Drum buffer rope (DBR) is a production scheduling method from the Theory of Constraints that sets the pace of a whole plant at its constraint. The drum is the constraint's schedule, the buffer is a block of protective time placed ahead of the constraint (and ahead of shipping), and the rope is the rule that releases raw material only when the constraint's schedule calls for it. The goal is a line where the constraint never waits for work and no other station buries it in inventory.
The Theory of Constraints explains why a system's output equals its constraint's output. This article covers execution: a schedule, a buffer and a release rule a shop floor can follow on Monday. For sizing and monitoring the buffer in more depth, see buffer management in the Theory of Constraints.
Where drum buffer rope comes from
Eliyahu Goldratt brought the Theory of Constraints to a wide audience in his 1984 novel *The Goal*. Plant manager Alex Rogo has three months to rescue a failing plant, and the book's best-known scene is a scout hike. The troop can only move as fast as its slowest scout, a boy named Herbie. Faster scouts cannot get the troop to camp sooner, they only stretch the gap between themselves and him. Alex puts Herbie at the front and lightens his pack, and the whole troop speeds up.
Later TOC authors added a literal rope tying the lead scout to the slowest one, so the fast scouts can never get further ahead than the rope allows. That picture is the whole method. The name drum-buffer-rope is usually traced to *The Race* (1986), which Goldratt wrote with Robert Fox and which uses soldiers marching to a drum. Sources differ on when the name first appeared, so treat 1986 as a common attribution.
DBR is also the five focusing steps applied to a production line: identify the constraint, exploit it, subordinate everything else to that decision, elevate it if needed, and return to step one when it moves. The drum is the exploit step, the buffer protects it, and the rope is the subordination.
What the drum, buffer and rope each do
The drum: the constraint's schedule
The drum is the schedule of the resource that limits what the plant can ship. Every other resource has more capacity than the constraint, so the constraint is the only place that needs a detailed, minute-by-minute schedule. Scheduling every station adds effort without adding output.
The drum fixes which jobs run, in what sequence and how many. Its rate is the plant's rate, and an hour lost at the drum is an hour of manufacturing throughput lost for the whole plant.
The buffer: protection measured in time
The buffer is an interval of time, not a pile of parts that looked about right: the lead the schedule gives material before the constraint needs it. WIP is the visible result, but the interval is the control. Measuring in time matters because jobs differ in size. A nine-day buffer holds nine one-day jobs or thirty-six quarter-day jobs, and the protection is the same.
Traditional DBR describes up to three buffers:
- Constraint buffer: time ahead of the drum, so the constraint is not starved when an upstream station stops.
- Shipping buffer: time between the constraint finishing a job and the customer's due date, so downstream variability does not make you late.
- Assembly buffer: time ahead of an assembly point where parts that passed through the constraint must meet parts that did not.
Sources disagree on counting: some treat the constraint and shipping buffers as two parts of lead time, newer descriptions fold them into one. The logic is the same either way.
The rope: the release signal
The rope controls when raw material enters the line. Material for a job is released one buffer-time before the constraint is scheduled to start it, no sooner. Releasing earlier adds WIP and lengthens lead time without making the constraint produce one extra part. Releasing later risks starving it.
In practice the rope is a release list tied to the drum schedule.
How DBR differs from kanban and CONWIP
All three are pull-style limits on work in the system, but they cap different things.
Kanban caps inventory between stages with cards or containers and suits repetitive, stable flow. CONWIP (constant work in process), proposed by Spearman, Woodruff and Hopp in 1990, caps the total number of jobs inside a whole line: a new job enters only when a finished one leaves. It suits mixed-model work where many products share resources. DBR ties release to one resource's schedule. As one practitioner summary puts it, the buffer in DBR is time while the buffer in kanban is space.
Rough guide to fit:
- Choose kanban when flow is repetitive, the mix is stable and no single station dominates.
- Choose CONWIP when the mix is high, products share equipment, and the aim is to stop chronic over-release without first proving where the constraint is.
- Choose DBR when one resource clearly limits output, stays the limit for months, and every lost hour there costs real revenue.
Weigh the criticism too. DBR assumes a stable constraint and addresses starving at the constraint but not blocking after it. One lean writer argues kanban or CONWIP usually brings an out-of-control system under control faster. Plants often combine them, using CONWIP or kanban across the line and DBR logic around the constraint.
What DBR needs from your shop
Reliable constraint data. You need the constraint's real rate, not the routing sheet's: measured minutes per part, changeover times, downtime and scrap. A constraint rated at 5 minutes per part that actually runs at 5.8 misses every drum schedule built on the routing. The OEE calculator shows what the constraint is delivering once stops, slow running and rejects are counted.
Discipline on release. The rope works only if nobody releases early. The temptation is real: the first station is idle, so someone feeds it a job, which adds WIP and nothing to constraint output. Utilization at non-constraints is not a goal. One practitioner suggests starting with 25 to 30 percent protective capacity at non-constraints and adjusting from buffer data, because that spare capacity is what lets upstream stations recover after a stop.
