Subordination
Concept: Vocabulary that names a phenomenon.
Aligning every non-constraint resource and policy to the rate the system constraint can consume, so the constraint’s exploitation is not undone by the rest of the system.
Subordination is the operational discipline that runs the system at the constraint’s pace. It is named for the third of the Five Focusing Steps, and it is the step practitioners most often skip. The omission is structural: subordination asks non-constraint resources to run below their local optimum, which conflicts with every scorecard a conventionally measured organization runs. Skipping it reinstalls the local-optimization antipattern.
Context
Every later TOC application depends on this alignment. Drum-Buffer-Rope is the production specialization: the rope releases material at the drum’s rate, which is the rate the constraint can consume. The Project Buffer and the Feeding Buffer are the CCPM expressions: feeder chains are subordinated to the critical chain, and the feeding buffer absorbs the variability the subordination produces. Replenishment Pull and Buffer Aggregation in Distribution carry the same logic into supply chains: stocking points hold the system buffer; upstream nodes release at the pull signal, not at their own batch-economic optimum. Even the Mafia Offer is the subordination move for a market constraint, with sales redirected at the offer the constraint can absorb.
The concept depends on three priors. The Goal supplies the direction the system is moving in (without a goal, non-constraints have no rate they could be subordinated to). Constraint supplies the resource the move aligns against. The Five Focusing Steps supply the discipline’s place in the sequence: subordinate happens after exploit and before elevate, and the order is not negotiable. Run subordinate before exploit, and the rest of the system is asked to align against an under-exploited constraint, locking in a system rate well below what the constraint could deliver. Run elevate before subordinate, and the system has been given more constraint capacity that the rest of the system, still running at its own local optimum, will keep starving of work or burying in WIP.
The TOC literature treats subordination as operationally central. The TOCICO Dictionary gives it a standalone entry. The TOC Handbook spends much of its production-management content on implementation. Srikanth’s DBR chapter (Cox & Schleier 2010, Ch. 8) names it explicitly as “the most challenging and difficult task” of every DBR implementation he has worked.
What It Is
The TOCICO Dictionary defines subordinate as the key word in the third focusing step and decomposes it into two operational components: non-constraint resources must deliver the constraint exactly what it needs when it needs it, and non-constraint resources must not process work the constraint will not absorb. The first component is what protects the constraint from starvation; the second is what prevents the work-in-process pileups that produce local efficiencies and global lateness. Both components are required; either alone is incomplete.
Three operational moves implement the discipline.
Setting the release pace at the constraint, not at non-constraint capacity. System input, whether material, project work, customer requests, or tasks, is released at the rate the constraint can consume. In DBR this is the rope: a release schedule pegged to the drum (the constraint’s schedule), with a time buffer offset that lets the constraint absorb variability without being starved. In CCPM the same idea governs project starts. New work is released only when the critical chain has capacity to absorb it, with feeding buffers absorbing feeder-chain variability. In supply chains it is the pull signal: an upstream node releases when the downstream stocking point’s buffer has been depleted, not when the upstream node’s batch-economic optimum says it should.
Leaving non-constraint capacity idle when the constraint cannot absorb its output. This is the move that most conflicts with the inherited measurement frame. A non-constraint resource with 120 units per day of capacity, asked to produce for a constraint that can absorb 100 units per day, is told to produce 100. Its measured utilization will be 83% instead of 100%. The cost-world reading is inefficiency and waste. The throughput-world reading is that the extra 20 units would have been work-in-process inventory the system did not need, charged against operating expense without contributing to throughput, and likely to disrupt the constraint by burying its queue in the wrong sequence. The non-constraint’s idle capacity is also where the system holds protective capacity: the cushion needed to catch up after a disruption, which a 100%-loaded non-constraint cannot supply.
