STRATEGYWeeks to result

Theory of Constraints (TOC)

Every system has exactly one constraint that limits its throughput -- find it, fix it, and find the next one

Problem it solves

it, and find the next one

Best for

Operations leaders and managers of any production or service system who need to dramatically increase output without proportional increases in investment or headcount

Not ideal for

Purely creative or exploratory work where throughput is not the primary objective, or early-stage startups still searching for product-market fit

Overview

Why this framework exists

The Theory of Constraints holds that every system, no matter how complex, is limited by a very small number of constraints -- often just one -- that determines the throughput of the entire system. The central insight is that optimizing anything other than the constraint is an illusion: improving a non-constraint resource does not improve the system and may actually harm it by creating excess inventory and confusion. Goldratt demonstrates this through the analogy of a boy scout hike where the slowest boy, Herbie, determines the pace of the entire troop regardless of how fast the other boys can walk. If the fastest boys speed up, the line simply spreads out -- the equivalent of excess work-in-process inventory -- without the group arriving any sooner. Traditional management obsesses over local efficiencies: keeping every machine and every worker busy. TOC rejects this, arguing that a plant in which everyone is working all the time is grossly inefficient because much of that work produces inventory that cannot be sold. The goal of any commercial organization is to make money, and TOC provides the logic for aligning every operational decision to that goal by focusing relentlessly on the constraint.

Core principles

8 total
  1. The goal of a commercial organization is to make money -- now and in the future
  2. Every system has at least one constraint that limits its throughput
  3. Optimizing a non-constraint does not improve the system and often makes it worse
  4. A plant where everyone is working all the time is grossly inefficient
  5. Local efficiencies are mirages -- only system-level throughput matters
  6. An hour lost at the bottleneck is an hour lost for the entire system
  7. An hour saved at a non-bottleneck is a mirage -- it produces nothing of value
  8. Statistical fluctuations combined with dependent events cause inevitable inventory accumulation and throughput loss

Steps

4 steps
  1. Define the goal of your system
    Before attempting any improvement, clearly articulate what the system exists to achieve. For commercial enterprises, the goal is to make money. Every improvement initiative must be evaluated against this goal, not against surrogate measures like efficiency, utilization, or cost reduction that may not correlate with money-making.
    Pro tipGoldratt insists on asking 'Did we sell more?' rather than 'Did we produce more?' Production that does not lead to sales is not progress -- it is just inventory.
    WarningSurrogate goals like 'efficiency' and 'quality' can become detached from the real goal. Quality is important only insofar as it supports throughput and reduces operating expense.
  2. Identify the system's constraint
    Find the single resource, policy, or market condition that limits the throughput of the entire system. In a manufacturing plant, this is the bottleneck -- the resource whose capacity is equal to or less than the demand placed on it. Constraints can be physical (a machine, a person) or policy-based (batch size rules, accounting practices, market limitations).
    Pro tipWalk the floor and look for where work-in-process inventory piles up. The bottleneck is typically upstream of the largest pile. Also ask expediters -- they know which resources cause the most delays.
    WarningDo not assume you know the constraint without data. The constraint may not be the oldest or slowest machine -- it is the one where demand exceeds capacity when the entire product mix is considered.
  3. Manage the constraint to maximize system throughput
    Once identified, ensure the constraint is never idle, never working on defective parts, and never processing work that is not immediately needed. Subordinate all other resources to the pace of the constraint. Protect the constraint with time buffers so that disruptions elsewhere do not starve it.
    Pro tipAt Alex's plant, the team discovered that the bottleneck machines were sitting idle during lunch breaks and shift changes. Simply keeping them running during those times increased effective capacity without any capital investment.
    WarningResist the temptation to keep non-constraint resources busy just because they can be. This creates excess inventory that clogs the system, increases lead times, and confuses priorities.
  4. Elevate the constraint if more throughput is needed
    If exploiting and subordinating are not enough, invest in increasing the constraint's capacity. This can mean buying additional equipment, hiring more people, outsourcing constraint work, or changing processes to offload work from the constraint to non-constraint resources.
    Pro tipAlex's team brought back an old Zmegma machine and rerouted some parts to older, less efficient processes to offload the bottleneck. The 'less efficient' routing was actually more efficient from a system perspective because it freed bottleneck capacity.
    WarningWhen you successfully elevate a constraint, the constraint moves somewhere else in the system. You must immediately identify the new constraint and not allow inertia to keep you managing the old one.

Checklist

Saved in your browser

Examples

2 cases
The boy scout hike

Alex takes his son's boy scout troop on a hike. The line of boys continually spreads out despite the same average walking speed because the slowest boy, Herbie, creates a constraint. The boys ahead of Herbie walk faster and create gaps (analogous to excess inventory serving no purpose), while the troop as a whole cannot arrive faster than Herbie can walk. Alex discovers that putting Herbie at the front and redistributing weight from his pack closes the gaps and improves the group's throughput.

OutcomeThe hike becomes Alex's epiphany about dependent events and statistical fluctuations. He realizes his plant behaves the same way: the bottleneck determines plant throughput regardless of how fast other resources work, and excess inventory from fast non-bottleneck resources just spreads the line without improving arrival time.
The UniCo plant turnaround

Alex's plant is given three months before closure. By identifying the bottleneck resources (the NCX10 machine and the heat-treat oven), exploiting them by eliminating idle time, subordinating all other resources to their pace using red and green tags, and elevating capacity by bringing back retired equipment and rerouting parts, the team dramatically improves due-date performance and throughput.

OutcomeThe plant goes from shipping almost nothing on time to achieving near-perfect on-time delivery. Lead times drop, inventory decreases, and throughput increases -- all without significant capital investment. The plant is saved and Alex is promoted.

Common mistakes

3 traps
Chasing local efficiencies
Traditional management measures each department or machine by its individual utilization. This drives behavior to keep everything busy, which produces massive amounts of work-in-process inventory, extends lead times, and obscures the actual constraint. Alex's plant installed robots that increased local efficiency by 36 percent but the plant shipped less and inventories grew.
Treating all resources as equally important
An hour lost at the bottleneck is an hour lost for the entire system because the bottleneck determines throughput. An hour lost at a non-bottleneck is meaningless because it has excess capacity. Yet traditional management treats idle time at any resource as equally wasteful, leading to misallocated effort and excess inventory.
Allowing inertia after breaking a constraint
When the team successfully elevated the physical bottlenecks (oven and NCX10), the constraint shifted to market demand and then to the material release system. The policies and procedures built to manage the old constraint became the new constraint because inertia prevented the team from questioning them. Stacey's team built finished goods inventory to keep bottlenecks busy even after the constraint had moved.

Origin story

How this framework came to be

Alex Rogo, a plant manager on the verge of having his factory shut down, encounters his former physics professor Jonah at an airport. Jonah asks a series of devastatingly simple questions that expose the flaws in Alex's management assumptions. When Alex boasts about his new robots increasing efficiency by 36 percent, Jonah asks whether the plant is shipping more products, whether inventories went down, and whether the plant laid anyone off. The answers reveal that despite higher local efficiency the plant is actually performing worse. Jonah introduces Alex to the idea that the goal of a manufacturing organization is to make money, and that every action should be judged by its impact on throughput, inventory, and operating expense. Through a series of encounters and phone calls, Jonah guides Alex to discover that his plant is governed by constraints, and that managing those constraints -- rather than chasing local efficiencies -- is the path to profitability.

Source

Traced to primary
Source · BOOK
The Goal: A Process of Ongoing Improvement, Third Revised Edition
Eliyahu M. Goldratt · 1984
Open source →

Related frameworks

Browse all Strategy →