Quality at the constraint. Scrap or rework there wastes constraint time directly. If quality is erratic, reruns consume the buffer and the drum schedule falls apart.
How to implement drum buffer rope, step by step
Step 1: Identify the constraint
Find the resource with the lowest capacity against demand: available minutes divided by real minutes per part, station by station. Confirm on the floor: the constraint usually has a queue in front. The bottleneck calculator does the arithmetic, and bottleneck elimination strategies covers cases where the answer is not obvious.
Step 2: Exploit the constraint
Before scheduling, remove avoidable loss there: cover breaks and handovers, do inspection and prep before the part arrives, and move changeover preparation off the machine. Every minute recovered is plant output.
Step 3: Build the drum schedule from constraint capacity
Take available time, subtract planned maintenance, breaks and changeovers, and divide by the real minutes per part. That is the drum rate and the most the plant should commit to. Sequence jobs by due date, grouped by changeover family where possible, in clock time.
Step 4: Set the buffer as time
Measure the current average lead time from material release to the job reaching the constraint, using tagged jobs. A widely quoted starting rule is to halve it, on the reasoning that most of that time is queue, not work. Divide the buffer into three equal zones: green (just released), yellow (mid-process) and red (the last third, where the job should be at or near the constraint). A practitioner summary suggests sizing so jobs reach red only about 5 percent of the time. Frequent red means the buffer is short or upstream reliability is poor. No red at all means you can try shortening it.
One academic study calls published buffer-sizing advice vague and dependent on managerial judgment, so treat the first figure as a hypothesis.
Step 5: Set the rope length and release rules
Rope length equals the buffer: release time is scheduled constraint start minus buffer. Write rules a supervisor can enforce:
- Release only jobs on the drum schedule, in drum sequence.
- Release at the scheduled time, not before, even if a station is idle.
- If material or tooling is short, hold the release and flag it. Do not release a part-kitted job to get started.
- Treat a rush job as a change to the drum schedule and re-sequence it there.
Step 6: Run buffer management daily
At each shift start, compare what should be in the buffer with what is. A job that should have reached its zone and has not is a hole. A hole in green is a note, in yellow means find the job now, and in red means expedite. Log the cause every time a job reaches red and review the log weekly. The same few causes repeat, and fixing them is how the buffer shrinks.
Step 7: Subordinate, then elevate
Everything else follows the drum. Add capacity to the constraint only once it is scheduled and protected. If another station becomes the constraint, return to step one.
A worked example: a five-station line
This is a hypothetical example with round numbers. A line makes a welded and painted assembly on five stations, each running 450 minutes per shift, one shift a day. Real minutes per part: Saw 2.0, Mill 3.0, Weld 5.0, Paint 4.0 and Pack 2.5.
Find the constraint and the drum rate
Capacity per shift is 450 divided by minutes per part: Saw 225, Mill 150, Weld 90, Paint 112, Pack 180. Weld is the constraint. It also has two 15-minute changeovers, so its run time is 450 minus 30, or 420 minutes. The drum rate is 420 divided by 5, or 84 parts per shift.
At 84 parts the other stations are loaded: Saw 168 minutes of 450 (37 percent), Mill 252 (56 percent), Paint 336 (75 percent) and Pack 210 (47 percent). Paint is the thinnest, with about a third more capacity than the drum needs, and it sits after the constraint, which matters for the shipping buffer.
What happens without a rope
If Saw runs at its own pace it releases 225 parts a shift into a Weld that consumes 84. The pile grows by 141 parts a shift, or 705 after a five-day week. That is about 8.4 shifts of Weld work, so a part released late in the week waits well over a week to reach the welder. Throughput did not change, only inventory and lead time.
Set the drum schedule
Weld's shift in minutes: changeover from 0 to 15, Product X (40 parts at 5 minutes, 200 minutes) from 15 to 215, changeover from 215 to 230, Product Y (44 parts, 220 minutes) from 230 to 450. That is 84 parts.
Size the buffer and set the rope
Tagged jobs show an average of 48 clock hours from release to the Weld queue. Halve it: a 24-hour buffer, three zones of 8 hours. Material for Tuesday's Product X job, which starts on Weld at minute 15 of Tuesday's shift, is released Monday at the matching time. Product Y material is released 24 hours before minute 230. Nothing is released that is not on the schedule, and nothing earlier.
The buffer holds about one day of Weld output, 84 parts, spread over Saw, Mill and the Weld queue. Compare that with the 705 that piled up when Saw set the pace.
Absorb a stoppage
Suppose Mill is down for 3 hours on Monday. Mill needs 84 times 3 minutes, or 252 minutes, to supply Tuesday's Weld jobs. After a 180-minute stoppage it has 450 minus 180, or 270 minutes left, enough with 18 to spare. Weld never notices, because the parts still reach the Weld queue by the end of Monday's shift, before the red zone begins.
