lets you get around it.

A great many employers — quite possibly including yours — have a standing policy of confirming dates of employment and job title and nothing else. That policy exists because a former employer who characterizes an ex-employee's performance carries some exposure if the characterization is wrong and damaging, and the cheapest way to have no exposure at all is to say nothing. Whether that policy is legally required, merely advisable, or simply what the handbook has always said varies by jurisdiction and by whoever wrote the handbook; Chapter 20 is where the employment-law framing lives. What matters operationally is that you should expect the wall, and that the wall is not about your candidate.

So build the call for it. Five minutes, in this order:

Get the right person. For a cook, that is the chef or the sous — the person who stood next to them — not a corporate line or a payroll office. The candidate can usually give you a direct number, and you should ask for one.

Say why you are calling in one sentence, and name the job. "They've applied to run sauté and pasta for me, five nights, about a hundred covers." A former chef who knows what you are actually asking about can answer usefully even under a restrictive policy, because now the question is about fit rather than about character.

Ask the closed question first: "would you hire this person again?" It is the whole reference in six words. There are three possible answers — a yes, a no, and a pause — and the pause is an answer.

Then the two open questions that survive a neutral-reference policy. "What would you tell somebody taking them on?" and "What did they need from a chef to be at their best?" Both are framed as advice rather than assessment, which is a real distinction and not a trick: a chef who will not grade a former employee will very often help a colleague avoid a preventable mistake.

Then ask about the one specific thing you are unsure of, and only that. If the stage showed a cook who went quiet under pressure, ask about a Saturday. Do not go fishing; you will catch something that is not true.

Close with a question that costs them nothing. "Anything I should know?" Then stop talking. A long silence here is information, and most managers fill it.

Two calls per finalist, ten minutes total. Run it even though it will usually change nothing — because of the small number of times it does, and because ten minutes against \$2,180 is not a close call. Note the answers on the same sheet as the interview scores, in the same words each time, for every candidate for the position.

One point on sequence. The safest and most common practice is to run any third-party background check after a conditional offer, never as a first-pass filter, and to run it identically for every candidate for a position. In a growing number of jurisdictions that timing is not merely good practice but prescribed — which is what the fair-chance and "ban-the-box" rules in the callout below are about. Verify locally.

Background checks — criminal history, credit, driving records — are governed by real law and should be approached carefully.

⚖️ Code and Compliance

What you may not ask, and what you must do if you run a check.

This area is heavily regulated, varies by jurisdiction, and changes. The structure to understand:

• Protected characteristics. Federal law prohibits employment discrimination on the basis of race, color, religion, sex (including pregnancy, sexual orientation, and gender identity), national origin, age (40+), disability, and genetic information. State and local law commonly adds categories. Do not ask about them, and do not ask questions that indirectly elicit them — about children and childcare, citizenship as distinct from work authorization, religious observance and weekend availability, or health conditions.
• Third-party background checks are consumer reports under the Fair Credit Reporting Act (FCRA), which imposes real obligations: standalone written disclosure, written authorization, and — if you intend to act adversely on the report — a pre-adverse-action notice with a copy of the report, a waiting period, and a final notice.
• "Ban-the-box" and fair-chance laws in many states and cities restrict when you may ask about criminal history and require an individualized assessment of whether an offense relates to the job.
• Salary-history bans exist in a growing number of jurisdictions; asking what someone currently earns may itself be unlawful where you operate.

None of this is a reason to skip due diligence. It is a reason to have a written, consistent process reviewed once by an employment attorney and then followed identically for every candidate. The consistency is the defense.

A note on judgment. This industry has historically been one of the few reliable routes into stable work for people with criminal records, unstable housing, or interrupted education, and that is genuinely one of the good things about it. An operator who screens out everyone with a record is both narrowing their pipeline in a tight labor market and opting out of something the industry does well. The relevant question is never "is there a record" but "does this specific history bear on this specific job" — which is, not coincidentally, close to what fair-chance law requires.

17.7 The offer, Form I-9, and the paperwork of day one

The offer should be in writing, even for hourly positions, and it should state: position, wage, schedule expectation, start date, reporting relationship, and whether employment is at will.

