Theory of Constraints

On a small production line, one machine is always running, always has a queue in front of it, and always gets blamed when an order ships late. The obvious fix looks like buying a second one. Theory of Constraints asks a less obvious question first: is that busy machine actually the thing limiting how much the whole line can produce, or does it just look that way because it is never idle?

Theory of Constraints, developed by Eliyahu Goldratt, treats a business as a chain of connected steps and argues that the whole system's output is limited by exactly one step at a time: the constraint, or bottleneck. Speeding up any step that is not the constraint does not increase what the whole system can produce: it just creates a bigger queue in front of the real bottleneck. The first of Goldratt's five focusing steps is simply to identify the constraint, which sounds obvious and is routinely got wrong. Goldratt introduced the idea through a business novel, The Goal, published in 1984, rather than through a conventional textbook, which is partly why the analogy of a factory floor still explains the model better than an abstract description does.

• Use it whenever output is being limited somewhere in a process and it is not obvious exactly where.
• Use it before spending money on capacity (new equipment, new hires) anywhere in the process, to check that money is going to the actual constraint.
• Use it to explain to a team why "everyone work harder" rarely fixes a throughput problem on its own.
• Do not use it on a one-off project with no repeating flow; the model is built for processes that run continuously.
• Do not assume the constraint you found last year is still the constraint: fixing one bottleneck reliably creates a new one elsewhere.

One-page model visual

The one-page visual

Use this visual as a quick reference. The card in the deck adds the questions and the steps to run the model in your next meeting.

See all 50 models as cards →

The mistake is assuming that the busiest-looking step is automatically the constraint. A machine can look permanently busy simply because operators feed it constantly, whether or not the rest of the line is actually waiting on it: visible activity and true bottleneck are not the same thing, and confusing them sends investment to the wrong place.

The deeper mistake is treating the constraint as fixed. Once a real bottleneck is addressed, output rises until a different step becomes the new limiting factor: the constraint moves, it does not disappear, and a business that fixed its bottleneck once and stopped looking is usually sitting on a new one it has not noticed yet.

This is also why buying capacity at the wrong step can make a line perform worse, not better, in ways that confuse everyone involved. A faster machine feeding into an already-overloaded next step just moves the queue earlier in the process, and the person who approved the purchase is left wondering why a clearly superior machine has not improved anything.

Illustrative example: a fictional company, not a customer case.

On the production line, timing each step for a full day, rather than just observing which machine looks busiest, shows something different: the machine everyone blames does have a queue, but it clears that queue within its shift. The step immediately after it, a manual inspection stage with only one trained person covering it, is the one where work genuinely piles up overnight.

Worth checking before concluding anything: whether that inspection stage is short-staffed by design, because it was never expected to handle this volume, or whether it has quietly become the busiest step in the line as order volumes have grown since the last time anyone reviewed staffing there.

That single measurement (queue length at the end of the day, not how busy a station looks during it) redirects the entire conversation away from buying a second machine. The card takes you through the remaining steps to a decision.

• Front: what Theory of Constraints is for and why the obvious bottleneck is often the wrong one.
• Back: the numbered steps to apply the model, plus a short worked example of its own. The company in this guide is a separate illustration, not the example printed on the card.
• Digital: a Miro board for mapping the process flow and marking queue points, with a video tutorial.

• Value Stream Mapping gives a fuller picture of the whole process, useful for spotting where a constraint might actually sit.
• Kanban Method is a practical way to visualise queues building up in front of a constraint in real time.
• Pareto Analysis helps rank which delays or defects are worth investigating first once a constraint is suspected.

BizDecks is the cheat sheet for business decisions: 50 models, one card each. See the toolkits.