Now suppose the stoppage is 5 hours. Mill has 150 minutes left against 252 needed, a shortfall of 102 minutes, about 34 parts. Those parts will miss the red zone, which is the signal to act: overtime at Mill, a split job, or work moved to another machine. Weld stays protected because someone acted at the red-zone threshold, not after Weld went idle.
Set the shipping buffer
Weld finishes Product Y at minute 450. Paint and Pack still have 44 parts to process: 176 minutes of Paint plus 110 of Pack, about 4.8 hours of touch time. A 24-hour shipping buffer covers that with room for queues and a Paint breakdown. Paint has the least spare capacity, so a Paint breakdown will not starve Weld but can still make the order late.
Simple and complex variants
Traditional DBR uses a drum schedule at the physical constraint plus constraint and shipping buffers. It suits plants with a clear physical constraint and a stable mix.
Simplified DBR (S-DBR) was proposed by Schragenheim and Dettmer around 2000. It assumes the market, not a machine, is the main constraint. It uses a single shipping buffer, releases material at the order's due date minus that buffer, and controls load by checking the planned load on the weakest resource before accepting an order. There is no separate constraint buffer, and the drum is effectively the due date, so jobs are not sequenced at any machine. Buffer management enforces one priority scheme on the floor. S-DBR is easier to run, but you must still use capacity well to maximize throughput.
Adaptations exist for job shops where the constraint moves with the mix, such as the Velocity Scheduling System described by its author Lisa Lang. Start with a single drum.
Limits and failure modes
- A moving constraint. DBR needs a drum that stays put. One practitioner suggests placing the constraint deliberately at the resource that is hardest or most expensive to expand, and treating any move as a business decision.
- Wrong buffer size. Too small, and the constraint starves and the plant loses throughput. Too large, and lead time, operating expense and cycle time climb while inventory turns fall.
- Ignoring the shipping buffer. The constraint can run perfectly while orders ship late because downstream stations lacked protective capacity. Loaded close to the drum rate, a non-constraint can turn into an interacting constraint, which is much harder to manage.
- Ignoring blocking. If a station after the constraint cannot accept its output, the constraint blocks. Add a space limit after the drum if that can happen.
- Thin academic evidence. Gupta and Snyder (2009), as summarized in general reference literature, argued the Theory of Constraints had not yet shown its effectiveness in academic research. Prove DBR on your own line with before and after lead time and on-time delivery.
Common mistakes
Naming the constraint from opinion instead of measured minutes. Building the drum from the routing sheet. Sizing the buffer once and never revisiting it. Releasing a whole week of material on Monday. Rewarding utilization at non-constraints. Changing the drum mid-shift for every rush request.
The constraint's schedule is the plant's schedule. Release material one buffer ahead of it, never earlier, and protect that interval with spare capacity on both sides.
A Monday-morning starter checklist
- Pick one line with a visible queue in front of one station.
- Time that station for a full shift: real minutes per part, stops, changeovers, scrap.
- Compute every station's capacity at real rates and confirm which is lowest.
- Calculate the drum rate: available minutes, minus changeovers and planned stops, divided by minutes per part.
- Write a one-shift drum schedule with clock times for each job.
- Tag a handful of jobs, measure release-to-constraint time, halve it for the first buffer and split it into three zones.
- Create a release list with each job's release time (constraint start minus buffer).
- Tell supervisors the single release rule: nothing goes in earlier than the list says.
- At each shift start, walk the buffer, mark holes by zone, and log the cause of every red.
- Review the log weekly and fix the top cause. Shorten the buffer only after red stays rare.
Frequently asked questions
Is drum buffer rope just a bottleneck schedule?
No. A bottleneck schedule sequences jobs at the constraint. DBR adds the buffer, which protects that schedule from upstream variability, and the rope, which stops material being released faster than the constraint consumes it. Without the rope the line still floods.
How big should the time buffer be?
A common starting rule is half the current average lead time from release to the constraint, split into green, yellow and red thirds. Then adjust from data: lengthen it or fix upstream causes if jobs reach red much more than about 5 percent of the time, and try shortening it if they never do.
Can I run DBR when the constraint moves?
Only with care. DBR assumes a drum that stays put for months. If yours shifts with the mix, decide where it should sit and protect it with capacity elsewhere, or start with CONWIP.
Does DBR replace kanban?
Not necessarily. Kanban caps inventory at each stage and suits repetitive flow, while DBR paces release to one constraint. Many plants use both.
What is simplified drum buffer rope?
Simplified DBR (S-DBR) treats the market as the constraint and uses one shipping buffer. Material is released at the due date minus that buffer, and the load on the weakest resource is checked before new orders are accepted. It is easier to run than traditional DBR.
Want our lean manufacturing guides first in Google? Add us as a preferred source.
Add LeanSuite as a preferred source on Google (opens in a new tab)
