Theory of Constraints: Prevent your bottlenecks from shifting

Applying the Theory of Constraints in a Factory: the “moving” bottleneck and the trap of local wins

A local improvement can make the numbers look good—and still inflate work-in-progress by 30% without increasing shipments. This paradox appears when you optimize workstation by workstation, without looking at the whole system. The Theory of Constraints, applied to industrial production, puts the spotlight back on a simple question: what is really limiting sellable throughput today? You then prioritize the constrained resource, rather than adding up scattered gains.

Key takeaway: stabilizing flow through synchronization and release rules reduces work-in-progress, shortens lead times, and avoids local “wins” that burn cash.

I- The “jumping” bottleneck: remove one constraint and create another

A constraint is not necessarily fixed. Depending on product mix, setups, and planning, the limiting point moves from one station to another. When a machine becomes faster without an adapted release rule, work-in-progress shifts and the customer sees the same delays. The management trap is rewarding local activity instead of system output: that is exactly the bias the Theory of Constraints aims to break.

A shopfloor symptom that doesn't lie: the WIP pile moves from station to station

Repeated observation: a station that produces more feeds a pile in front of the next station, and shipped pace does not improve. Proof sits in the correlation between useful saturation and shipped throughput. Without this check, you confuse visibility with the real constraint. Measuring WIP (Work In Progress, work-in-progress) provides an early signal.

Mini shopfloor case: faster machine, same plant output, more WIP

What: replacement of a machine with a 25% cycle-time gain. How: unchanged release rules and increased upstream feeding. Impact: shipped throughput +3% and WIP +30% with a deterioration of lead time. Conceptual sources: Goldratt, The Goal and principles from Factory Physics (Hopp & Spearman).

II- Dynamic constraint: product mix, variability, and utilization move the critical point

The bottleneck becomes nomadic when the product mix varies. A complex product constrains machining; a standard product constrains assembly. Managing by averages hides the variability that creates queues. At high utilization, the system loses its absorption margin and the bottleneck jumps more often: Factory Physics and Kingman's formula (queueing theory) describe this non-linear explosion of waiting time beyond ~85–90% utilization, depending on variability.

When the product family decides the limiting station

Two product families on the same line can impose two different constraints depending on their cycle times. The good practice is to segment by flow and by comparable families, and to find the constraint along the shipping path. Control must connect the resource to sellable throughput, not to utilization rate.

Factory Physics: why variability makes the bottleneck unstable

Factory physics shows that variability turns into waiting.

Beyond 85–90% utilization, queueing theory (Kingman's formula, as presented in Factory Physics) predicts a rapid increase in waiting times for the same drift in variability. The consequence is instability of the limiting point and a fast deterioration of lead time (lead time).

III- Measure and qualify the constraint: from theory to shopfloor proof

A constraint is proven by tests and data; it is not decided by vote.

Operational definition: the system constraint limits shipped throughput over a given horizon. A simple test is to temporarily improve a resource and verify whether sellable output increases. This sequence avoids reflex investments and creates a common language across production, methods, quality, supply chain, and finance.

Distinguish bottleneck, market constraint, and policy constraint

Three types of constraints exist: internal capacity, market, and policy.

A policy constraint comes from a rule or parameter, such as a batch size imposed by the ERP (Enterprise Resource Planning, integrated management software). Treating a policy constraint goes through the rule—not only through CAPEX (capital expenditure).

Shopfloor diagnosis: minimal data, fast tests, cross-validation

Collect cycle times, stoppages, setups, scrap, downstream blockages, and upstream starvation over a period representative of the mix. A station that is starved for long periods cannot be the constraint over the studied horizon.

Cross-check visual observations, machine logs, and ERP extracts to confirm the diagnosis.

Useful indicators and toxic indicators: avoid OEE everywhere

OEE (Overall Equipment Effectiveness) helps improve a station; it does not manage a flow. Relevant indicators: shipped throughput, WIP, and lead time.

High OEE outside the constraint can create work-in-progress and degrade service; this is a known effect in queued systems (Hopp & Spearman), where over-feeding upstream mostly increases waiting.

