K
KnowMBAAdvisory
OperationsIntermediate7 min read

Theory of Constraints

Theory of Constraints (TOC), developed by Eliyahu Goldratt in his 1984 book The Goal, says that every system has exactly ONE bottleneck at any given time — and the throughput of the entire system equals the throughput of that bottleneck. Improving anything that isn't the constraint is a waste of money. The five focusing steps: (1) Identify the constraint, (2) Exploit it (squeeze max output from existing capacity), (3) Subordinate everything else to it, (4) Elevate the constraint (add capacity), (5) When the constraint moves, repeat. A factory that doubles the speed of every machine EXCEPT the bottleneck still ships the same number of units — but now has more inventory piling up.

Also known asTOCGoldratt's TheoryBottleneck TheoryConstraint ManagementThe Goal Method

The Trap

Teams optimize the wrong station because it's easier or more visible. The CEO buys a faster CRM because the sales team complained, but the actual bottleneck is the 3-week onboarding process — so faster sales just means more deals stuck in onboarding limbo. Local efficiency metrics ('utilization at every workstation') actively HARM throughput by encouraging non-bottleneck stations to over-produce, creating WIP that buries the real constraint. If you don't know which step is the bottleneck right now, every 'improvement' is a coin flip on whether you helped or hurt.

What to Do

Walk the value stream end-to-end. The bottleneck is the station with the largest queue in front of it (literally — inventory piles up before constraints). Once identified, do this in order: (1) EXPLOIT — make sure the bottleneck never starves and never works on defects; stage a buffer in front of it, run it through lunch breaks, do quality checks BEFORE the bottleneck so it doesn't waste time on garbage. (2) SUBORDINATE — set the pace of every other station to match the bottleneck (no more, no less). (3) ELEVATE — only AFTER exploiting, invest in more capacity at the constraint. Then re-identify, because the constraint will move.

Formula

System Throughput = Throughput of the Bottleneck (everything else is noise)

In Practice

When Eliyahu Goldratt consulted with a heating-element manufacturer in the 1980s (the basis for The Goal), the plant was missing 90% of its delivery dates with millions in inventory and a 'we need more machines' culture. He found the constraint was a single $250K furnace. Instead of buying another, they: stopped scheduling lunch shutdowns on the furnace, moved QA inspection to BEFORE the furnace (so it never melted defective parts), and cut batch sizes everywhere else. Throughput rose 40% in 3 months without any capital expenditure. The 'we need more machines' problem turned out to be a 'we waste the machine we have' problem.

Pro Tips

  • 01

    Goldratt's rule: 'An hour lost at the bottleneck is an hour lost for the entire system. An hour saved at a non-bottleneck is a mirage.' Stop measuring utilization at non-constraints — it incentivizes the wrong behavior.

  • 02

    Use Drum-Buffer-Rope scheduling: the bottleneck (drum) sets the pace, a small inventory buffer keeps it fed, and a 'rope' (signal) tells upstream stations to release work only when needed. This prevents WIP explosion.

  • 03

    In knowledge work, the bottleneck is almost always a person, not a process. The senior engineer who reviews every PR, the legal team that approves every contract, the CEO who signs every offer letter. Find the human queue and either delegate, batch, or invest in their capacity.

Myth vs Reality

Myth

If we improve every step by 10%, the whole system improves by 10%

Reality

False. The system improves by however much the bottleneck improved, period. Improving non-bottlenecks just creates more inventory and longer lead times. This is the single most counterintuitive result in operations.

Myth

More automation always increases throughput

Reality

Automating a non-constraint accelerates the rate at which work piles up at the constraint, making lead times WORSE. Automation only helps throughput if it's at the bottleneck — and even then, it might just move the constraint elsewhere.

Try it

Run the numbers.

Pressure-test the concept against your own knowledge — answer the challenge or try the live scenario.

🧪

Knowledge Check

Your SaaS onboarding pipeline: (1) Sales hands off in 1 day, (2) Solutions Engineer configures in 3 days, (3) Customer Success Manager kicks off in 8 days (queue), (4) Training delivered in 2 days. The CEO wants to invest $200K. Where should it go?

Industry benchmarks

Is your number good?

Calibrate against real-world tiers. Use these ranges as targets — not absolutes.

Throughput Improvement After TOC Implementation

Manufacturing and service operations applying the five focusing steps

Strong

30-50% in 6 months

Typical

15-30% in 6 months

Weak (likely wrong constraint identified)

< 10%

Source: Goldratt Institute case data / APICS

Real-world cases

Companies that lived this.

