How to open a restaurant step by step: the before and after of putting service on the critical path

Step 1 · Write the service model before you sign the lease

The first deliverable of an opening is not the kitchen layout, it is a six-to-eight-page document defining the service model: number of stations, covers per server per shift, service-step sequence, target time from appetizer to main, and the upselling script with its three ticket anchors. It is done when you can answer, with a number, how many tables each station covers and how many minutes it takes to clear and reset a table of four. That document then governs payroll, layout and even bar size, which is why it comes before the signature. A 90-seat room that turns tables 2.4 times per service instead of 1.8 moves 33% more covers on the SAME rent and the same fixed payroll. Verification: two people on your team read the document separately and draw the same station map. If they draw two different maps, the model does not exist yet.

Step 2 · Certify the floor team in a simulator from week −8

Train the dining room before the kitchen is ready, and here I was wrong for years, putting training in week −2, after the equipment test, arguing that until there is food there is nothing to serve. That is false. Service steps, reading a table, handling a complaint, closing the check and resetting can all be trained in a simulator without a single burner lit, and whoever does it reaches day one with a team that already made its mistakes where mistakes cost no cash. The deliverable is a roster with name, assigned station and certification score by module; the checkpoint is 100% of the floor team above 85 points in service sequence and upselling. With hospitality turnover at 79.6% during 2025 according to the National Restaurant Association, that simulator also covers your month-four replacement, when nobody has time left to train by hand.

Step 3 · Soft opening with data, not with friends

A soft opening exists to measure, not to show off, and the difference is whether you walk out of it with four numbers written down: average time from appetizer to main, real average ticket against projection, share of tables with an incident, and covers per hour per server. Run it at controlled capacity —between 55% and 70% of seating, three or four services— with one observer who does not serve and only times and records. The deliverable is the timing sheet by station with three bottlenecks identified and an owner for each. Checkpoint: appetizer-to-main under 22 minutes on 90% of tables. The Bogotá group in this case skipped the step, sent four tables away on opening night over waits above 34 minutes, and watched the group's Google average fall from 4.6 to 4.2; recovery took five months. Five months of reviews against three measured services.

Step 4 · Week 1 with automated preshift and a daily board

Week one is won with a fifteen-minute preshift every day and a three-number board on the kitchen wall. The preshift is not a pep talk: it is the dish of the day, the ticket target, yesterday's incident and the station assignment, written and sent the night before so the shift lead is not improvising at 11:40. The board carries covers, average ticket and average service time, updated at close. The deliverable is seven documented preshifts and seven closings with those three figures; the checkpoint is that the average ticket on day seven is not below day one, because a drop there almost always means upselling was dropped the moment pressure arrived. Diego F. Parra has run openings in 43 countries with this same format, and the pattern repeats identically in a 40-seat room and in the fifth unit of a chain. The most expensive mistake is cutting training, because it is the only budget line with no vendor demanding a deposit: construction bills you, equipment bills you, licenses bill you, and practice hours can be postponed with nobody calling.

The five mistakes that sink an orderly opening

You pay for them anyway, with real customers. The other four: hiring the floor in week −2 and calling that a schedule; opening at full capacity on day one to «ride the buzz»; leaving each soft-opening bottleneck without an owner, so the timing sheet gets filed away; and launching a 60-item menu before the kitchen has run one full service, which blows up ticket times and scrambles inventory. An opening menu stays between 22 and 28 dishes. If you hesitate between training and decorating, train: a guest forgives a bare wall, a guest does not forgive 34 minutes of waiting with an empty table in front. Here is the sequence.

What happens if the kitchen is ready and the floor improvises?

For the first three days sales run high because curiosity does not depend on service, and the owner reads that number as validation; by week two the first week's reviews land and reservations drop between 15% and 25%;

by month two table turnover freezes wherever it stalled, since nobody retrains a room that is already operating, and with staff turnover at 79.6% (National Restaurant Association, 2025) by month five you have half a new team learning from a team that never learned properly either. That is broken scaling: it did not break in the kitchen, it broke on the floor, and fixing it costs twice what doing it early would have. The tension is genuine —training burns paid hours before you invoice a single peso, and opening cash flow is the tightest there is— and the bridge is moving practice into the simulator, where the hour costs a fraction.