At-will employment means either party may end the relationship at any time, for any lawful reason or no reason. It is the default in nearly every U.S. state, subject to significant exceptions — you may not fire for an unlawful reason, in retaliation for protected activity, or in breach of a contract you have created. Be careful about creating one accidentally: an offer letter promising "continued employment as long as performance is satisfactory" can undercut at-will status. Say what you mean, briefly.

What the offer should say, and what it should not

The offer letter is a one-page document that takes four minutes and prevents most of the arguments this chapter's arithmetic is otherwise going to pay for.

SAMPLE OFFER — LINE COOK, SAUTÉ / PASTA STATION           [constructed teaching example]

  We are offering you the position of LINE COOK on the sauté and pasta station.

  RATE           $20.00 per hour, paid every two weeks. Overtime at the legally
                 required rate for hours over 40 in a workweek.
  POSITION       Line cook, sauté/pasta. Hourly, non-exempt.
  REPORTS TO     The sous chef.
  SCHEDULE       Five shifts a week, Tuesday through Saturday, on at 2:30 p.m. Both
                 weekend nights are required. Scheduled hours vary with business
                 volume. Schedules are posted [__] days in advance.
  START DATE     [date]. Please arrive at 1:30 p.m.; the sous chef will meet you at
                 the back door.
  BEFORE DAY ONE Bring documents you choose from the attached list, for the
                 employment-eligibility verification every employee completes.
  ALSO           Shift meal on every shift worked. Paid time off accrues per the
                 handbook. We pay for your food-handler course.

  This is at-will employment: either you or the restaurant may end the employment
  relationship at any time, for any lawful reason or for no reason. Nothing in this
  letter is a contract for a fixed term, and nothing in it guarantees a number of
  hours.

  This letter states the whole of what has been offered. Anything discussed and not
  written here is not part of the offer.

  Please sign and return by [date].

Six fields and two sentences. The schedule-posting line is bracketed on purpose: whether advance notice is required at all, and how much, is a question of local law that Chapter 20 owns. Fill it in with what you will actually do, then do it.

Now what the letter must not say, which is where operators create problems out of goodwill:

• No conditional promises of continued employment. "As long as things are going well" is the phrase that undoes the at-will paragraph two inches above it.
• No promises about the future you have not decided to keep. "We see you on the hearth by summer" is a lovely thing to say at 11:40 p.m. and a genuine grievance in August. If you mean it, write it with a date and a condition. If you do not mean it, do not say it.
• No guaranteed hours you cannot guarantee. "Full time" is not a number. Write the number of scheduled shifts and write that hours vary.
• And no compensation figure that quietly includes tips. Telling a server they will "make about thirty an hour" is not a rate, it is a forecast, and it is a forecast you do not control. State the cash wage, state that it is a tipped position, and let Chapter 20's rules about tipped compensation govern the rest.

⚠️ Where the Money Leaks

The verbal offer is where nearly every employment dispute in a small restaurant begins.

Here is how it actually happens. The chef calls at 11:40 p.m. after service and says, "we'd love to have you — twenty an hour, start Tuesday." The candidate says great. That is the offer. Nothing is written, nothing is sent, and nobody is lying.

Six weeks later there are two honest and irreconcilable memories in the building:

What the restaurant remembers	What the new cook remembers
\$20.00 an hour | \$20.00 an hour, going to \$22 at ninety days
Five shifts, both weekend nights	Five shifts, one weekend night off a month
"We'll look at the hearth eventually"	Promised the hearth
Hours vary with volume	Forty hours a week

No single row of that is worth a lawsuit. All four rows together are exactly how a good hire becomes a resentful one by month three — and how a separation at month five becomes an unemployment hearing, a wage claim, or simply a story told in every kitchen in the neighborhood, which in a mid-size city is the most expensive of the three.

The letter costs four minutes and removes all four rows. Send it the same night, before anybody has slept on a different version of the conversation.