Sequencing non-constraint work to maximize the constraint’s exploitation rate, not the non-constraint’s local efficiency. When a non-constraint has a queue of items waiting to be processed, the sequence is chosen by the impact on the constraint’s throughput, not by what is cheapest for the non-constraint to run next. A non-constraint setup choice that minimizes its own changeover time at the cost of delaying the next constraint batch is the wrong choice. The non-constraint’s changeover time is excess capacity the system already has more of than it needs; constraint hours are the resource the system has the least of.
Each TOC application parameterizes the move differently. DBR uses the drum schedule. S-DBR uses planned load against a planned shipping date. CCPM uses the buffer color reading on the critical chain. Replenishment Pull uses the buffer status at the stocking point. The underlying structure is the same: one rate-setting signal originates at the constraint, propagates outward through the system, and aligns non-constraints to that signal instead of to their own local optima.
How It Is Measured or Recognized
Three operational tests separate a system that has implemented subordination from one that has only declared it.
The first test is the non-constraint idle test. In a subordinated system, non-constraint resources are visibly idle some of the time, and the idleness is not treated as a defect to be eliminated. Operators of non-constraints can name the reason they are idle: the constraint cannot yet absorb the next batch, the buffer is full, or the schedule says hold. Supervisors do not pressure them to find make-work.
Run the test in reverse: a system in which every resource is observed running at full utilization is a system that has not subordinated, because the non-constraints’ full utilization is producing inventory the constraint cannot absorb. The 95-to-100-percent figure Srikanth cites, with non-constraints constituting the overwhelming majority of resources in any non-trivial system, is what makes the test useful. There should be many idle non-constraints, not few.
The second test is the release-signal test. Walk the system backwards from the constraint and ask, at each upstream resource, what triggers release of the next unit of work. In a subordinated system the answer ties back to the constraint’s pace: a rope signal, a drum schedule, a buffer status, or a pull request from the next downstream node. In an un-subordinated system the answer is local: “we run when the previous batch is finished,” “we run when the operator is available,” “we run when the start-of-shift checklist is clear.” Both answers can come from a confident operator with years in the system. The difference is whether the release decision propagates the constraint’s rate or the resource’s own.
The third test is the measurement-set test, Goldratt’s diagnostic at the system level. Read the scorecards the system actually publishes. A subordinated system has retired or substantially recalibrated the non-constraints’ local-efficiency, utilization, and unit-cost measures, because those measurements reward the inverted behavior. If the system still publishes 100%-utilization targets for non-constraints, the measurement frame is pulling the operation back toward the un-subordinated default no matter what the implementation plan declared.
The standard implementation move is to replace local-efficiency reports on non-constraints with buffer-status reports tied to the constraint’s exploitation. The Continuous Improvement Auditing discipline (Barnard & Immelman, Cox & Schleier 2010 Ch. 15) is built around watching for that regression.
A diagnostic at the planning level is also useful: ask whether release decisions are made by a non-constraint manager applying that resource’s standard release rules or by a central scheduling function reading the constraint’s signal. In a subordinated DBR plant a single rope schedule governs material release at the gating operation, and individual workstation managers do not authorize releases on their own initiative. In a subordinated CCPM portfolio a single critical-chain manager controls project starts against the staggering schedule, and individual project managers do not begin work on their own clock.
Why It Matters
Subordination is the move that converts exploitation into throughput. Exploitation maximizes what the constraint can produce; subordination prevents the rest of the system from undoing the exploitation. The two are paired by design. The empirical record across TOC implementations is consistent: implementations that complete Step 2 without Step 3 produce a more efficient constraint embedded in a system whose throughput does not move, while implementations that complete both produce the system-throughput gains the case literature documents.
Subordination is also the move that prevents the Local Optimization antipattern. Local optimization is named for the pattern of running each part of the system at its own local optimum without regard to the system rate; subordination is the explicit refusal of that pattern. Subordination is similarly the defense against Improving Non-Constraints: a non-constraint that has been subordinated has visible excess capacity; the diagnosis of that excess capacity as “improvement opportunity” is the antipattern, and the corrective is to read the excess as the protective capacity the system needs to subordinate well.
