Kitchen ticket time control: before vs after with Masterestaurant

Step 1: split the clock by station before you touch anything else

Kitchen timing control starts by breaking ticket time into four independent clocks —cold station, grill, fryer and pass— because the global average hides the exact number you need. If the appetizer leaves in 4 minutes and the main in 19, an 11.5-minute average declares everything fine while the table stares at empty plates for twenty. What this step delivers is a sheet with four columns holding two weeks of point-of-sale history, station by station, time band by time band. Verify it this way: anyone on the team should be able to tell you, without opening a report, how long the grill takes between 8:30 and 9:30 pm. One Bogotá restaurant running 140 tickets per service found through this single split that its delay lived at the pass, with nobody assigned during peak, and fixed it by moving one person for twenty minutes. A written ceiling is what turns a complaint into a measurable deviation, and without it AI has nothing to watch.

Step 2: write the ceiling in minutes, station by station, and post it on the wall

Set the maximum yourself: cold station 5 minutes, grill 12, fryer 8, pass 3 from the moment the plate is ready until it leaves the kitchen. Those figures are not universal —they depend on your menu and your equipment— but they must exist in writing before anyone measures a thing. The deliverable is an A4 sign at every station with its number, plus the same number loaded into the ticket system so it triggers the alert. Verification is brutally simple: ask three different cooks what their ceiling is; if you get three answers, the step is not done. Some 62% of independent restaurants never measure a single pass time, almost always because nobody wrote the number down first. An 18-minute ticket may have spent 12 waiting at the printer and 6 actually cooking, and that distinction decides whether you have a kitchen problem or a staffing one.

Step 3: tell queue apart from execution or you will blame the wrong cook

Pushing the grill cook when the bottleneck sits in the inbound queue burns the team and moves the clock not one minute, made worse because hourly turnover in full service already runs at 96% (Black Box Intelligence 2024) and every departure costs you weeks of learning curve. Deliverable: two timestamps per ticket, print time and real start of preparation, even if the cook marks the start with a button. Check it by subtracting both columns across three consecutive services. When queue waiting exceeds 30% of total time, your problem is not at the stoves and no amount of training will fix it. A kitchen screen corrects behavior that no head chef achieves by shouting, and that is where the Masterestaurant method moves from data to operation. Diego F. Parra has spent twenty years walking into kitchens with the same question —how long does your signature dish take— and almost always gets the same empty answer; once that number sits on a monitor with the ceiling beside it, tickets turning amber at eight minutes and red at twelve, the team regulates itself without supervision.

Step 4: put the clock where the team sees it, not in your Monday report

The deliverable is one screen per station, or at minimum a shared one whose four columns read clearly from any position. Verify it by standing in the corner through a full service: if nobody looks at the screen in forty minutes, it is badly placed or the colors mean nothing to them. Automating the watch only makes sense after the previous four steps, because an algorithm that does not know your limit cannot warn you about anything. With the ceiling loaded, the system reviews 400 tickets per service without tiring, crosses each deviation against the day of the week and tells you that on Thursdays between 9:00 and 9:40 pm the fryer runs 40% above its ceiling, a pattern no human eye catches. In 2026 that layer costs less than one hour of cook overtime per month. The deliverable is an automatic alert to the manager's phone whenever three consecutive tickets from the same station break the ceiling, not a weekly report.

Step 5: AI belongs here, once the ceiling is written and not before

Test it by forcing the situation on a quiet Tuesday: if the alert takes longer than two minutes, your threshold is misconfigured. The costliest mistake is measuring the average and celebrating it: a 14-minute mean ticket time with 20% of orders above 25 is still a dining room full of irritated guests, so always read the 90th percentile alongside the mean. Second comes using the clock to punish; the moment your team smells sanctions, it learns to mark plates ready before they leave and you lose the data permanently. Third is measuring only at peak —the 8:00 pm bottleneck is not the 10:30 pm one— and fourth, changing the menu without recalculating the affected station's ceiling. Add that 20% of early departures trace back to poor onboarding (meez 2025): a new cook who does not understand the screen on day one will not use it on day two.

Four mistakes that wreck the rollout, and how to dodge them

Measurement is information, never discipline. Suppose you ignore the station and keep the aggregate number: the average would indeed drop, because the team learns to fire the easy items fast —salads, fridge desserts— while the expensive main lags behind, so ticket time improves on paper as margin erodes. By month three you would see more cold plates sent back, since the appetizer leaves out of sync and waits at the pass; by month six, the manager would be convinced the kitchen runs well and the trouble sits in the dining room. There lies the paradox of this trade: optimizing the indicator the guest sees, without opening the station that produces it, worsens the very thing you set out to fix. The only way out is holding both readings at once, the aggregate for the floor and the station for the kitchen, and deciding always with the second. You will know timing control is in place when you can tick six items with nobody's help.

Closing checklist: how to know it all landed

One: four written ceilings exist and three cooks picked at random give you the same number. Two: every ticket records print time and start time, and the gap reads off a single column. Three: a screen is visible from each station with the colors working. Four: the automatic alert reached the manager's phone at least once last week and somebody acted on it. Five: you review the 90th percentile and not just the mean. Six: average ticket time on an 80-order service dropped from 22 to 14 minutes within three or four weeks without hiring anyone. If item six fails while the other five hold, do not roll anything back: the bottleneck is equipment or menu, and that is a different conversation. Measure by station, not by ticket. Global ticket time hides the problem: if the starter leaves in 4 minutes and the main in 19, an 11.5 average says everything is fine while the table stares at empty plates for twenty minutes.

Three differences that decide the outcome

Split cold line, grill, fryer and pass and the real bottleneck surfaces, almost never the one you suspected. A Bogotá venue running 140 orders per service found its delay lived at the pass — nobody assigned there during peak — and fixed it by moving one person for twenty minutes, hiring nobody. Separate queue from execution. An 18-minute order may have spent 12 minutes waiting at the printer and 6 actually cooking. Pushing the cook in that scenario burns money and burns the team, because the fault sits in the sequencing rather than in anyone's hands. Decent KDS platforms have logged both timestamps since 2023, yet according to Jon Taffer, founder of Bar Rescue and president of the Nightclub & Bar Media Group, most operators buy the screen and never open the report that justifies it. That is where 90% of the return evaporates. Marginal efficiency beats total efficiency. Do not chase every minute: chase the expensive one.

Three differences that decide the outcome — in practice

Cutting from 22 to 18 minutes on a plate worth 4% of sales changes little; shaving two minutes off the plate worth 28% frees real tables and lifts turn rate. I got this wrong for years, pushing even improvements across the whole menu because it felt fairer, until the till taught me that operational fairness is measured in contribution margin instead of equal shares.