Form I-9 is the federal employment-eligibility verification every U.S. employer must complete for every employee. The mechanics that matter operationally:

• Section 1 is completed by the employee no later than their first day of work.
• Section 2 is completed by the employer within three business days of the start date.
• The employee chooses which acceptable documents to present. You may not specify which documents you will accept — doing so is document abuse and is itself unlawful.
• Retain the form for the required period and store it separately from the personnel file, so it can be produced for inspection without exposing unrelated records.

The week between yes and day one

There is a gap in every hiring process that almost nobody manages, and it sits between the moment a candidate accepts and the moment they walk in.

It is usually one to three weeks, because a decent candidate is giving notice somewhere. During that window you have made a hire and they have not started — which means the schedule you are about to write depends on something that has not happened yet. It is also, reliably, the window in which candidates evaporate: their current employer counter-offers, or another restaurant simply moves faster than you did.

Four things close that gap, and none of them takes ten minutes:

• Send the offer letter the same day. There should be a document with their name on it before they go to sleep.
• Send the practical details a week out: which door, where to park, what to wear, what time, who will meet them, and what the first shift will consist of. A person who does not know where to park at 1:20 p.m. on day one is already having a bad day one.
• Put them on the posted schedule the day they accept. A name on a schedule the whole staff can see is a commitment running in both directions, and it is how the team learns someone is coming before the person is standing in the kitchen.
• Have one human being make contact during the wait — the sous, the cook they staged next to, anyone at all. Thirty seconds of "see you Tuesday" is the cheapest retention spend in this chapter.

Then day one arrives, which is where the actual work of keeping this person begins. §17.8 is about that day and the twenty-nine after it, and Chapter 18 builds the training that runs underneath them. Everything up to this point has been acquisition. None of it counts until somebody is still here in March.

⚠️ Where the Money Leaks

Day-one paperwork is where good hires start to leave.

A new cook arrives at 2:00 p.m. for a 3:00 p.m. shift. Nobody is expecting them. There is no locker, no apron in their size, no name on the schedule, and the manager who hired them is off. They fill out forms standing at the end of the bar while prep happens around them.

That person has now learned the most durable thing they will learn about this restaurant: it is disorganized, and they are not important. Some fraction of them will be gone within six weeks, and the \$2,180 in §17.1 will be spent again.

The fix costs nothing: a printed packet ready the day before, an apron and a locker assigned, a name on the schedule, and one person whose job that shift is to be responsible for them.

17.8 Onboarding: the first shift, the first week, and the thirty-day check

Onboarding is the structured process of bringing a new hire from offer to full productivity. It is not orientation, which is paperwork, and it is not training, which is Chapter 18's subject. It is the deliberate management of the period in which a new person decides whether to stay.

That period is short, and the data operators collect on their own separations — when you bother to collect it — tends to show a striking concentration of departures in the first thirty to ninety days. Which means the money in §17.1 is not mostly lost to people who worked out for two years and moved on. It is mostly lost to people who were hired, trained partway, and left before returning any of the investment.

FIGURE 17.3 — WHERE A NEW HIRE IS ACTUALLY LOST       [constructed teaching example]

  investment
  recovered
     ▲
     │                                                    ┌──────── productive
100% │                                              ┌─────┘
     │                                        ┌─────┘
     │                                  ┌─────┘
  50%│                            ┌─────┘
     │                      ┌─────┘
     │                ┌─────┘
   0%├────────┬───────┘
     │  ▼     ▼      ▼
     └──┴─────┴──────┴────────────────────────────────────────────►
       day 1  wk 2   day 30                              month 6

       ▼ = the three moments people quit

       day 1   — nobody expected them; no apron, no locker, no plan
       week 2  — the honeymoon ends and nobody has checked in
       day 30  — they can do the job but cannot see a reason to stay

  The investment is spent BEFORE the return begins. Every departure to the
  left of the break-even point is a total loss, and all three common exit
  points sit there.