Subordination is the discipline the standard accounting frame works against. The cost-world reading of an idle non-constraint as inefficiency is the readout the Cost-World Thinking antipattern is built on. The recommendation that follows (“eliminate the idle capacity, reduce headcount at the non-constraint, outsource the spare hours”) is exactly the move that breaks subordination and turns excess into a thin operating margin the next disruption will expose. Holding subordination across an accounting close requires the Throughput Accounting frame (Budd 2010, Cox & Schleier Ch. 13) and explicit naming of the cost-world recommendation as a misreading of protective capacity.
For a TOCICO certification candidate, subordination is the operational hinge between the Fundamentals and the domain-specific Practitioner credentials. The Fundamentals exam tests the candidate’s ability to state the Five Focusing Steps and the three global measures. The Practitioner exams in Supply Chain, CCPM, Throughput Accounting, and Thinking Processes test whether the candidate can implement subordination in a specific domain, distinguish the implementation mechanism from the underlying logic, and defend the mechanism to a sponsor who reads cost-world reports.
Common Misunderstanding
“Subordination means running everyone less.” It means running each non-constraint at the rate the constraint can absorb. Sometimes that is less than the non-constraint’s local maximum; sometimes it is more, during a catch-up after a disruption; sometimes it is the same. The principle is rate-matching, not rate-reduction. A subordinated system uses its non-constraints harder during a catch-up than an un-subordinated system would, because the protective capacity held back is the capacity available for catch-up. It does not run them harder when the constraint cannot absorb the output.
“Subordination is a scheduling problem.” It is also a measurement problem, an incentive problem, and a vocabulary problem. The scheduling mechanism (the rope, the critical-chain schedule, the pull signal) propagates the constraint’s rate, but it cannot hold in a system whose scorecards reward the inverted behavior. The implementations the Handbook documents as durable, including Mabe, Boeing’s CCPM rollout, and Procter & Gamble’s distribution redesign, all retired or substantively recalibrated local-efficiency measurements alongside the scheduling change.
“Subordinate to the constraint, then move on to elevate.” Subordination is not a one-time configuration; it is an ongoing daily-control discipline. Buffer Management is the daily-control instrument: the buffer’s color reading (green, yellow, red, black) signals whether the system is on the rate the constraint requires. A subordinated system reads the buffer every day. An un-subordinated one reads it once during implementation and treats the configuration as complete.
“Non-constraint subordination is the same as resource leveling.” Resource leveling, in conventional project management, smooths the workload across resources to even out utilization. Subordination does the opposite: it accepts deliberate unevenness in utilization, with high utilization at the constraint and low-to-moderate utilization at non-constraints. The two practices look superficially similar (both adjust release rates and timings) but optimize against different goals. Resource leveling optimizes against resource utilization; subordination optimizes against constraint exploitation.
“Subordination breaks down when the constraint moves.” The discipline is built to expect constraint motion. Step 5 of the Five Focusing Steps directs the system to return to Step 1 when the constraint has moved, and subordination is then re-pegged to the new constraint. The failure mode named by the Step 5 warning clause (do not allow inertia to become the system’s constraint) is the failure to re-subordinate when the binding constraint has shifted. The fix is not to abandon subordination; it is to keep subordination tied to whichever resource is currently binding the system, recalibrating each time the binding shifts.
“In services or knowledge work, there is nothing to subordinate.” The mechanism changes but the principle does not. In a software organization with a constraint at code review, subordination means feature work starts at the rate review can absorb, not at the rate squads can begin sprints. In a hospital A&E with a constraint at the assessment step, subordination means upstream triage and downstream bed allocation align to the assessment pace. The Oxfordshire Radcliffe implementation (95%+ four-hour target from a 50 to 60% baseline in roughly two months, with no added staff) is the canonical service-sector demonstration that subordination is the move that produces the throughput gain.
Related Articles
Complements: Exploitation Without Elevation — Step 2 (exploit) and Step 3 (subordinate) are the paired moves: exploitation maximizes what the constraint can produce; subordination prevents the rest of the system from undoing the exploitation.