IV- The 5 focusing steps: a shopfloor operating model, deliverables, and metrics

The five steps enforce a logical order and concrete deliverables. They reduce unnecessary spending and turn improvement into measurable results. Success is measured by three indicators: shipped throughput, WIP, and lead time (traversal time). A deliverable-free approach becomes talk, and the effect evaporates.

• Step 1: identify the system constraint, without confusing noise and signal
  Define scope, product families, and horizon. Find the resource whose effective capacity caps output. Validate with a test: does improving the resource increase sellable output?

• Step 2: exploit the constraint, before buying a machine
  Exploit means reducing losses at the limiting point: setups, micro-stops, first-pass quality, and tooling availability. List dominant losses and build a short removal plan. Measure useful availability and first-pass yield.

• Step 3: subordinate the rest, or stop “producing to produce”
  Subordinate means aligning releases to the constraint and limiting batch sizes. Define WIP limits by area. If WIP drops and throughput stays stable or improves, the rule works.

• Step 4: elevate the constraint, with a clear financial referee
  Elevate combines operational solutions and investment. Compare OPEX (operating expenditure) and CAPEX (capital expenditures) with one criterion: incremental sellable throughput. Use simulation to quantify ROI (return on investment) and anticipate the next limiting point.

• Step 5: repeat, because the constraint always moves
  A moved constraint signals a living system. Restart identification and adjust buffers and rules. The maturity metric is how fast you stabilize after the shift.

V- Drum - Buffer - Rope: day-to-day bottleneck control

The drum (drum) - buffer (buffer stock) - rope (rope) system translates the Theory of Constraints into daily control, without fluff. The drum sets the actually deliverable output pace, not the “desired” pace. The buffer protects that pace against variability (breakdowns, setups, quality, micro-stops). The rope limits upstream releases to prevent WIP from swelling, tying up cash, and stretching lead times.

A simple signal: if shipments stagnate while “everyone is running,” you don't have an effort problem—you have a synchronization problem. The point is not to keep resources busy; it is to feed the constrained resource at the right time with the right priorities. That's where local wins become expensive.

Define the drum: single pace, prioritization rules, and sequence freeze

Choose a pace based on effective capacity: real cycle time, availability, first-pass good-part rate, and changeover time. Then impose a readable prioritization rule: what protects the drum goes first; the rest waits. Freezing the sequence over a short window reduces replanning and “panic” setups. Measure two things: sequence adherence and shipped throughput (not produced volume).

Size buffers: protect throughput, without drowning the shopfloor

A buffer is sized from measured hazards, not gut feel: downtime, cycle dispersion, quality instability, logistics delays. Express it in time (e.g., hours of protection) or pieces, then visualize an alert code (green/orange/red) to trigger the right actions. Too small, the drum starves; too large, the shopfloor clogs. The useful indicator: drum feeding rate and fewer emergencies.

• Green: normal feeding, no action

• Orange: targeted acceleration of upstream operations

• Red: top priority, immediate maintenance/quality support

Set the rope: release rules, WIP limits, and execution discipline

The rope sets how much and when to release so WIP stays under control while guaranteeing drum feeding. Define WIP limits by area and a single release rule executable by shopfloor teams. The benefit is direct on working capital requirement (BFR, Besoin en fonds de roulement): less WIP, less cash tied up, shorter lead times. In practice, +20% WIP often translates into +20 to +40% cash tied up, depending on turns and the purchased-content share in WIP.

With a digital twin, like at Dillygence, you test these settings before enforcing them on the shopfloor and quantify the impact on throughput, lead times, and costs.

VI- Why simulation beats intuition for a nomadic bottleneck

Intuition sees a pile; simulation shows the dynamics. A model connects resources, rules, variability, and product mix. It quantifies impact on throughput, WIP, and lead time, and anticipates the new limiting point after action.

Test scenarios: exploit vs elevate, without disrupting operations

Compare exploitation scenarios (low cost) and elevation scenarios (investment). Rank them by ROI and risk. Validate robustness under disturbances to avoid an “average” scenario that collapses at the first incident.

Link throughput, WIP, and lead time in a single model

The non-linear relation between utilization and waiting comes from queueing theory (queueing) and Kingman's approximation. Simulation turns opinion debates into quantified trade-offs and enables buffer sizing from data.

