Operations & process
Theory of Constraints
Goldratt's argument that every system has one binding constraint that sets its throughput. Find it, wring the most from it, subordinate everything else to it, invest to elevate it, then start again, because once a constraint is broken the constraint moves somewhere else.
Also known as TOC, Five Focusing Steps, Constraint management, Process of ongoing improvement. First set out by Eliyahu M. Goldratt in 1984; the primary source is cited in full below.
- Format
- Process / loop
- Level
- Business unit · Team
- Best for
- Plan execution · Allocate resources · Prioritise
- Decision stage
- Diagnose · Plan · Execute
- Difficulty
- Intermediate
- Time to apply
- Days to identify and begin exploiting a constraint; embedding subordination, buffer management and the measurement changes takes months.
Plate · The model
The components
Identify the constraint
Locate the one resource or policy that limits the throughput of the whole system. Everything else in TOC depends on getting this diagnosis right.
Signals of strength
Work-in-progress accumulates in front of one stage · Downstream stages are starved and wait · Expediting and overtime always converge on the same resource · Promised lead times are set by one stage's queue
Exploit the constraint
Extract maximum output from the constraint without capital spend: eliminate its idle time, keep defective work away from it, and reserve it for work only it can do.
Signals of strength
Constraint utilisation measured and losses categorised · Breaks and shift handovers staggered so the constraint never stops · Quality checks moved upstream of the constraint · Constraint time freed from work other resources could do
Subordinate everything else
Align every other resource, schedule and measure to the constraint's pace, typically via drum-buffer-rope release. Non-constraints deliberately carry spare capacity to protect the constraint's flow.
Signals of strength
Work released to the floor at the constraint's pace, and no faster · A sized time buffer protecting the constraint, actively managed · Local efficiency targets on non-constraints relaxed or removed · Falling work-in-progress everywhere except the buffer
Elevate the constraint
Increase the constraint's capacity through investment, once exploitation and subordination have proven insufficient for demand. Options are valued in throughput per constraint-hour, the currency of throughput accounting.
Signals of strength
Exploitation gains exhausted and documented before spend · Investment cases stated in throughput terms, beyond unit cost · Capacity added in steps, with the effect on flow re-measured · A view of where the constraint will move once elevated
Repeat: find the new constraint
Return to step one when the constraint breaks, and dismantle the rules that grew up around the old one. Inertia, running the system to a constraint that no longer exists, is the failure mode this step guards against.
Signals of strength
The new binding resource identified after each change · Old constraint-era policies reviewed and retired · Buffers and release rules re-sized to the new drum · Recognition when the constraint moves into the market, making sales the drum
When it earns its keep
- Work piles up reliably in front of one stage of a process while stages downstream sit idle. That queue is the system announcing its constraint.
- Local efficiency looks good everywhere and yet the whole system delivers late. TOC exists precisely for organisations that have optimised every department and starved the flow.
- You must decide where a limited improvement or capital budget goes. An hour saved at the constraint raises the whole system's output; an hour saved anywhere else is, in Goldratt's phrase, a mirage.
- Lead times and work-in-progress are growing and you need a scheduling logic. Drum-buffer-rope, pacing release of work to the constraint's drumbeat with a protective time buffer, follows directly from the analysis.
And when it doesn't
- The system has no stable bottleneck: demand is far below capacity everywhere, or the constraint shifts week to week with product mix, in which case the five steps chase a moving target and flow-levelling approaches may serve better.
- The real limit is outside operations, a market constraint of too few orders, or a policy constraint such as a batching rule. TOC handles these conceptually, but the Five Focusing Steps in their physical-bottleneck form will point you at the wrong machine.
- Quality or variation is the dominant problem. TOC maximises flow through the constraint; if a third of what flows is defective, DMAIC or a fishbone-led investigation comes first.
- You need certified cost figures. Throughput accounting is a decision-making lens, deliberately at odds with absorption costing; it will not satisfy statutory reporting and should not be asked to.
How to run it
Before starting, gather the inputs the analysis depends on:
- A mapped flow of the process from order to delivery, with actual demonstrated capacity at each stage, since nameplate capacity lies.
