Scheduling Around the Bottleneck Machine: Drum-Buffer-Rope for Job Shops
An hour lost at the bottleneck is an hour lost for the whole shop. Here is how to identify the constraint machine and build the week's schedule around it with drum-buffer-rope.

When Every Machine Is Busy and the Shipment Is Still Late
It is 6:45 on a Tuesday morning and the traveler stack on the supervisor's desk looks healthy. The saw is caught up. Both lathes are running. The deburr bench is empty, with two people standing at it waiting for work. Meanwhile the 5-axis mill, the only machine in the building that can cut the housings due Thursday, has eleven jobs parked in front of it, and its first-shift operator lost forty minutes yesterday hunting for a fixture.
By ten o'clock the customer service rep will call about the housings. By noon someone will carry a rush order to the front of the mill's queue and push two other jobs back without telling anyone. The shop is not short of effort. It is short of hours on exactly one machine, and nothing in the schedule treats that machine any differently from the rest.
This article shows how to do bottleneck machine scheduling with drum-buffer-rope: find that machine, schedule it first, protect it with a buffer, and release work to the rest of the shop at its pace.
One Machine Sets the Output of the Whole Shop
Drum-buffer-rope is the scheduling method from Eliyahu Goldratt's theory of constraints. It rests on one observation: in any routing, one resource has the least spare capacity relative to the work that wants to flow through it. That resource is the constraint. Every other resource can only ship what the constraint lets through.
The consequence is lopsided. An hour lost at the constraint is an hour of output lost for the whole shop, because nothing downstream can make it up. An hour lost at a machine with slack costs nothing, because that machine has spare hours to recover it. A saw idle for an hour delays nothing if the mill is the limit. A mill idle for an hour delays everything.
Theory of constraints teaches five focusing steps:
- Identify the constraint.
- Exploit it: make sure it never waits for material, an operator, or a program, and never runs work that is not needed.
- Subordinate everything else to it: schedule and release the other machines to serve the constraint's plan.
- Elevate it: only after steps 1–3, add capacity, cut its setups, or move work off it.
- Repeat, because the constraint may move.
Most job shops jump to step 4, buying a machine or adding a shift, before doing step 2. Drum-buffer-rope is the mechanism for steps 2 and 3.
How to Find the Bottleneck Machine
Start with arithmetic rather than opinion, because finding the bottleneck machine is a load calculation first. For each machine, add up the hours of work that the open jobs will demand from it over the planning window, then divide by the hours that machine can actually be available in that window after shifts, planned maintenance, and known downtime. The machine with the highest ratio is your candidate. For the available-hours side, the machine capacity planning guide covers it, and the machine utilization rate calculator shows how to measure what a machine actually delivers against what it could.
Then check the candidate against the floor. A real constraint leaves fingerprints:
- A queue that never clears. Jobs wait in front of it, and the queue is there on Monday and still there on Friday.
- Starved downstream stations. The operations after it keep waiting for parts while the operations before it pile up WIP.
- Expediting that always points at it. Rush orders and overtime requests land on the same machine.
A caution on measurement: utilization alone can fool you. Per OEE-Benchmark.org (2026), typical OEE in machine shops and metal fabrication runs 45–65%, and setup and idling are the dominant losses. A constraint can look modest on a utilization report because it spends its hours in changeovers and waiting, which is precisely why load against available hours, and the physical queue, are the better tests.
The constraint also depends on the job mix. A shop limited by the 5-axis mill when housings dominate may be limited by inspection when a different customer's work comes through. That is the reason step 5 says repeat.
Bottleneck Machine Scheduling: Drum, Buffer, and Rope
The drum is the constraint's schedule, and it sets the beat for everything else. Build it before you schedule anything else.
- Sequence the constraint's work first. Take the jobs that need the constraint, order them by due date and priority, and lay them end to end on that machine's calendar, honoring its real available hours.
- Group setups where you can. Changeover time on the constraint is output you never get back, so sequence similar jobs together when due dates allow. The finite capacity scheduling explainer covers why one job per machine window is the rule that makes this honest.
- Do not overload the drum. If the committed hours exceed the available hours, the schedule is fiction. Either some jobs move to the following week, or you decide deliberately what to do about the shortfall.
- Schedule everything else around it. Machines before the constraint must finish each job in time to reach it. Machines after it must be ready to take each job when it comes off.
This reverses the usual habit. Instead of loading every machine to the brim and hoping the work lines up, you commit the constraint's hours with care and let the other machines absorb the variation. Bottleneck machine scheduling treats the other machines as support for one machine's calendar.
Because an hour lost at the constraint is lost for good, the constraint must never run out of work. That is the buffer's job.
The buffer is a block of time, not extra inventory. You release a job so that it arrives at the constraint ahead of its scheduled slot by a margin, the constraint buffer. If a drill press breaks down upstream or an operator is out, the buffered job still shows up in time and the constraint keeps cutting.
