Most constraints get elevated before they are ever exploited

Eliyahu Goldratt set out the five focusing steps in The Goal in 1984, and they have been quoted ever since: identify the constraint, exploit it, subordinate everything else to it, elevate it, and then go back to the start, because the constraint will have moved.

The trouble is that they are quoted as a list, and a list invites you to work down it. In practice most improvement efforts identify the constraint correctly, glance at step two, and land on step four, because step four is the one that comes with a capital request and a project plan. Buy another machine. Hire another two people. Add a shift.

Step two is where the money is, and it usually costs nothing.

The line

Five stations. Somebody has measured what each one can do in an hour when it is running.

Weld is the constraint, and identifying it took one morning with a stopwatch. This is step one, and it is genuinely the easy step: an hour of the constraint is an hour of the whole line, and every other station’s spare capacity is worth exactly nothing.

Which is where the first decision gets made badly. Thirty eight an hour is what weld produces. It is not what the welder can do. The welding itself takes one minute twelve seconds a unit, which is fifty an hour. The gap between fifty and thirty eight is not skill and it is not the machine. It is the shift.

---

	Shift length, eight hours	480 min
	Setups, four at ten minutes	40 min
	Unmanned while the welder is on a break	30 min
	Waiting for material from form	25 min
	Reworking units rejected at finish	20 min
	Time the constraint actually spends welding	365 min

At one minute twelve seconds a unit, 365 minutes is 304 units a shift, which is the 38 an hour the line reports. The measured rate and the loss account reconcile, and if they had not, one of the two would have been wrong.

That reconciliation matters more than the individual figures. A loss account that does not add back up to the observed output is a loss account with something missing from it, and the missing item is usually the largest one.

Step two, in full

Exploit means one thing: get every unit you can out of the constraint without spending anything. Not “run it harder”. Stop wasting it.

Look at the four losses and ask which of them are the constraint doing work.

Setups, forty minutes. Real work, and reducing it is a genuine project. Leave it for now.

Unmanned over breaks, thirty minutes. The welder takes their breaks when everyone else does, so the constraint stops when the canteen opens. Staggering the break rota so the station is covered costs a conversation. Thirty minutes.

Waiting for material, twenty five minutes. Form has spare capacity, so this is not a supply problem, it is a sequencing one. Kitting the next run ahead of time, which form has the hours to do, costs nothing. Twenty five minutes.

Rework, twenty minutes. The constraint is welding units twice because they failed inspection at finish. Worth pursuing, but the cause sits at cut and form and the fix is not immediate. Leave it for now.

---

	Time welding today	365 min
	Recovered by covering breaks	30 min
	Recovered by kitting material ahead	25 min
	Time welding, at one minute twelve a unit	420 min
	Line output a shift, was 304	350 units

Fifteen percent more from the whole line, for a break rota and a kitting change. No capital, no headcount, and the two changes that were left alone are still available.

Thirty eight an hour becomes 43.75. Nobody bought anything.

Step three is the one that gets resented

Subordinate everything else to the constraint. In plain terms: the other four stations stop being measured on their own output.

Cut can do sixty two an hour. Over a shift that is 496 units against a line output of 350. If cut is told to run flat out, because its supervisor is measured on utilization, it will produce 146 units a shift that nothing downstream can take. Those units become a pile in front of form, then a pile in front of weld, then a working capital line somebody will eventually ask about.

The pile is not a symptom of cut working well. It is the visible form of a measurement error, and it will continue for exactly as long as cut is rewarded for making it.

This is the step that costs political capital rather than money, and it is why so many improvement efforts stall between step two and step four. Nobody minds spending money. People mind being told their number no longer counts.

Step five is not a footnote

Go back to the beginning, because the constraint has moved.

At 43.75 an hour, finish is running at forty seven. The line has bought itself 7.4 percent of headroom and no more. Do the rework project as well, and weld goes past forty seven and the constraint becomes finish, at which point every rule on this page applies to a different station and the break rota that mattered so much last month is worth nothing.

Goldratt’s own wording for step five is the useful part, and it is the part that gets dropped from the summaries: do not let inertia become the constraint. The rules you wrote to protect weld will still be in force when weld is no longer the problem, and rules outlive their reason quietly.

The constraint is often not a machine

The example above is a physical station because a physical station is easy to draw. In most processes, and in nearly every service process, the binding constraint is a policy.

A rule that welded parts move downstream only in batches of twenty five adds waiting to every part in the batch and shows up as a capacity problem at finish. A rule that a manager must approve anything over a threshold turns one person’s diary into the constraint on a whole department. A rule that reports are produced monthly makes every decision that depends on them monthly too.

Those constraints do not appear in a capacity study, because nothing in a capacity study measures a rule. They appear as waiting, which is why the first thing worth doing on a process nobody has studied is not a capacity survey at all: it is separating the time work spends being worked on from the time it spends waiting between steps. Whatever holds the biggest queue in front of it is the constraint, whether it is a machine, a person, or a sentence somebody wrote in a procedure four years ago.