- Evidence of where queues, expediting and waiting actually occur: work-in-progress counts, wait times, and which stage overtime always goes to.
- Throughput-accounting basics: selling price minus truly variable cost per unit, so that decisions at the constraint can be valued in money per constraint-hour.
- Honesty about policies: batch sizes, shift patterns, incentive schemes and measures that may themselves be the constraint.
- 1
Identify the system's constraint
Find the resource that limits throughput for the whole system. The physical evidence is a queue in front of it and starvation after it; the organisational evidence is that everyone expedites around it. There is usually one, and it is frequently a policy rather than a machine.
- 2
Exploit the constraint
Get the most from the constraint as it stands, before spending anything. Stop it processing defective or unnecessary work, cover its breaks, cut its changeovers, and make sure it is never idle waiting for material. Exploitation is where TOC pays for itself; most constraints are losing hours nobody has counted.
- 3
Subordinate everything else
Run every non-constraint to serve the constraint's schedule, and accept that they will show idle time. This is drum-buffer-rope: the constraint sets the drum, a time buffer protects it from upstream disruption, and the rope releases work into the system only at the pace the constraint can consume. It is also the step organisations resist, because it makes efficient departments look inefficient.
- 4
Elevate the constraint
Only after exploitation and subordination, add capacity: new equipment, more people, offloading work, outsourcing. Elevation costs money, which is why it comes fourth; throughput accounting, valuing options in throughput per constraint-hour, tells you whether the investment pays.
- 5
Repeat: find the new constraint
When the constraint breaks, another stage becomes binding, and rules built around the old constraint become active liabilities. Goldratt's warning at this step is about inertia: do not let yesterday's bottleneck run today's factory. The loop is the point; TOC calls it a process of ongoing improvement.
Reading the result
A diagnosis of the system's binding constraint, a sequenced plan to exploit, subordinate to and finally elevate it, a scheduling logic (drum-buffer-rope) built around it, and decision measures, throughput, inventory and operating expense, that value actions by their effect on the whole system rather than on local efficiencies.
- Judge every proposed action by its effect at the constraint. Improvements elsewhere may be genuinely worthless for throughput, however satisfying locally.
- Expect the paradox: a well-run TOC system shows idle time on most resources. That idleness is the price of flow, and the instinct to eliminate it is how systems clog.
- Read the money through throughput accounting: throughput (sales minus truly variable cost), inventory and operating expense. Products are ranked by throughput per constraint-hour; conventional unit costs actively mislead at the constraint.
- Treat the fifth step as the most important and least practised. If the same 'bottleneck' has been named for years while the factory around it has changed, inertia has set in.
A worked example
An upholstered furniture factory finds its drum in the sewing room
A UK manufacturer of upholstered sofas and armchairs, around 260 units a week across frame shop, fabric cutting, sewing, upholstery benches and final inspection, is quoting six-week lead times, missing a third of promised delivery dates, and turning away trade orders. Every department reports healthy efficiency. The operations director, fresh from reading The Goal, walks the flow with the five steps rather than commissioning a new ERP module.
- Identify the constraint
- Three days of cut fabric sits racked in front of the sewing room, while upholstery benches downstream wait for covers most mornings. Overtime records confirm it: sewing has run Saturdays for eight months. The constraint is the sewing room's eleven skilled machinists, and no useful capacity number for it exists until the walk-through produces one: roughly 275 cover sets a week demonstrated.
- Exploit the constraint
- A week of observation finds the machinists losing about 15% of capacity to work that should never reach them: re-sewing covers cut from flawed fabric, machining simple cushion casings that a trainee could sew, and waiting while thread and fabric-type changeovers are batched badly. Fixes cost almost nothing: fabric inspection moves upstream of cutting, cushion casings go to two trainees, and like-fabric jobs are sequenced together. Demonstrated capacity rises to around 310 sets.
