When orders back up, the delay may be in production, an unchecked ticket, missing packaging or a finished order waiting for handover. Each calls for a different response. Observe where work waits before deciding whether the kitchen needs another appliance, a different assignment or a clearer handoff.
This guide shows how to observe order flow, identify the current constraint and test one controlled operating change. Use the kitchen layout guide alongside it when the evidence points to a physical space or route problem. Every number below is an original simplified illustration, not a benchmark for a real kitchen.
The method in brief
Map the steps and queues, then estimate each resource’s capacity against the work it must do. In a simple stable serial process, the limiting step sets the theoretical throughput ceiling; increasing another step’s capacity alone does not raise that ceiling. Real kitchens also face interruptions and variation, so changes elsewhere can help by preventing the limiting step from being starved or blocked. Recheck the whole flow after a change.
Applying this in a kitchen needs care, because people share tasks, orders vary, and food safety controls cannot be traded for speed.
Separate work from waiting
Separate active work from waiting before a task and waiting after an order is ready. Tag rework separately: a correction can involve both active work and waiting, so it is not a fourth duration to add without checking overlap. Record what an observer actually sees rather than inferring every delay from the ticket system.
Each state has a different cause and a different fix. A long total time with little active work points to queues or handover problems, not slow cooking. Record when the order was promised as well as when it actually reached the guest, so you can see the gap that matters to them rather than only the time the kitchen was busy.
Bound the observation
Compare like with like. Observe one channel, such as counter or delivery, one menu and one service window at a time, and choose cases that represent normal trading rather than the single worst night. Count order units and items separately, because one order with eight items behaves differently from eight single-item orders, and time large batches separately from individual orders.
Record the order mix, interruptions, stock shortages and applicable food-safety controls. Explain the observation to the team and collect only necessary operational information. Use anonymised ticket references and task timings; guest names, phone numbers and other identifying details are not needed for this review.
- Do not treat one person covering two tasks as two separate capacity resources. Add their workload together and compare it with their available time.
- A shorter queue at one stage may simply move the delay further along. Retest the full flow after any change.
- Do not hide waiting by marking orders ready early, and do not quietly narrow the menu without telling guests.
Illustrative example: three stages
Imagine a simplified process with three serial stages, each performed by one dedicated person working on one order unit at a time, with no rework, breaks or changeovers. Capacity per hour is 3,600 seconds divided by the seconds per unit.
Under these assumptions the ideal flow capacity is limited to 80 orders per hour by assembly and verification. That is a ceiling, not a guaranteed actual throughput. If 30 orders arrive in 15 minutes, the arrival rate is 120 per hour. Even if assembly works for the entire 15 minutes, it can finish at most 20 orders, so at least 10 will not have passed assembly by the end of that window. Start-up time and downstream delays can make completed handovers slower still, so treat this as a direction, not an exact queue prediction.
| Stage (illustrative) | Seconds per order unit | People dedicated | Capacity per hour |
|---|---|---|---|
| Order check | 18 | 1 | 200 |
| Assembly and verification | 45 | 1 | 80 |
| Packing and handover | 30 | 1 | 120 |
What changes if the constraint improves
Suppose an approved trial brings assembly and verification to 30 seconds per unit while every check is still carried out. Its capacity becomes 120 per hour, tied with packing and handover. That is a hypothetical result for this example, not a promised saving.
At that point there are two constraints, and any variation at either stage will show quickly. Re-evaluate the other stages, the people who are shared between tasks and the variation in the order mix before deciding what to try next.
Under the same assumptions, a single order with no queue would receive 18 plus 45 plus 30, which is 93 seconds of active work. That figure describes the work content only. It does not predict how long a guest will wait, which depends on queues and handover.
Reading real ticket times
Stage observations help estimate capacity under stated assumptions. Ticket timestamps show elapsed intervals: record minutes from a common observation start when an order was accepted, work started, it was ready and it was handed over. The start-to-ready interval is preparation elapsed time; it may include active work, internal waiting and rework. A separate observation is needed to distinguish those. These sample tickets contain no real customer data.
These tickets are a separate illustration from the three-stage capacity example. Their four-minute preparation intervals are not measures of four minutes of active work and should not be compared directly with the earlier 93 seconds. Missing or out-of-order timestamps leave a ticket unresolved.
| Ticket (illustrative) | Accepted | Work started | Ready | Handed over | Waiting | Preparation elapsed | Handover delay | Total |
|---|---|---|---|---|---|---|---|---|
| A | 0 | 4 | 8 | 10 | 4 | 4 | 2 | 10 |
| B | 2 | 3 | 7 | 8 | 1 | 4 | 1 | 6 |
Trial one controlled change
Pick one change aimed at the current constraint, agree it with whoever is responsible for food safety, and trial it for a defined period. Useful examples include staging approved packaging and labels before service, a clearer handoff point, an assigned expeditor or changing the sequence in which orders are released to the line.
Food safety and service quality are independent gates. A change that improves speed but bypasses a holding check, reduces verification or alters approved preparation and storage controls fails the trial, whatever the timings show.
- Record the baseline, the change, its owner, the trial period and the criteria for success.
- Retest the whole flow, not just the stage you changed.
- Keep, adjust or reverse the change, and log the follow-up date.
When the constraint is physical
Sometimes the observation shows that the limit is genuinely physical: not enough bench space at the bottleneck, a crossing route or a station that cannot be staffed without blocking another. Then the evidence you collected becomes the basis for a design conversation, rather than a guess about what to buy.
Put it into practice
Your action checklist
Your progress saves in this browser. No sign-up needed.
0 of 8 complete
From reading to doing
Kitchen workflow observation
The Workflow Review worksheet records stage times in seconds per observed unit and the number of dedicated people at each stage, then calculates capacity per hour and the limiting stage. A separate ticket log takes accepted, started, ready and handed-over times in minutes from a common start and calculates waiting, preparation elapsed time, handover delay and total. Missing or reversed times stay unresolved. An action log tracks the baseline, change, owner, trial criteria and follow-up.
Open the Workflow Review tab in the launch workbook. Keep a copy for your own business.
Download the free workbook See what’s inside the toolkit →Sources & further reading
- Retrospective: Introduction to Operations Management, 15.760 Operations Management, Spring 2002 (MIT OpenCourseWare) ↗
Archived lecture, retrieved 7 November 2026. Background on process flow maps, capacity and queue variables, and bottleneck management. The kitchen examples and numbers in this guide are original illustrations, not benchmarks from the lecture.
Published by Keystonne Grow. Product links connect to the Keystonne ecosystem; check the linked product or service for its current scope.
