Aaron N. Tubbs bio photo

Aaron N. Tubbs

Dragon chaser.

Twitter Facebook Google+ LinkedIn Github

To reward me for my flattering praise of Cablevision’s concerted efforts to protect its users, they’ve decided to keep yanking my connection to the net just long enough to terminate my socket connections every 30-40 minuets for the last day or so. Oh well. Can’t have everything.

Last night I was playing Metal Gear Solid 4, and inadvertently quit without saving my progress in Act 4. This was depressing as I’d nearly completed the entire act (after about three and a half hours), and my last save is at the very beginning of the act. I’d picked up a deagle, a railgun, a 25-mm grenade launcher and 40,000 Drebin points along the way, and I really didn’t want to do it all again. Save early and save often, I guess. Oh well, I get to kill crying tiger again. And skip all the cutscenes. I’m starting to get really tired of cutscenes.

Oh, I went there. The cutscenes in MGS4 are boring and predictable. Flame on!

Whatever.

Frustrated, and well past one in the morning, I should have gone to bed. Sad thing was that I was still pretty wide awake, and I knew I wasn’t going to be able to fall asleep; I had that feeling I’d had the previous nightw hen I couldn’t sleep (was awake considering money issues, whether it makes sense to rent an apartment or not, and what my financial goals should be in the long run).

But I couldn’t sleep, in any event, so I started writing.

Three hours later, I gave up on what I was writing and went to bed. When I woke up this morning, I realized it’s something that I’ll never publish. The core of what I was debating was the theory of compensation packages and performance reviews (yes, it’s that time of year), with special attention to whether the two should be linked, and if so, how much of a link there should be. Having been exposed to several different systems over the years, I thought I was qualified to make some grand observation about the whole mess. I’ve chewed on this sort of stuff for a while, and written about it half a dozen times, if ou dig through the archives. It’s not that interesting, don’t bother.

I was able, first, to convince myself that the review process should be independent of the compensation process, and that its importance is in providing a framework for feedback, and feedback only. While the review might summarize the reasons for a compensation change, there is no link between the two. An implicit link shows up over time, but the two systems remain arbitrary and with out direct correlation.

Despite what seems like a dismissive summary, I did a good job of convincing myself that it all made sense in the end, given the expectation of a long tenure, faith in the system, and if the world turns out to be fair.

So, it was also easy to argue that the two processes should have a very explicit link. Good review means good salary, bonus, whatever. By tying money to the performance review directly, there is no ambiguity or impedance mismatch between the feedback and the paycheck.

So I started taking things to the next step: Making the entire process completely transparent and mechanical, reducing feedback, more or les, to compensation itself. While this sounds like a pretty shitty deal, it has the advantage that there are no monetary surprises, because everything is just a big algorithm; there isn’t room for the subjective. This was fun as an intellectual tool, but pretty much sucked otherwise.

I tried to factor the subjective back in by using a distributed ranking system that directly modified the previous system; this was a mess. I ended up exploring three different schemes for this, and then destroyed each with a quick game theory excercise, pointing out that each was subject to a major flaw and a suboptimal steady state, unless certain players decided to “do the right thing” rather than do the “right” thing, if you know what I mean.

You probably don’t, but that’s okay. It took me three paragraphs to explain it, and I’m not going to bother; this is all meta-eta at this point.

So I backed up, and I think, in the end, I was convinced that there should be some link. That’s about all I’d really settled on when I gave up.

Looking back over it this morning, I realized the problem was I don’t know what the right approach is. I could argue with myself forever; lots of stuff about fairness, merit, randomness, luck, subjective wellbeing, and stupid nonsense factors into it.

In the process, I developed an interesting (to me) side discussion of what it takes to be a good manager. I was reminded what a manager is supposed to do (develop people), and not supposed to do (everything else). I came up with lists of the sort of stuff that is important, and the sort of stuff that is irrelevant. I was doing this all on autopilot, until I went through the list and realized that I was, as judged by my own criteria, a really shitty manager.

Oh well, better luck next year.