Mini case: lead-time reduction by limiting releases

What: a multi-reference line showed an average lead time of 12 days with peaks at 25 days.

How: release limitation via the rope and implementation of a buffer validated by simulation.

Impact: lead time -30%, WIP -25%, shipped throughput stable then +8% after stabilization.

Key assumption: release discipline maintained.

VII- Turning the method into executive decisions: growth, capacity, and ROI

An executive committee invests based on a quantified trajectory: truly sellable throughput, inventory level, and operating expenses.

The approach from the Theory of Constraints links these three levers directly to EBITDA (Earnings Before Interest, Taxes, Depreciation and Amortization, earnings before interest, taxes, depreciation, and amortization).

It limits reflex CAPEX decisions that look reassuring on paper but do not increase shipments.

Throughput, inventory, and operating expense: the financial reading that avoids “reflex” CAPEX

Treat WIP as tied-up cash. A gain outside the constraint can increase working capital requirement and degrade cash. Simple example: +€500k average WIP (material + value-added) at 8–10% cost of capital is €40–50k/year burned in carrying costs, not counting obsolescence and space.

Capacity investment decisions: where to put €1 to create €3 of margin

An investment is judged by its impact on the real constraint and the risk of transfer.

Classic case (anonymized): €2M CAPEX on a false bottleneck → €0 incremental throughput, zero ROI, +depreciation and often +WIP because upstream “feeds better.”

Opposite case: €120k tooling + re-organization at the constrained station → +6% shipped throughput in 8 weeks, with ROI < 6 months when unit margin follows.

Simulation helps estimate incremental throughput, and therefore associated margin, before deciding.

Constraints and Sales & Operations Planning: align demand, capacity, and promises

Industrial and sales planning (PIC, plan industriel et commercial) becomes realistic when it integrates the constraint by product family. At the S&OP (Sales and Operations Planning, sales and operations planning) and S&OE (Sales and Operations Execution, sales and operations execution) levels, protect the drum and stabilize the sequence. A frequent executive mistake: promising on “average” capacity; at 90–95% utilization, one disturbance is enough to blow up lead times (Kingman / Factory Physics).

VIII- Multi-site chain: avoid transferring the constraint from one plant to another

Optimizing one site without modeling the end-to-end flow can shift the constraint and degrade group performance. Map physical and information flows, measure effective capacity at each point, and identify the drum by flow—not by org chart. Governance must align local priorities with global flows.

Govern inter-site buffers: stock, time, and escalation rules

Between sites, a buffer can be stock, time, or flexible capacity. The choice depends on variability and shortage cost. Define clear escalation rules to avoid each site optimizing its own survival at the expense of the global flow.

Mini supply chain case: stabilize global throughput by controlling one point

What: chronic delays between sites despite high inventories. How: identification of the downstream control station as the drum, aligned upstream release rule, and an inter-site time buffer. Impact: service level +12 points, total WIP -18% without adding capacity. Conceptual source: TOC approaches and inter-site synchronization studies (Theory of Constraints).

IX- Lean, Six Sigma, and constraints: when to combine, when to choose

The useful question is about order and where to apply. Lean (lean manufacturing) quickly reduces local waste but, outside the constraint, it can increase WIP.

Six Sigma reduces variability and delivers strong ROI when applied at the limiting point.

The Theory of Constraints helps decide where to focus those efforts.

What lean improves fast—and what it can degrade without a pacing point

Lean reduces motion, overstock, and changeover time. If gains don't serve the constraint, they can increase releases and WIP. Combine lean and drum-based control to make gains last.

What Six Sigma stabilizes—and why it should be decided at the bottleneck

Six Sigma targets variability and defects through statistical methods. Its maximum impact is at the drum. Outside the constraint, ROI can be low if the effort doesn't affect sellable output.

X- The 9 rules: rule, common mistake, and countermeasure

1. Balance flow, not capacities.
A plant that looks “well balanced” on paper can ship less than a plant intentionally unbalanced around the bottleneck.
Mistake: equalizing utilization until non-constrained stations saturate and create useless queues.
Countermeasure: keep slack on non-constraints to absorb disturbances (quality, setups, micro-stops) and protect the drum—i.e., the actually deliverable pace.