Verified narratives with the numbers that prove (or break) the concept.

✈️

Boeing (787 Dreamliner Final Assembly)

2010-2013

success

Boeing's 787 program was missing delivery commitments by 3+ years and burning $1B+/quarter. Internal teams insisted the bottleneck was supply chain (electrical, lithium batteries). Application of TOC analysis showed the actual constraint was the final assembly station in Everett — specifically, the bay where systems integration happened. Boeing applied exploit-first thinking: pre-staged parts kits at the constraint, ran 24/7 instead of two shifts, moved quality inspections upstream so the constraint never wasted time on rework. Production rate climbed from 2.5 to 10 aircraft/month over three years.

Production Rate (2011)

2.5/month

Production Rate (2014)

10/month

Capex on Final Assembly

Minimal — exploit before elevate

Backlog Reduction

~$50B in deferrals cleared

The most expensive resource is rarely the constraint people first blame. Walk the line and look for the queue.

Source ↗
🛠️

Hypothetical: Mid-Market SaaS Implementation Team

Recent

success

A 200-person SaaS company was missing every Q2 implementation target. Leadership was about to hire 12 more implementation consultants at $180K each ($2.2M annual cost). A TOC walkthrough revealed the constraint wasn't consultants — it was the single Solutions Architect who had to design every new customer's data model. Queue in front of him: 6 weeks. They split his role into 3 specialized SAs by industry vertical and built a self-serve template library to handle 60% of cases. Cycle time dropped from 11 weeks to 5 weeks WITHOUT hiring 12 consultants.

Avoided Hiring Cost

$2.2M/yr

Cycle Time

11wks → 5wks

SA Capacity

1 → 3 (effectively 5x with templates)

Customer NPS

+22 points

When you instinctively want to hire more of the visible role (consultants), check if the invisible role behind them (the architect everyone routes through) is actually the constraint.

Decision scenario

The Manufacturing Bottleneck Investment

You're COO of a precision-parts manufacturer. Quarterly throughput is flat at 2,400 units despite year-over-year demand growth of 18%. The CEO has authorized $500K capex. The plant manager wants to upgrade the cutting machines (currently 92% utilized). The shop floor lead says the polishing station has a 10-day backlog of half-finished parts.

Throughput

2,400 units/qtr

Demand Growth

+18% YoY

Capex Available

$500K

Polishing WIP

10-day backlog

Cutting Utilization

92%

01

Decision 1

You walk the floor. Cutting machines are humming at 92%. Polishing has a wall of half-done parts queued up. The polishers work through lunch and stay late. The cutting team takes long breaks because their output piles up downstream. Where do you put the $500K?

Upgrade the cutting machines — 92% utilization is high and the plant manager has experience justifying this kind of capexReveal
You buy faster cutting machines. Cutting capacity rises 30%. Throughput rises 0%. The polishing backlog grows from 10 days to 14 days because more work is piling up faster. Lead times get worse. Customers escalate. You spent $500K to make the symptom more painful. The classic Goldratt mistake: improving a non-constraint accelerates the rate at which the real constraint chokes the system.
Throughput: 2,400 → 2,400 units/qtrPolishing WIP: 10 days → 14 daysCapex Wasted: $500K
First exploit polishing (move QA upstream so polishers never touch defective parts, add a third shift); then use the $500K to add a second polishing lineReveal
Exploitation alone (QA upstream + third shift) lifts polishing throughput 35% in 6 weeks at almost zero cost. Then $500K on a second polishing line elevates capacity another 80%. System throughput rises from 2,400 to 3,800 units/qtr — well ahead of the 18% demand growth. The constraint moves to assembly, which is your next focus. You followed the five focusing steps in order.
Throughput: 2,400 → 3,800 units/qtrPolishing WIP: 10 days → 1.5 daysConstraint: Polishing → Assembly (next cycle)

Related concepts

Keep connecting.

The concepts that orbit this one — each one sharpens the others.

Beyond the concept

Turn Theory of Constraints into a live operating decision.

Use this concept as the framing layer, then move into a diagnostic if it maps directly to a current bottleneck.

Typical response time: 24h · No retainer required

Turn Theory of Constraints into a live operating decision.

Use Theory of Constraints as the framing layer, then move into diagnostics or advisory if this maps directly to a current business bottleneck.