Global Optimum

The global optimum is the best possible state for the system as a whole, with respect to its goal (usually throughput).

Why global optima matter

• Most improvements aren't global. They improve one part without helping the whole.
• Focus on global optima. Aim your changes there.
• Local optima look like improvements but are usually global detours.

CF's adoption

CF treats global vs. local as a recurring dichotomy:

• Bias often pushes us to local optima (helping me, not us).
• Most factors aren't borderline — the global optimum is usually at the bottleneck.
• The Five Focusing Steps exist to keep you aimed at global optima.

How to know if you're at the global optimum

• You found the constraint and fixed it.
• You found the next constraint and fixed that.
• The new constraint is no longer the original problem.
• Throughput has actually increased.

Common false global optima

• "Best of all options I've tried."
• "Best within my department."
• "Best this quarter."
• "Best for me personally."

These are usually local in disguise.

"Most improvements aren't very effective. They're optimizing local optima, but that doesn't significantly help the big picture (the global optima)." — criticalfallibilism.com