2. Utilization of a non-constraint is not a local decision.
When a non-constrained station “runs to amortize,” it mostly creates intermediate stock and tied-up cash.
Mistake: “run to amortize.”
Countermeasure: release according to the rope (rope), i.e., the drum need and WIP limits (Work In Progress, work-in-progress), not available capacity.

3. Activation is not utility.
Producing more only has value if the customer receives more, faster, with the right quality.
Mistake: confusing produced parts with sellable parts.
Countermeasure: manage by shipped throughput, WIP, and lead time (traversal time), then arbitrate actions that improve this trio.

4. An hour lost at the bottleneck is lost for the system.
Every minute of stop on the constraint translates into less sellable throughput, rarely “catchable up.”
Mistake: letting the drum wait for parts, tools, or support.
Countermeasure: priority support to the drum (maintenance, quality, methods) and upstream preparation to avoid starvation.

5. An hour gained outside the bottleneck is often an illusion.
Speeding up upstream without release discipline can improve a local indicator and degrade global lead time.
Mistake: speeding up upstream without subordination.
Countermeasure: reduce releases, smooth arrivals to the drum, and remove artificial urgencies that consume real time.

6. Transfer batch ≠ production batch.
Transferring only in full batches increases downstream waiting and extends lead time.
Mistake: transferring only full batches.
Countermeasure: create sub-batches that feed the drum earlier, while keeping a production batch adapted to changeovers.

7. Production batch should be variable, not fixed.
A single batch size imposed by the ERP (Enterprise Resource Planning, integrated management software) freezes control and ignores real buffer status.
Mistake: single ERP batch size.
Countermeasure: adjust batch size based on the current constraint, variability, and buffer level (green/orange/red).

8. Bottlenecks govern throughput and WIP.
Planning each station “as best as possible” creates conflicting priorities and cascading replans.
Mistake: planning stations independently.
Countermeasure: plan the drum first, then subordinate other resources, with a stable sequence over a short window.

9. Planning must include all constraints.
Plans built on averages hold until the first disturbance—then the shopfloor goes into firefighting mode.
Mistake: average plan ignoring disturbances.
Countermeasure: test via simulation, integrating product mix, breakdowns, setups, and cycle dispersion to choose robust rules, not “pretty” ones.

XI- The five deadly traps of “anti-flow” control

1. Local optimization that slows the global
   Problem: an improvement outside the constrained station (e.g., +10% upstream speed) mostly creates work-in-progress and lengthens queues. Typical result: shipments nearly flat, but WIP rising and lead times drifting.
   Countermeasure: subordinate local targets to the drum (drum (reference pace)) and manage by flow: shipped throughput, WIP, and lead time (traversal time).

2. KPIs that push overproduction
   Problem: OEE (Overall Equipment Effectiveness) and “successful” machine hours push teams to produce to keep resources busy—even if the customer does not receive more. This bias increases tied-up cash and creates artificial urgency.
   Countermeasure: replace local KPIs with a trio readable both in the executive room and on the shopfloor: truly shipped throughput, WIP level, and traversal time.
   
   

3. Poorly governed buffers
   Problem: a buffer stock that is too small starves the constraint; too large drowns the shopfloor and hides sources of variability (breakdowns, setups, quality). In both cases, deliverable pace becomes unstable.
   Countermeasure: size buffers from measured disturbances (cycle dispersion, downtime history, scrap) and manage them with alert thresholds (green/orange/red) and standardized actions.
   
   

4. Investing too early
   Problem: buying a machine (CAPEX, capital expenditure) before exploiting the constraint often means paying to move the problem. You increase theoretical capacity, not sellable throughput.
   Countermeasure: first exploit (reduce losses, changeovers, defects at the limiting station), then subordinate (release rules). Only then validate elevation via simulation to quantify ROI (return on investment).
   
   

5. Undetected constraint shift
   Problem: after a successful action, the limiting point migrates (product mix, variability, planning) and performance drops with no obvious explanation. You celebrate a “win” while the bottleneck changes address.
   Countermeasure: set a routine to re-qualify the constraint (test + data) and a flow dashboard linking shipments, WIP, and traversal time to spot the “jumping” bottleneck.