The countermeasures are unglamorous and cheap:

• Day one: someone expects them, by name. Equipment is ready. They are introduced to the team rather than left to introduce themselves. They shadow rather than produce. They leave knowing what tomorrow looks like.
• Week one: a named person is responsible for them — not "the team," which means nobody. Chapter 18 builds the training plan itself; what matters here is that somebody owns it.
• Week two: a deliberate check-in, held whether or not anything seems wrong, asking two questions that actually produce answers: "What has surprised you?" and "What is harder than it should be?" Both invite specifics; "How's it going?" invites "fine."
• Day thirty: a real conversation. Is this working, from both sides? A restaurant that ends an obviously-wrong hire at day thirty rather than month five saves months of drag and does the person a kindness.

🤝 Hospitality

You cannot serve guests generously with a staff you hire carelessly.

This book's third theme is that hospitality — not service, but the feeling a guest carries out the door — is what produces the second visit that makes a restaurant profitable. Chapter 23 works the arithmetic.

Here is the connection that makes hiring a hospitality decision rather than an administrative one. The experience your guest has is delivered, without exception, by the people you hired. A server three weeks into a job they were dropped into cannot read a table, cannot recommend the wine that lifts the check by four dollars, and cannot recover a mistake gracefully — not because they are inadequate, but because nobody gave them the ninety days it takes.

Danny Meyer's argument in Setting the Table runs in this direction and further: that the order of priority runs to employees first, because everything the guest receives passes through them. You may find that too strong. But the operational version is not arguable — a restaurant hires its guest experience, one person at a time, and then either invests in it or doesn't.

17.9 Reading your own hiring log

Everything so far in this chapter has been process. This section is about the one document that tells you whether any of the process is working, and it is a document that almost no independent restaurant keeps.

A hiring log is a spreadsheet with five columns, updated on the day somebody leaves. It takes about ninety seconds per departure. At the end of a year it is worth more than every benchmark in this chapter, because it is the only thing that turns §17.1's model of turnover into an account of it.

