Throughput

Throughput is Goldratt’s name for what a system actually achieves at its goal: the rate at which resources move all the way through to the end, like a factory turning raw materials into sold products. It is the global measure of success, distinct from local activity. In The Goal, Goldratt pairs it with inventory and operating expense as the three measures that matter, and argues that improvements should be judged by whether they raise throughput, not by whether they make some workstation busier.

The central insight is that only the constraint limits throughput. Speeding up a non-constraint station cannot raise the system’s output; it merely produces work-in-progress that waits on the bottleneck. So optimizing away from the constraint is wasted effort — sometimes worse than wasted, since it adds clutter and risk.

CF takes throughput out of the factory and uses it to define “good enough”. A factor is good enough when it no longer constrains throughput toward the goal; beyond that point it has excess capacity and small changes to it should be ignored. This grounds CF’s attack on weighted-factor analysis and credences: a method that nudges its score whenever any local factor improves is registering changes that do not increase throughput, and is therefore recording an error rather than added precision. The right move is to find breakpoints — where a quantity crosses into making a qualitative difference — and otherwise optimize attention and goal-success rather than every dimension.