The rope is the release rule that ties the front of the shop to the drum. Raw material is released to the first operation only when the drum says it is needed: the constraint's slot minus the buffer minus the lead time of the operations in between. Release earlier and you build WIP that clogs the floor. Release later and the buffer is gone.
The rope is the part shops resist, because it means letting an able saw sit idle when it could be cutting bars for jobs not needed yet. Parts the constraint cannot use this week add no shipments and only lengthen the queue.
The constraint sets the beat, the buffer protects the beat, and the rope keeps everyone else from running ahead of it.
A Worked Example With Round Numbers
This is an illustration with invented round inputs, not a measurement from any real shop. Suppose a shop runs two shifts, five days a week, so each machine has 80 available hours next week after planned maintenance is removed. The open jobs demand the following hours:
| Machine | Hours demanded | Available | Load |
|---|---|---|---|
| Saw | 40 | 80 | 50% |
| Lathe | 62 | 80 | 77.5% |
| 5-axis mill | 91 | 80 | 113.75% |
| Deburr | 35 | 80 | 43.75% |
| Inspection | 48 | 80 | 60% |
The arithmetic points at the 5-axis mill: 91 ÷ 80 = 1.1375, so it is committed beyond its capacity by 11 hours. No other machine reaches 100%. The mill is the drum.
Now apply the rules.
Exploit it. Eleven hours do not fit. The scheduler decides on purpose which jobs slip a week, by due date and customer priority, instead of letting the floor decide by shouting.
Buffer it. Say the shop sets the constraint buffer at 16 working hours, one two-shift day. Every job destined for the mill should be sitting in front of it a day before its slot.
Rope it. If the saw and lathe operations ahead of the mill take about two days of elapsed time for a job, then material for that job is released three days before its mill slot: two days of routing plus one day of buffer. The saw does not cut bars for jobs scheduled at the mill two weeks out.
Count the hours that matter. If deburr loses an hour to a broken cart, the shop loses nothing, because deburr has 45 spare hours. If the mill loses an hour waiting for a fixture, the shop loses an hour of shippable output. This is bottleneck machine scheduling in one comparison.
Then elevate. Suppose the mill spends 8 hours of that week in changeovers. Structured SMED work typically cuts changeover time by 30–50% on a first pass, according to iFactory and Oxmaint (2025). Taking the low end, 30% of 8 hours is 2.4 hours handed back to the constraint each week. Those hours count in a way 2.4 hours at the saw would not.
For a mold and die shop whose constraint is a single EDM machine, the same pattern applies, and the EDM bottleneck scheduling article works through it.
Where Drum-Buffer-Rope Breaks Down
The method is simple to describe and hard to hold, for three practical reasons.
The constraint moves and the schedule does not. A new customer, a different mix, or a machine going down changes where the limit sits. A drum rebuilt only on Mondays is wrong by Wednesday.
The rush order problem. A rush job that needs the constraint has to displace something already on the drum, and every displaced job has a due date. Do the displacement on paper, with the consequences visible, before you make the promise.
Manual upkeep. Maintaining a drum by hand means recomputing finish times for every job behind a change, and that is where whiteboards and spreadsheets quietly stop being trusted. A scheduling board that enforces finite capacity and recalculates when you drag a job helps here. MachineScheduler's finite-capacity enforcement and drag-and-drop rescheduling are available on Professional and above, but the method works on a spreadsheet if you are disciplined about the load table and the release dates.
Whatever tool you use, the method only pays off if the buffer is respected. A buffer raided to "keep the saw busy" is just WIP by another name.
Next Step: Put Your Own Numbers in the Load Table
The five-machine table above is the whole first step. With real numbers, list the open jobs, total the hours by machine, divide by available hours, and see which row crosses 100%. The Machine Capacity Planning Workbook is built to quantify your scheduling waste and build the ROI case for a scheduling system, with a capacity dashboard, a cost estimator, and a one-page summary. You can download the Machine Capacity Planning Workbook and use the capacity dashboard to see where your hours go before deciding what to change.
Ready to go beyond the guide?
Most shops are on a live Gantt board within 60 minutes of sign-up, with their existing job list imported from Excel.
Get shop floor scheduling guides in your inbox
Practical articles for production managers — no spam, unsubscribe anytime.
Related articles
CNC Setup Time Reduction: A SMED Walkthrough for Job Shops
In a high-mix shop, setups eat the capacity the schedule thinks it has. This walkthrough applies SMED to CNC changeovers…
Preventive Maintenance Scheduling in a Machine Shop: Treating PM as a Capacity Block
Preventive maintenance that lives on a separate calendar collides with the production schedule. This guide shows how to …
Machine Utilization Benchmarks by Machine Type: Mills, Lathes, and EDM
A shop-wide utilization number flattens out machines that behave nothing alike. Here's how mills, lathes, and EDM typica…