🧾 Read the Numbers

```text FIGURE 17.4 — "Twelve months of hires and separations" [constructed teaching example] THE ARTIFACT Bellwether's hiring and separation log for year one — the twenty-seven rows that closed. Rows for people still on the roster at year end are omitted. Dates are weeks from opening day; negative weeks are pre-opening hires. The cost column applies Figure 17.1's per-position replacement cost to each event. THE CONTEXT Bellwether, end of year one, read in January before next year's plan is written. Twenty-nine non-owner positions on the roster.

POSITION       HIRED  LEFT   TENURE  REASON RECORDED                     COST
─────────────────────────────────────────────────────────────────────────────────
Line cook      wk -4  wk 12  16 wks  wage; hearth job across town      $2,180
Line cook      wk -4  wk 19  23 wks  moved out of state                $2,180
Line cook      wk 14  wk 17   3 wks  released; couldn't hold Friday    $2,180
Line cook      wk 21  wk 23   2 wks  no-call, no-show after 5 shifts   $2,180
Server         wk -3  wk  8  11 wks  second-job conflict               $1,500
Server         wk -3  wk 40  43 wks  left for a fine-dining room       $1,500
Server         wk  6  wk  9   3 wks  quit; sections not as described   $1,500
Server         wk 11  wk 33  22 wks  returned to school                $1,500
Server         wk 25  wk 26   1 wk   no-call after 2 shifts            $1,500
Server         wk 30  wk 48  18 wks  released; cash-handling variance  $1,500
Busser/runner  wk -2  wk 16  18 wks  school schedule changed           $  700
Busser/runner  wk  3  wk  5   2 wks  no-call after 3 shifts            $  700
Busser/runner  wk  9  wk 12   3 wks  quit; closing shift ran too late  $  700
Busser/runner  wk 19  wk 23   4 wks  warehouse job, fixed hours        $  700
Busser/runner  wk 27  wk 41  14 wks  promoted to server (backfilled)   $  700
Dishwasher     wk -2  wk 22  24 wks  factory job with benefits         $  700
Dishwasher     wk  2  wk  3   1 wk   no-call after 2 shifts            $  700
Dishwasher     wk  7  wk  9   2 wks  quit; no reason given             $  700
Dishwasher     wk 15  wk 18   3 wks  released; attendance              $  700
Dishwasher     wk 24  wk 35  11 wks  moved                             $  700
Host           wk -2  wk 20  22 wks  graduated; left the city          $  900
Host           wk 10  wk 14   4 wks  quit; weekend availability        $  900
Host           wk 22  wk 46  24 wks  took a server job elsewhere       $  900
Prep cook      wk -3  wk  9  12 wks  wanted full time; we had part     $1,400
Prep cook      wk 12  wk 42  30 wks  left for a bakery                 $1,400
Bartender      wk -3  wk 32  35 wks  left to build a bar program       $2,400
Sous chef      wk -6  wk 44  50 wks  executive chef role, new opening  $5,200
─────────────────────────────────────────────────────────────────────────────────
27 separations                                                        $37,820

WHERE IT CONCENTRATES
  Line cook + server            10 of 27 events (37%)   $17,720 of $37,820 (47%)
  Busser + dish + host          13 of 27 events (48%)   $ 9,700 of $37,820 (26%)
  Tenure of four weeks or less  11 of 27 events (41%)   $12,460 of $37,820 (33%)
  Largest single event          the sous chef           $ 5,200 of $37,820 (14%)

WHAT IT SHOWS Three things, and the third is the one to act on. First, the realized year came in at $37,820 against the $38,070 the ledger in Figure 17.1 modeled — and the near-agreement is a coincidence, not a validation. The ledger models fractional separations that no real year can produce. Here the sous chef actually left (+$2,600 against model), the assistant manager did not (-$2,000), a bartender stayed (-$1,200), and a fifth dishwasher went (+$350). Those four variances net to -$250, which is the whole difference. Second, the money is concentrated in two positions. Line cook and server are 37% of the events and 47% of the cost. The three entry-level positions are 48% of the events and 26% of the cost — half the churn, a quarter of the money. A manager chasing headcount churn is working the cheap half. Third: eleven of twenty-seven people never reached day thirty, and they took $12,460 with them. That is Figure 17.3's curve, populated — every dollar of it spent on somebody who returned none of it. WHAT IT DOESN'T It does not say why. The reason column is the weakest data in the building: one line, typed by a manager who is sometimes part of the answer. "Moved" and "school" are the two most common entries in every restaurant's log and also the two easiest things to say instead of the truth. It does not separate regretted from unregretted. The wk-17 line cook and the wk-44 sous chef are each one row; one is a correction working as designed and the other is the most expensive loss of the year. It does not distinguish a separation from a transfer. The wk-41 busser was promoted to server — the position was refilled at $700, but the restaurant kept the person. A log that cannot tell those apart will make internal promotion look like failure. It does not show the vacancy. Between a separation and a replacement's first productive shift somebody covered, and that cost is in overtime and in the tables §17.1's leak callout described, not in this column. And it charges every separation a flat per-position figure. A dishwasher who left on day 7 and one who left on day 200 both cost $700 here. That is convenient, and it is not true. THE DECISION Three moves, and not one of them is "hire better." 