Not everything at once.
The right thing first.
A system runs as fast as its narrowest point. Improvements anywhere else cost money and change nothing. This is the logic that decides where both core systems intervene — and where they deliberately do not.
The constraint sets the throughput.
Eliyahu Goldratt formulated it for production systems in 1984: the capacity of a chain hangs on its weakest link. Strengthen a different link and you have spent money without moving anything — throughput stays where it was. The sentence sounds trivial until you apply it to a business that is optimising in five places at once.
In practice: before anything is improved, it must be established which point constrains. Everything else is activity with an invoice attached.
When the constraint decides what gets changed at the constraint.
In an owner-led business the owner is frequently the narrowest point: decisions, approvals, key relationships and critical knowledge run through one person. That differs from a machine in one fundamental respect — here the constraint has a say in which improvements happen at all.
The result is self-reinforcing. Every new task lands with the person who can decide fastest. Every relief requires time that precisely this person does not have. So the problem does not dissolve through more effort at the same point, only through a structural change set in motion from outside.
Five steps, and their order is the point.
Goldratt's five focusing steps, translated to owner dependency. The most common expensive shortcut is jumping to step four.
- 01
Identify
Which point actually constrains? Not the loudest, not the most unpleasant — the limiting one. That is a question of measurement, not of opinion.
- 02
Exploit
Whatever happens at the constraint without having to happen there has to go. Existing capacity is cleared before new capacity is bought.
- 03
Subordinate
Everything else aligns to the constraint. This is the uncomfortable step: areas that run efficiently on their own are deliberately slowed when they add load to the constraint.
- 04
Elevate
Only now do you invest — hire, outsource, buy a system. Start here and you are buying capacity for a problem not yet understood. It is the most expensive order there is, and the most common.
- 05
Repeat
Once the constraint is resolved, it sits elsewhere. The second pass usually yields less than the first because the system has adapted. Whoever does not know the loop mistakes the second constraint for a setback.
Five dimensions, unweighted.
In owner-led businesses constraints typically appear in these five dimensions. Deliberately unweighted: which of them constrains your business is exactly what a diagnosis answers — not a table in advance.
- Systemic
- Process queues, undefined handovers, breaks at interfaces.
- People
- Key persons, knowledge monopolies, no cover.
- Technical
- System breaks, performance limits, missing integration.
- Financial
- Payment terms, liquidity squeeze, stalled investment.
- Market
- Demand swings, seasonality, growth stall.
Why we do not document everything.
One rule follows from the model, and it stands as a gate condition in Build to Leave™: as little systematisation as possible, as much as necessary. The goal is not the largest number of standard procedures but the few that reduce dependency. A fully documented business in which every approval still runs through the owner has generated effort and changed nothing.
- Standards are set where they reduce dependency — not across the board.
- Measures that do not pay into the prioritised dependency are deliberately discarded.
- A second constraint is not failed work, it is the next pass.
Where the model comes from.
- BGoldratt (1984)
"The Goal: A Process of Ongoing Improvement", North River Press
The original work. The five focusing steps come from there, not from us.
- AMabin & Balderstone (2003)
"The performance of the theory of constraints methodology", International Journal of Operations & Production Management 23(5/6), 568–595
Meta-analysis of more than 80 documented applications. Median improvements: lead time around 70 %, inventory around 49 %, revenue around 63 %.
Limit: These figures come from manufacturing. They show that the methodology holds — they are not an expectation for a consulting or service business, and we do not use them as one. The logic transfers, the value does not.
It tells you how to look. Not what you will find.
The constraint logic is a thinking tool and has been public for forty years. It does not answer which point constrains your business — that requires counted figures rather than an assessment, and that is what the Diagnostic is. Get the constraint wrong and you apply the method cleanly while improving nothing.
Which point constrains yours?
The Diagnostic measures the four areas where owner dependency becomes visible and names the point where the first intervention belongs.