Depends on: Constraint — Subordination is undefined until a constraint has been identified; the move is the alignment of everything else to whatever the named constraint needs.
Depends on: The Goal (concept) — Non-constraints are aligned to the constraint's needs because the constraint sets the system's rate of goal-unit production; the goal supplies the direction the subordination move points in.
Enforces: Buffer Management — Buffer Management is the daily-control discipline that enforces subordination: the buffer's color reading is the signal the rest of the system is on or off the rate the constraint requires.
Implements: Drum-Buffer-Rope (DBR) — DBR is the production implementation of subordination: the rope releases material at the drum's rate, which is the rate the constraint can consume.
Implements: Project Buffer — The project buffer is the CCPM expression of subordination at the project level: the buffer absorbs feeder-chain variability so the critical chain is subordinated to and protected.
Part of: Five Focusing Steps — Subordination is the third of the Five Focusing Steps; the standalone entry expands the move the parent pattern names but does not spell out in operational detail.
Prevents: Improving Non-Constraints — Improving a non-constraint's capacity raises that resource's local efficiency without raising system throughput; subordination is what makes the futility of the move visible.
Prevents: Inventory (TOC Definition) — Subordination prevents the work-in-process pileups that produce excess inventory: non-constraints that run at their own local optimum produce material the constraint cannot absorb.
Prevents: Local Optimization — Local optimization is what happens when non-constraints are run at their own local optimum instead of at the rate the constraint can consume; subordination is the named refusal.
Produces: Throughput — Subordination produces throughput by ensuring the constraint is never starved and never asked to process work the rest of the system will not absorb.
Sources
• Cox, TOCICO Dictionary (2nd ed., 2012), entries “subordinate,” “non-constraint,” “protective capacity,” “excess capacity,” and “drum-buffer-rope” — the certification-grade definitions this entry cross-checks against. The Dictionary is distributed through TOCICO membership rather than the trade book channel and does not carry an Open Library record.
• Goldratt and Cox, The Goal (1984; revised 1986, 1992) — the originating treatment. The dramatized subordination move is the boy-scout-hike chapter, in which Herbie’s pace sets the rate of the column and the faster scouts are deliberately positioned behind him; the move is then transferred to the plant in the chapters on rope and drum.
• Goldratt, What Is This Thing Called Theory of Constraints and How Should It Be Implemented? (1990) — formalizes the five-step structure and gives the operational specification of the subordinate step that the Dictionary later canonized.
• Cox and Schleier, Theory of Constraints Handbook (2010), Ch. 1 (Goldratt, “Introduction to TOC — My Perspective”) and Ch. 8 (Srikanth, “DBR, Buffer Management, and VATI Flow Classification”) — Srikanth’s DBR chapter names subordination as the most challenging task of every DBR implementation he has worked, and the chapter spends most of its operational content on the mechanisms that propagate the constraint’s rate through the rest of the plant.
• Cox and Schleier, Theory of Constraints Handbook (2010), Ch. 13 (Budd, “Traditional Measures in Finance and Accounting, Problems, Literature, and TOC Measures”) and Ch. 15 (Barnard and Immelman, “Continuous Improvement and Auditing”) — the measurement-frame chapters that name the cost-world reading of non-constraint idle capacity as the antipattern subordination has to be defended against.
• Schragenheim, Dettmer and Patterson, Supply Chain Management at Warp Speed (2009) — applies subordination logic to supply chains and articulates the simplified-DBR (S-DBR) variant that uses planned load against a planned shipping date as the subordination signal.
• Mabin and Balderstone, The World of the Theory of Constraints (2000) — meta-review across more than 80 documented TOC implementations; the cases that completed Step 3 in full versus those that skipped it are the empirical basis for the claim that subordination is the operational hinge of TOC.
• Dettmer, Goldratt’s Theory of Constraints (1997) — extended treatment with explicit handling of policy and paradigm constraints; the subordination move for non-physical constraints (the alignment of policies, measurements, and beliefs to the binding constraint) is named and worked at chapter length.