1. Move the check-in. Figure 17.3 put a deliberate conversation at day thirty; this log says all eleven early departures were gone before it happened. Pull the first real conversation to day seven for every position, and keep the thirty-day one. 2. Stop opening line-cook and server searches on a general board. Those two carry 47% of the cost and are exactly where the bench, the rehire list, and the referral bonus in §17.3 are worth most. Work those channels for ten days before spending the $150. 3. Treat the sous chef as its own problem. One person was 14% of the year's turnover cost, and no sourcing channel touches that. It belongs to Chapter 21. THE LESSON Turnover has a shape, and the shape is not the average. Two positions carry half the money; half the events are worth a quarter of it; and two-fifths of the departures happen inside a window that costs almost nothing to manage. None of that is visible in a 75% turnover rate — and the log that shows it takes ninety seconds a departure to keep. ```

Nothing about that document is sophisticated. It is five columns in a spreadsheet, and it is the highest-return page in this chapter — higher than the scorecard, higher than the job description — because it is the only one that can tell you whether the others are working.

Keep it from opening day, for one blunt reason: you cannot reconstruct it later. Payroll can tell you who was issued a W-2 and when the last check cleared. It cannot tell you the tenure, it cannot tell you the reason, and by the time you want to know, the manager who could have told you has also left.

Three columns people leave off, and should not:

• Source. Which channel produced this hire. Without it, §17.3's retention column is received wisdom rather than your own data. Two years of it and it is the most valuable page in the binder, because it tells you where your good hires come from in your market — which may not match the table at all.
• Who hired them. This one is uncomfortable, which is the point. If one manager's hires have a median tenure of seven weeks and another's is eleven months, that is worth knowing before it costs you \$38,070 again.
• Regretted or not. One yes-or-no field, entered on the day of the separation, before hindsight has had time to soften it. A restaurant losing twenty-seven people a year of whom four were regretted has a very different problem from one where twenty were.

🔍 Check Your Understanding

1. Figure 17.1 models 26.8 separations at \$38,070. Figure 17.4's realized year records 27 at \$37,820. Which number should a manager plan next year against, and why is the close agreement misleading?
2. Eleven of twenty-seven separations happened inside four weeks, at a cost of \$12,460. Which of this chapter's steps is most likely to be failing, and which is least likely?
3. The reason column records "moved" and "school" twice each. Why should that make you less confident in the column rather than more?

(1: The ledger — because it is built from rates and can be re-priced when a wage, a headcount, or a channel changes, while a single year's log is one draw from a distribution. Twenty-seven events is a small sample, and agreement to within \$250 was produced by a sous-chef departure and an assistant-manager retention that happened to cancel each other. Use the log to correct the ledger's rates, not to replace it. 2: Most likely the offer-to-day-thirty sequence — §17.7's clarity about what the job is, and §17.8's first shift — because someone who leaves at week two or three has usually found a gap between what was described and what the job is, not discovered they cannot do it. Least likely the interview: an interview cannot fail this fast, because the failures it screens for do not surface in fourteen days. 3: Because they are the two socially costless answers. A person leaving over a schedule, a wage, or a manager will usually say "moved" rather than say the true thing to the person it is about — so the column is a floor on schedule-and-wage departures, never a count of them.)

17.10 What hiring cannot fix

Everything in this chapter has argued that hiring is a discipline with an arithmetic behind it, and that doing it deliberately is worth about \$38,070 a year of attention at Bellwether's size. All of that holds. Here is the limit on it, stated plainly — because a chapter that ends without one has sold you a formula and kept its failure modes.

Hiring cannot fix a schedule nobody can plan a life around. Read Figure 17.4's reason column again: second-job conflict. Closing shift ran too late. Warehouse job, fixed hours. Wanted full time; we had part. Weekend availability. Five of twenty-seven separations, in a column that understates. No posting, no scorecard, and no stage recovers a person from a schedule that will not let them see a child, hold a second job, or know on Monday what Saturday looks like. Chapter 19 owns the schedule. What belongs here is that the schedule is doing your hiring for you whether you intend it to or not.