- Subordinate everything else
- Cutting had been optimising fabric yield by cutting whatever nested best, which is why the wrong covers were always plentiful. Now cutting releases to a sequence set by sewing's schedule, the drum, with a one-shift buffer of cut sets held in front of sewing and work released into cutting by that rope. The frame shop stops building frames to forecast and drops its output 8%, which the weekly report initially records as a productivity failure until the measure itself is changed.
- Elevate the constraint
- With demand still above 310 sets, the firm elevates: a second twin-needle machine and a machinist apprenticeship for two upholsterers. The 28,000 pound outlay had previously failed a unit-cost appraisal; valued in throughput per sewing-hour it repays in under five months, the clearest local demonstration of why throughput accounting exists.
- Repeat: find the new constraint
- Within a quarter the queue reappears, now in front of the upholstery benches: the constraint has moved. The like-fabric sequencing rule built for sewing is now hurting bench flow and has to be retired, a small, instructive case of Goldratt's inertia warning. The five steps begin again, one stage downstream.
The read. Weekly output rises from about 260 to just over 300 units and quoted lead times fall from six weeks to three and a half, with the largest gains coming from the free steps, exploitation and subordination, before any capital was spent. The honest reading: TOC delivered focus, and the fifth step delivered humility, because the reward for breaking a constraint was a new one, plus a set of obsolete rules to unlearn. The next constraint may well be orders, at which point the drum moves to the market and the factory's problem becomes sales.
Pitfalls
- Elevating before exploiting: buying capacity while the constraint loses hours to bad sequencing, rework and unattended breaks. Capital is the most expensive way to discover the schedule was the problem.
- Refusing to let non-constraints idle. Managers who chase utilisation everywhere rebuild the work-in-progress mountain that subordination just removed.
- Hunting the constraint among machines when it is a policy: a batch-size rule, a month-end shipping push, or an incentive scheme paying for local output. Goldratt held that the most damaging constraints are usually policies.
- Keeping cost-accounting measures alongside TOC scheduling. If departments are still judged on absorption efficiency, subordination will be quietly sabotaged in self-defence.
- Freezing at step four: treating the once-identified bottleneck as permanent. The fifth step, with its warning against inertia, is the difference between a process of ongoing improvement and a one-off project.
What the critics say
Trietsch argues that TOC's scheduling claims contain an internal inconsistency: drum-buffer-rope depends on forbidding a balanced plant, yet the elevation step drives the system towards balance, and that much of what TOC presents as new, bottleneck-focused scheduling and criticality, restates results long known in operations research and PERT/CPM practice without acknowledgement.
Trietsch, D. (2005) 'From Management by Constraints (MBC) to Management By Criticalities (MBC II)', Human Systems Management, 24(1), pp. 105-115.
The single-constraint assumption is the framework's load-bearing simplification. In systems with shifting product mix, high variability or several near-critical resources, the binding constraint moves faster than the five steps can cycle, and scholars examining TOC's status as theory find its concepts loosely defined and its evidence base dominated by advocate-reported cases rather than controlled comparison.
Naor, M., Bernardes, E. S. and Coman, A. (2013) 'Theory of constraints: is it a theory and a good one?', International Journal of Production Research, 51(2), pp. 542-554.
Throughput accounting's dismissal of conventional cost accounting is itself contested. Accounting researchers note that treating direct labour and most overhead as fixed is a short-run assumption; used for long-run product and pricing decisions, throughput ranking can mislead just as surely as the absorption costing it mocks.
Dugdale, D. and Jones, T. C. (1998) 'Throughput accounting: transforming practices?', British Accounting Review, 30(1), pp. 73-91.
Sources and further reading
- Goldratt, E. M. and Cox, J. (1984) The Goal: A Process of Ongoing Improvement. Great Barrington, MA: North River Press.
- Goldratt, E. M. (1990) Theory of Constraints. Croton-on-Hudson, NY: North River Press.
- Trietsch, D. (2005) 'From Management by Constraints (MBC) to Management By Criticalities (MBC II)', Human Systems Management, 24(1), pp. 105-115. ↗
- Rahman, S. (1998) 'Theory of constraints: a review of the philosophy and its applications', International Journal of Operations and Production Management, 18(4), pp. 336-355.