Constraint Analysis

Origin. Goldratt's theory of constraints; linear-programming duality.

Mechanism. Distinguishes binding from non-binding constraints. Only a binding constraint has a shadow price; relaxing a non-binding constraint changes nothing. Systems have few binding constraints at a time, and improvement anywhere else is measurement error.

Procedure. Enumerate constraints. For each, ask what changes if it is relaxed by ten percent. Nothing changes → non-binding, ignore. Rank the rest by shadow price. Attack the top one. Then re-identify, because the binding constraint moves.

Applies to. Throughput, capacity, cost, and any resource-limited system.

Limitations. Local optimization elsewhere feels productive and is measured; the discipline is refusing to do it.