Hiring cannot fix a wage below the market you are actually competing in — and the market is not the restaurant down the street. The dishwasher who left at week 24 for "a factory job with benefits" did not leave for another kitchen. The busser who took a warehouse job did not either. Your real competitive set is everything else a person with those hands and that availability can do on a Tuesday, and in most American cities that set is larger and better-paying than the industry likes to admit. A wage that is competitive against three other restaurants and uncompetitive against the distribution center will produce exactly the log above, indefinitely, no matter how good the interview is.

Hiring cannot fix a kitchen where the standard is not enforced. This one runs opposite to intuition, so it is worth saying carefully. A loose kitchen does not lose its weak cooks — they are comfortable. It loses the good ones, and it loses them for a reason §17.5 already identified: the cook who takes a correction well, who names their own mistake before you find it, who cleans as they go, is precisely the cook who will not stay somewhere that lets a plate leave at sixty percent. You cannot hire your way past a standard you are not holding. You will simply hire better cooks and lose them faster, which is the more expensive version of the same problem.

And hiring cannot fix a manager people are leaving. You will not see this in the log, because nobody writes it in the reason column — not to the person's face, and not on a form the person may read. You will see it as a position that turns over faster than the same position in the same building under a different manager, and as five separations in a row that all say "moved."

So, plainly, and this is the sentence to carry out of the chapter:

A restaurant with a hiring problem usually has a retention problem, and a restaurant with a retention problem usually has a management problem.

That is not a moral observation. It is a diagnostic sequence, and it runs in that order for a practical reason: hiring is the symptom that surfaces first, because it is the one with an invoice attached. Nobody notices a schedule quietly costing them five people a year. Everybody notices being short a line cook on the second Friday in October. The invoice arrives at the last stage of a process that failed three stages earlier, which is why so much management energy in this industry gets spent on the recruiting end of a problem that lives at the other end.

The test is simple. If you are hiring constantly, the pipeline is not the thing that is broken. The pipeline is the thing being asked to compensate.

Which is Chapter 21 — the chapter this one has been pointing at since §17.1, and the real reason the \$38,070 was worth computing. You do not price turnover in order to get better at replacing people. You price it to find out what you would have to change to stop.

🍽️ The Business Plan

Checkpoint 17 of 40 — the Staffing Plan.

The plan now has a menu, a kitchen, a beverage program, and a revenue forecast. It does not yet have anyone to execute them. This checkpoint specifies who Bellwether employs; Chapter 19 prices it against the \$500,000 labor line.

The organization.

BELLWETHER — ORGANIZATION AT OPENING                      [the Bellwether plan]

                      ┌────────────────────────┐
                      │  The two owner-partners │
                      └───────────┬─────────────┘
                    ┌─────────────┴──────────────┐
          ┌─────────▼─────────┐        ┌─────────▼──────────┐
          │  CHEF / PARTNER   │        │   GM / PARTNER     │
          │  (back of house)  │        │  (front of house)  │
          └─────────┬─────────┘        └─────────┬──────────┘
                    │                            │
              ┌─────▼─────┐              ┌───────▼────────┐
              │ SOUS CHEF │              │ ASST. MANAGER  │
              │ (salaried)│              │  (salaried)    │
              └─────┬─────┘              └───────┬────────┘
                    │                            │
      ┌─────────────┼──────────┐      ┌──────────┼───────────┬──────────┐
   ┌──▼───┐    ┌────▼───┐  ┌───▼──┐ ┌─▼──────┐ ┌─▼────────┐ ┌▼───────┐ ┌▼──────┐
   │ LINE │    │  PREP  │  │ DISH │ │SERVERS │ │BARTENDERS│ │ HOSTS  │ │BUSSER │
   │  × 4 │    │   × 2  │  │  × 3 │ │  × 8   │ │   × 3    │ │  × 3   │ │  × 4  │
   └──────┘    └────────┘  └──────┘ └────────┘ └──────────┘ └────────┘ └───────┘

Headcount at opening: 31 people, about 22 full-time equivalents.

The shape is driven by three constraints established earlier in the book. The kitchen line is four at peak, which sets the line-cook count. Service runs seven times a week — five dinners and two brunches — which is why the dishwasher and host counts look high for the seat count: those positions cannot be consolidated across services. And the hearth caps the kitchen near 144 covers, which means there is no volume upside that would justify a fifth line cook.

The hiring plan against turnover. At the rates in Figure 17.1, Bellwether should expect roughly 27 separations in year one — nearly its entire headcount — at a cost near \$38,000. The plan therefore commits to four things that cost little and are the countermeasure:

1. A referral bonus of \$300, half at hire and half at ninety days.
2. A standing bench of interviewed-but-not-hired candidates, contacted quarterly.
3. Paid stages for every hands-on position, budgeted at roughly \$88 each.
4. A thirty-day check for every hire, on the calendar before the start date.