XII- Direct answers: definitions, benchmarks, checklist, and team training

• Operational definition: the Theory of Constraints aims to increase system performance by identifying what limits shipped throughput, then organizing control around that point. Shopfloor benchmark: if improving a station does not increase sellable output, that station was not the constraint over the chosen horizon. Diagnostic checklist: measure WIP, blockages, starvation, stops, setups, and scrap, then connect these data to shipped throughput.

• Useful training: diagnostic practice, release rules, and daily control. Teach constraint qualification (capacity, market, policy) and the coupled reading of throughput–WIP–lead time. Add a simulation module, because scenario testing produces more robust decisions than averages.

• 10-working-day implementation checklist:

• day 1–2 scope the flow, product families, and perimeter boundaries (workshop, line, plant);

• day 3–5 shopfloor measurements (real cycles, stops, setups, scrap, upstream starvation, downstream blockage) and ERP extraction (Enterprise Resource Planning, integrated management software);

• day 6 proof by facts: identify the constraint and run a falsification test (if we help this station, do shipments increase?);

• day 7 drum exploitation plan: priorities, short sequence, focused maintenance/quality actions;

• day 8 release rule and WIP limits (Work In Progress, work-in-progress) by area, to stop “producing to produce”;

• day 9 initial buffer sizing (buffer stocks) in time or pieces, based on measured disturbances;

• day 10 daily routine and flow dashboard (shipped throughput, WIP, lead time (traversal time), sequence adherence).
  Success criterion: more stable shipped throughput, lower WIP, decreasing lead time.

Conclusion — Theory of Constraints: manage the nomadic bottleneck, reduce WIP, and increase shipped throughput

The Theory of Constraints doesn't reward “being busy”: it measures what actually leaves the factory—what gets shipped.

Without control via the drum and the rope, work-in-progress accumulates, working capital requirement worsens, and lead times stay high—even if each station shows “good” indicators.

The lesson is simple: a plant can run at 100%… and still not ship more. The lever is flow synchronization: exploit the constraint, subordinate the rest with release rules, govern buffer stocks, and only then elevate if sellable throughput increases.

Dillygence uses its digital twin to validate Drum – Buffer - Rope logic, arbitrate exploitation vs capacity increase, and quantify the impact on shipped throughput and WIP before any shopfloor change.

FAQ — Theory of Constraints

What is the Theory of Constraints?

A management method that focuses efforts on the point that limits shipped throughput and organizes the system around that point. It prioritizes flow synchronization and WIP reduction over local optimization.

What are the 5 steps of the Theory of Constraints?

Identify the constraint, exploit the constraint, subordinate the rest to the drum, elevate the constraint, and repeat. The order prevents unnecessary spending and maximizes impact on sellable throughput.

What are the 9 rules of the Theory of Constraints?

The nine rules remind that flow matters more than balancing capacities, that local activity is not the same as system utility, and that an hour lost at the bottleneck costs the whole system. They guide batch management, planning, and buffer sizing.

What are the 3 types of constraints?

Internal capacity constraint, market constraint, and policy constraint. The countermeasure depends on the type and time horizon.

How do you apply the Theory of Constraints to a multi-site chain?

Map flows by product family, identify the global flow drum, define inter-site buffers and escalation rules. Synchronization prevails over local optimization.

How does the Theory of Constraints help reduce WIP and working capital requirement?

Limiting releases to what the drum can absorb via the rope and WIP limits reduces WIP and therefore the working capital requirement (BFR, Besoin en fonds de roulement). Result durability depends on buffer governance and respecting priorities.

How does the Theory of Constraints improve Sales & Operations Planning?

It aligns the PIC (plan industriel et commercial) with effectively deliverable capacity by integrating the constraint and variability. At S&OP, it reduces commercial over-commitment and frequent replanning.

How does the Theory of Constraints guide capacity investment decisions?

It forces funding what increases sellable throughput and requires exploiting then subordinating before elevating. Simulation compares OPEX and CAPEX and provides quantified ROI and a risk scenario.

How do you use the Theory of Constraints to reduce lead time on the shopfloor?

Reducing lead time goes through lowering WIP. Limiting releases via the rope and protecting the drum via a calibrated buffer reduces waiting. Sequence stability reduces emergencies, setups, and rework, further shortening traversal time.