3 ms·

Up or out is a supremely effective way to drive people to meet organizational goals. On the down side, you have to be super careful about what you measure. Enro

by jfoutz 11y ago

Up or out is a supremely effective way to drive people to meet organizational goals. On the down side, you have to be super careful about what you measure. Enron made extensive use of up or out. It worked really really well, right up until it didn't. If that's really going to happen, executives are going to need to be at least as precise than an assembly programer when setting metrics.

jjoonathan 11y ago
Bad things happen when the "competition knob" is turned to 0 (everyone free-rides) and 11 (everyone cheats). > you have to be super careful about what you measure. Do you really think you can design metrics so perfect that they won't be gamed? Even if it were possible (I'm pretty sure it isn't), could you do it without undermining cooperative team dynamics that create organizational value? I'm not saying all metrics are bad, but I am saying that turning up the heat with up-or-out is the equivalent of turning the "competition knob" to 11 and that it has a large number of obvious failure modes which make me puzzle over your declaration of its efficacy (unless you meant it in the tautological sense, but that would be silly?). I'm further puzzled that you brought Enron up as an example of incentives done correctly. Most people think of them as a case-study in runaway incentives and how turning the competition knob to 11 inevitably causes, well, exactly what happened to Enron. When the choices are "cheat or get fired," your company will inevitably fill itself with cheaters. Unless your metrics are perfect, of course, but are you willing to bet your company on it?
- jfoutz 11y ago
  > Do you really think you can design metrics so perfect that they won't be gamed? Nope. I think it's super super hard. I think a lot of executives will say, oh we'll just add "Up or out" to dial up the competition knob, and then they'll self destruct. Executives make great decisions all the time. Unfortunately most of those decisions involve "close enough" thinking. Up or out introduces razor sharp edges. Everything appears better than fine, hell it appears fantastic. Until it isn't. and when it isn't it's terminal. I used assembly specifically to highlight the care that must be taken. I wanted to evoke images of bricking devices. Very few executives have the mindset that if i'm not certain i'm right, then i'm wrong. Often their thinking is squishy at the edges, and that's just the cost of doing business. I do think if it's possible to do "Up or out" effectively, that corporation would be formidable. If gaming the metric advanced the corporation's goals, that would be amazing. But, super hard. Folks selected for being the smartest guy in the room are most likely to fool themselves into thinking they can handle it.
  - jjoonathan 11y ago
    I think where we fundamentally disagree is that you believe it's generally possible to design an incentive system which is well-behaved in the asymptote of extreme competition whereas I do not, except under select conditions where the sum total of value provided by an individual is trivially quantifiable and separable from the contribution of his peers (e.g. assembly line work). "Care must be taken" simply doesn't cut it; the comparison to assembly programming is ridiculous. For most devices it's hard to screw up badly enough to brick them and even if you do you are out, what, a few hundred dollars? Also, the feedback is unambiguous and instantaneous. Screw up a big organization and you are out millions or billions with only vague, ambiguous feeback that arrives years after you institute the policy. Even a comparison to rocket science doesn't come close to capturing the difficulty involved because rockets are designed to be separable into subsystems which obey well-characterized physical laws whereas most creative processes can't be broken down so rigorously. The hypothetical perfect incentive system is like the hypothetical analog computer which gives you infinite accuracy and precision for free: it only appears to be an exciting possibility if you neglect all of the factors that limit its efficacy in practice.
    - jfoutz 11y ago
      I think it works really well in adversarial contexts. I'd guess the best is probably law. Win cases, bill hours. Lack of ethics are exploited ruthlessly by the opposition - that's grounds for appeal. Somewhat less effectively, sales. Car salesmen have horrible reputations, but lots of technical sales are smart, knowledgeable and helpful. I think the long term relationship aspect helps. Even less effective, but still clearly in use is finance. Insider trading and collusion are tough to prove, and there's not much incentive to keep the system in check. But finance clearly takes advantage of up or out. Bricking a device is a fine analogy. Sometimes the slippery slope argument is not a fallacy. I think this is one of those times. You seem to agree, we're just not clear on stakes. I think the slippery slope exists in a 2 person startup all the way up to microsoft or google. Perfect anything is a pipe dream. "It has been said that democracy is the worst form of government, except all those others that have been tried". If somebody wants to try a social experiment with their investors money, best of luck to them. It's unlikely i'd invest or be interested in working there though.

