January 20th, 2011

Recently was given an axiom which turns out to be folks engineering wisdom, namely:

Optimizing a system requires sub-optimizing the subsystems. Conversely optimizing a subsystem sub-optimizes the global system.

Chewing on that as it relates to technology and organizations.


(In particular I was thinking about the related corollary this afternoon, namely, “We work hard to be able to be this dumb.”)

7 responses to “Sub-optimization”

  1. Jon Stahl says:

    Worse is the new better. Good enough is better than perfect.

  2. Andy says:

    You could have added “… in the short term” to the axiom’s last sentence. Am thinking about it in relation to software development and to training for bike racing.

  3. “Optimizing a system requires sub-optimizing the subsystems.”


    Seems so self-evident I want the tee-shirt. In queuing you narrow (de-optimise) stages to maximise overall thruput, I imagine there’s a similar result for a decomposed system.

  4. Kellan says:

    Well I work about 10ft from a screen printing lab these days, maybe we could arrange shirts.

  5. ryan king says:

    I think the initial quote is just a fancy way of saying “every optimization has a tradeoff”.

  6. Ben says:

    The general rule is that you can’t optimize along all dimensions at once. You’re trying to move to one stable point, reevaluate among your closest set of stable points, and then choose another. It’s only teleological to the extent that your goals and values determine the direction in which to move. Oh, and be careful: at night the manifold weasels come out.

  7. Have you been reading Eliyahu Goldratt lately? Optimizing subsystems is a great way to kill otherwise good companies and it’s one of the things that makes MBAs so dangerous to organizations.

    I suggest reading ‘The Goal’ for a very approachable introduction to the Theory of Constraints and related thinking.