What this checkpoint does not settle. It does not price any of it — Chapter 19 builds the labor model and confronts whether this organization can run at 32.3% of sales, and at the tighter figure the ramp demands after week thirteen. It does not address wage structure, the tip model, or scheduling law, all of which belong to Chapter 20. And it assumes the two partners work in the business full time and are compensated inside the labor line, which is an assumption worth flagging now and testing in Chapter 31.

Open questions carried forward:

1. Can 31 people at these positions be paid inside \$500,000 — including \$12,035 of workers' compensation? (Chapter 19)
2. What is the tip model, and is it lawful where Bellwether operates? (Chapter 20)
3. What happens to this organization on the Friday the grill cook does not come in? (Chapters 19 and 21)

Conclusion

Hiring is not an administrative task that interrupts running a restaurant. It is roughly a \$38,000-a-year line item at Bellwether's size — 2.5% of revenue, 7.6% of the labor line, more than half the annual debt service — and it is invisible, scattered across nine accounts under other names.

Making it visible is the first move. Everything after that is unglamorous: a description that states the wage and the schedule, sourcing that leans on referrals rather than boards, a scorecard used identically for every candidate, a paid stage that tests the hands and the correction, a reference process that is consistent enough to defend, and thirty days of deliberate attention to a new person during the window in which they decide whether to stay.

None of it is difficult. All of it competes for a manager's time against a broken dish machine and a delivery that came in short, which is exactly why it gets done badly, and why doing it merely adequately is a genuine competitive advantage in this industry.

Chapter 18 takes the people you have now hired and makes them competent — which is the difference between a staffing plan and a restaurant.

Key Terms

Turnover rate — separations in a period divided by average positions in that period, usually annualized. A useful headline and a poor diagnostic on its own, because it treats a two-week dishwasher and a four-year sous chef identically. (Ch. 17)

Cost of turnover — the fully loaded cost of replacing one departure: advertising, management time, the stage, orientation, unproductive training wages, the trainer's diverted attention, reduced output, learning-curve waste, and vacancy overtime. At Bellwether, about \$2,180 for a line cook. (Ch. 17)

Job description — the document stating what a position does, requires, pays, and reports to. The cheapest screening tool available, and a brand document read by every candidate. (Ch. 17)

Sourcing channels — the routes by which candidates reach you (referrals, schools, boards, walk-ins, direct approach, returning alumni), which differ in cost, volume, and retention. The cheapest generally produce the longest tenure. (Ch. 17)

Structured interview — the same predetermined questions, in the same order, scored against a defined rubric, for every candidate for a position. Substantially more predictive than an unstructured conversation. (Ch. 17)

Stage — a trial shift in a kitchen, from the French stagiaire, in which a candidate works so both sides can assess fit. Should be paid. (Ch. 17)

Working interview — the front-of-house equivalent of a stage. Same principle, same wage-and-hour analysis. (Ch. 17)

Reference check — contacting prior employers, whose single most informative question is "would you hire this person again?" (Ch. 17)

Background check — third-party screening of criminal, credit, or driving history. A consumer report under the Fair Credit Reporting Act, carrying disclosure, authorization, and adverse-action obligations. (Ch. 17)

Form I-9 — the federal employment-eligibility verification required for every employee: Section 1 by the first day, Section 2 within three business days, employee's choice of acceptable documents, retained separately from the personnel file. (Ch. 17)

At-will employment — the default U.S. rule that either party may end employment at any time for any lawful reason, subject to significant exceptions and easily undercut by careless offer-letter language. (Ch. 17)

Onboarding — the structured management of a new hire from offer to full productivity, concentrated on day one, week two, and day thirty — the three points at which people actually quit. (Ch. 17)

Spaced Review

1. Compute the cost of turnover for a bartender at \$2,400 a replacement, at a 50% annual rate across three positions. What is the annual figure, and where on the P&L does each component hide?
2. From Chapter 11: a restaurant's food cost runs three points over target and the chef blames portioning. Name two ways a hiring problem could produce that variance.
3. From Chapter 14: the Friday-night grill cook no-show cost \$116.70 in hard money. Using this chapter's arithmetic, what is the total cost of that event if the cook does not return?
4. From Chapter 9: Bellwether's pre-opening budget carries \$31,510 of labor before a dollar of revenue. What does §17.1's turnover arithmetic imply about hiring the opening team too early?
5. The recurring question: a manager proposes raising the line-cook starting wage by \$1.00 an hour to reduce turnover. Does this move prime cost, and in which direction? What would you need to measure to find out whether it was worth it?