by jfoutz 11y ago

jjoonathan 11y ago

Bad things happen when the "competition knob" is turned to 0 (everyone free-rides) and 11 (everyone cheats). > you have to be super careful about what you measure. Do you really think you can design metrics so perfect that they won't be gamed? Even if it were possible (I'm pretty sure it isn't), could you do it without undermining cooperative team dynamics that create organizational value? I'm not saying all metrics are bad, but I am saying that turning up the heat with up-or-out is the equivalent of turning the "competition knob" to 11 and that it has a large number of obvious failure modes which make me puzzle over your declaration of its efficacy (unless you meant it in the tautological sense, but that would be silly?). I'm further puzzled that you brought Enron up as an example of incentives done correctly. Most people think of them as a case-study in runaway incentives and how turning the competition knob to 11 inevitably causes, well, exactly what happened to Enron. When the choices are "cheat or get fired," your company will inevitably fill itself with cheaters. Unless your metrics are perfect, of course, but are you willing to bet your company on it?

jfoutz 11y ago
> Do you really think you can design metrics so perfect that they won't be gamed? Nope. I think it's super super hard. I think a lot of executives will say, oh we'll just add "Up or out" to dial up the competition knob, and then they'll self destruct. Executives make great decisions all the time. Unfortunately most of those decisions involve "close enough" thinking. Up or out introduces razor sharp edges. Everything appears better than fine, hell it appears fantastic. Until it isn't. and when it isn't it's terminal. I used assembly specifically to highlight the care that must be taken. I wanted to evoke images of bricking devices. Very few executives have the mindset that if i'm not certain i'm right, then i'm wrong. Often their thinking is squishy at the edges, and that's just the cost of doing business. I do think if it's possible to do "Up or out" effectively, that corporation would be formidable. If gaming the metric advanced the corporation's goals, that would be amazing. But, super hard. Folks selected for being the smartest guy in the room are most likely to fool themselves into thinking they can handle it.
- jjoonathan 11y ago
  I think where we fundamentally disagree is that you believe it's generally possible to design an incentive system which is well-behaved in the asymptote of extreme competition whereas I do not, except under select conditions where the sum total of value provided by an individual is trivially quantifiable and separable from the contribution of his peers (e.g. assembly line work). "Care must be taken" simply doesn't cut it; the comparison to assembly programming is ridiculous. For most devices it's hard to screw up badly enough to brick them and even if you do you are out, what, a few hundred dollars? Also, the feedback is unambiguous and instantaneous. Screw up a big organization and you are out millions or billions with only vague, ambiguous feeback that arrives years after you institute the policy. Even a comparison to rocket science doesn't come close to capturing the difficulty involved because rockets are designed to be separable into subsystems which obey well-characterized physical laws whereas most creative processes can't be broken down so rigorously. The hypothetical perfect incentive system is like the hypothetical analog computer which gives you infinite accuracy and precision for free: it only appears to be an exciting possibility if you neglect all of the factors that limit its efficacy in practice.
  - jfoutz 11y ago
    I think it works really well in adversarial contexts. I'd guess the best is probably law. Win cases, bill hours. Lack of ethics are exploited ruthlessly by the opposition - that's grounds for appeal. Somewhat less effectively, sales. Car salesmen have horrible reputations, but lots of technical sales are smart, knowledgeable and helpful. I think the long term relationship aspect helps. Even less effective, but still clearly in use is finance. Insider trading and collusion are tough to prove, and there's not much incentive to keep the system in check. But finance clearly takes advantage of up or out. Bricking a device is a fine analogy. Sometimes the slippery slope argument is not a fallacy. I think this is one of those times. You seem to agree, we're just not clear on stakes. I think the slippery slope exists in a 2 person startup all the way up to microsoft or google. Perfect anything is a pipe dream. "It has been said that democracy is the worst form of government, except all those others that have been tried". If somebody wants to try a social experiment with their investors money, best of luck to them. It's unlikely i'd invest or be interested in working there though.