Systems vs hiring more staff: the numbers that decide for you

When does hiring one more person stop working?

It stops working at the exact moment person number N+1 fails to move your plates-per-labor-hour figure, and that point arrives earlier than almost any owner calculates.

The Bogotá steakhouse in this case added two kitchen assistants and pushed payroll up 11.4 percentage points without moving average check or Friday delays, because the constraint was an oven, not a pair of hands. That arithmetic hurts more now: according to Toast, payroll already absorbs more than 26% of revenue in the United States, up from roughly 23% in 2021, so every wrong hire weighs three points heavier than four years ago. Measure it yourself: count plates fired during the two peak hours, divide by effective labor hours in that window, and repeat the count four weeks after the hire. If the number does not climb at least 8%, you bought perpetual fixed cost without buying a gram of capacity.

Labor cost is rising for everyone and that changes the equation

Eighty-eight percent of operators reported higher labor spending during 2024, according to the National Restaurant Association, and that figure reorders the whole decision because it turns a once-negotiable variable into structural pressure. When 88% of the market faces the same increase, hiring better stops being a competitive edge: everyone pays more for the same profile. The same source records that 57% of operators ran short-staffed by more than 10% of their headcount that year, which means not even paying above market guarantees you shift coverage. Here is the concrete call: if annual turnover clears 60% and payroll passed 26% of revenue, the money behind your third hire of the year returns more buying sequence —station timing, measured mise en place, tickets sorted by cook time— than buying one more body who will walk out in eight months. Some functions always belong to a person, and the clearest one touches food safety, because there the cost of an error is not measured in margin but in lives.

Where hiring still wins and no dashboard replaces it?

Food Safety Magazine documented that deaths tied to recall-linked outbreaks went from 8 in 2023 to 19 in 2024, more than double in twelve months.

No camera judges whether that hake holds up for tomorrow's service; no algorithm reads the face of the guest at table 12 when the plate came out wrong. Dock receiving, core temperature reading, a head chef's judgment on a protein at its limit: that is trained instinct, and automating it is expensive fraud. The practical rule I apply with clients: if the task demands judgment over perishable product or over an upset human being, hire and pay well; if the task is repetitive, predictable and measurable, systematize it and stop arguing. A new line cook reaches service speed in about eight weeks and carries that entire learning out the door the day he resigns, while a process written inside the system survives every departure.

Turnover takes the training; the written process stays

That asymmetry defines the economics of this sector: with 57% of operators understaffed in 2024 per the National Restaurant Association, you are paying the learning curve several times a year for the same station. Run the full cash math. Eight weeks of training at half productivity, on a hot-line wage, plus the head chef's supervision time, plus waste during the first three weeks: that lands somewhere between 1.5 and 2 monthly salaries per hire. Repeat it four times a year and you are quietly financing the equivalent of a recipe-and-station management system that never resigns, never calls in sick and never asks for December off. Roughly 75% of restaurant traffic now happens outside the dining room, according to Circana, and that share tilts the balance toward systems because delivery is a coordination problem, not a manpower problem. In limited service, the off-premise mix hit 83% in 2024 against 76% in 2019, per the National Restaurant Association: seven points of migration in five years.

Off-premise traffic already settled this argument for you

A delivery order needs synchrony across ticket, cook time, pickup window and courier status, and none of those four improves by putting another cook in front of the flat top. If more than half your tickets leave through the door, your next investment should pull the platforms onto one screen before adding a station. Hiring solves capacity; the system solves coordination, and confusing the two against 75% external traffic costs you the entire shift. Up to 20% of reservations end as no-shows in the United States and Canadá according to OpenTable, and across the United Kingdom, Australia and New Zealand the band runs from 15% to 18%, numbers no host solves with phone calls. The same source reports that platform-booked reservations show roughly 20% fewer no-shows than those taken by phone. Translate that into cash.

Reservations: one case where the system pays for itself

A 120-seat dining room turning twice on Friday with a $30 check moves $7,200 on a full night; a 20% no-show rate takes $1,440 off that, and trimming that rate by a fifth hands back about $288 per Friday, somewhere near $15,000 a year. That exceeds the annual subscription of nearly any serious reservation system, and it requires hiring nobody and expanding the room by not one meter. The three scenarios split by operational maturity, not by size, and the dividing line is simple: do written processes exist before you buy the tool? Small restaurant, one location, fewer than 15 employees: forget integrated management software and first write your standardized recipes and station mise en place; with 96% of Mexico's restaurant sector made up of microenterprises according to INEGI and CANIRAC, this is the region's majority scenario and the one losing most money on platforms nobody feeds.

How to read these numbers in YOUR operation?

Mid-size, two locations or 15 to 40 employees: the lever here is a point of sale with daily inventory control, because leakage shows up between locations.

Group, three outlets and up: labor cost is linear and perpetual while system cost is stepped and decreasing per location, so from the third one the gap turns structural and hiring better no longer closes it. These figures come from four public sources —National Restaurant Association, Circana, OpenTable, Toast and INEGI/CANIRAC— and you deserve to know their limits before making a payroll decision with them. The labor cost and staffing shortfall data describe the 2024 United States market; Toast's 26% payroll-to-revenue figure comes from its own client base, which over-represents already digitized operations. OpenTable's no-show numbers come from restaurants with managed reservations, not from the full universe. The only Latin American figure here is the INEGI and CANIRAC count on Mexican microenterprises.

Where these benchmarks come from and what they don't tell you?

At Masterestaurant, Diego F. Parra treats these benchmarks as directional reference, never as a target: the number that governs is yours, measured in your own register across four consecutive weeks.

Start this week by counting plates per labor hour at the Friday peak. Hiring solves capacity; systems solve coordination. Mixing them up is expensive: if three cooks already collide at the same flat top, the fourth cook makes the number worse. Labor cost is linear and perpetual; system cost is stepped and decreasing per unit. From the third location onward that gap turns structural and no hiring strategy closes it. A new hire learns in 8 weeks and takes that learning along when resigning; a written process inside the system survives the 79% turnover that defines the sector in 2026. Systems fail exactly where people shine: no camera judges whether a protein survives tomorrow's service, and no algorithm reads the face of the guest at table 12 when something went wrong.

Where the two paths genuinely split?

Hiring reverses at a high legal and human cost; switching off a badly chosen system costs one month of license and an afternoon of data export.

That risk asymmetry rarely enters the conversation and it should. Owner-independent operation exists only when judgment is written down. No amount of staffing produces independence if decisions still climb through WhatsApp at eleven at night.