Why order weighs more than budget?

Two openings with identical capital and different order do not produce identical results, and the industry proves it at scale:

franchised restaurants grew through system, not spend —the franchise model added 210,000 new jobs in 2025 (+2.4%) according to the IFA Economic Outlook 2025, and FRANdata projects 845,000 franchised establishments in the United States for 2026, up 1.5% from 832,521. Nobody replicates 800,000 units on individual talent; you replicate with a written sequence, standardized training and daily measurement. Spain shows the same pattern: 390 brands and 7,967 franchised restaurant establishments in 2024 (Tormo Franquicias Consulting). What franchising understood thirty years ago —order is the asset— is what an independent group still renegotiates at every opening. Budget sets the ceiling of your room; order decides whether you reach it in month three or in month fourteen. Your opening is healthy when you can tick six boxes without hunting for paperwork.

Closing checklist: how to know everything landed

One: the service-model document exists and two people draw the same station map. Two: 100% of the floor is certified above 85 points, with a signed roster. Three: the soft-opening timing sheet lists three bottlenecks with an owner and a date. Four: appetizer-to-main runs under 22 minutes on 9 out of 10 tables. Five: seven documented preshifts and seven closings with covers, ticket and average time. Six: the day-seven average ticket matches or beats day one. If one box is missing, do not open another location yet; if three are missing, you have an operation that depends on your best server never getting sick. Start today with box one: sit down for two hours and write how many covers each station handles. Expansion CapEx is not recovered with pretty construction, it is recovered with table velocity. A 90-seat room turning 2.4 times per service instead of 1.8 moves 33% more covers on the SAME rent, the same fixed payroll and the same kitchen brigade.

The differences that decide whether the unit survives year two

The gap between 1.8 and 2.4 almost never lives in the kitchen: it lives in how long a server takes to clear, close out and reset, and that is either trained or it is not. Here is a tension few operators resolve well: training costs paid hours before a single peso is billed, and opening cash flow is the tightest money there is. So the owner cuts training —the only budget line with no vendor demanding a deposit. The bridge is the simulator: it moves most practice hours into an environment with no food cost, no dedicated supervisor and no review risk, leaving floor hours for what can only be learned standing up. In unit economics the floor shows up twice and most people count it once. It shows up as payroll cost —between 26% and 32% of sales in table service— and as a revenue lever through average check and turns.

The differences that decide whether the unit survives year two — in practice

An owner who reads it only as cost optimizes by cutting people, drops the service level, loses check size and lands on a worse margin. That is the trap: the saving appears in month one and the loss surfaces in month five, when it gets blamed on the market. On menus: MASTERESTAURANT always recommends keeping the printed menu alongside the QR. The printed menu controls the experience —it sets service rhythm, carries the menu narrative, enables suggestive selling when a server points at a dish— while the QR is a complement for delivery, accessibility, price changes and analytics on what gets viewed without being ordered. Opening QR-only saves roughly 900 dollars in printing and costs average-check points every single night. Serious investor due diligence no longer stops at a projected P&L. It asks about team retention, about ramp time for a new server, and whether the service standard is documented or lives in the manager's head.

The differences that decide whether the unit survives year two — key points

A certification dashboard with percentages per person and station turns a promise into a verifiable asset, and that moves valuation when you go looking for capital for unit number three. MTIE —internal margin of tolerated error— is the number separating a trained operation from one that merely survives. It defines how many errors per hundred tables you accept before intervening. Unmeasured, the real MTIE of a traditional opening runs 18 to 24 errors per hundred tables in week one; with a certified team it drops to 6 or 8, and that gap is precisely what customers post on Google.