Sam Gentle.com

Worse is slower

There's a trope you see sometimes in movies or TV shows, where a Technical Dweeb is doing a very important technical thing for the plot-dictated imminent deadline. "Damnit", the Important Leader Who Tells People What To Do says, "this is taking too long! Isn't there any way to go faster?" – "W-w-well, if we bypass the safety systems and reroute the engine power through this coathanger, but that would..." – "Just do it, Technical Dweeb! There's no time!". And, of course, everything works out fine.

There is a scary implication there, beyond just that coathangers are not rated for engine power, which is that there is a neat and clear negative relationship between speed and quality. That is, if you want better quality, it will take more time, and if you want it faster you can just sacrifice quality. Dangerously, it is somewhat true, in the sense that there are situations where you can clearly point to a speed/quality tradeoff. And, if you have a single event and the consequences of insufficient quality on that single event are acceptable, then it can be a good trade to make. But you have to be very careful with those assumptions.

For example, if you make screwdrivers, and each screwdriver needs to undergo a careful and lengthy sequence of heating and cooling cycles to be tough but not brittle, that's a clear speed/quality tradeoff. You could just skip that step and have a crappy screwdriver that either shatters or bends when you use it. Maybe it normally takes 8 hours, and this way you can get it down to 4 but it'll probably break the first few times you use it. That would make sense if someone wanted a screwdriver in half the time, and they only need it for ceremonial reasons, or they're giving it to someone they don't like. That's a single event where the consequences are acceptable.

Now imagine that you needed to use a screwdriver multiple times, but you're on a really tight deadline. You get the four-hour crappy screwdriver, use it a few times, and it breaks. Now you need a new screwdriver, but you're still on a really tight deadline. So you order another four-hour special. Of course, it breaks again. You're now up to 16 hours. In the single-event case, the crappy screwdriver would have saved you time, but in this case it's cost you time. If you did it right to begin with, you'd be 4 hours better off. Keep in mind, this didn't require taking some far-future view where the screwdriver had to last a thousand years, this showed up after the event repeated itself a few times.

But things get incomparably worse if you work in a screwdriver factory. You make screwdrivers, but you also use screwdrivers in their construction. Now you have a clever idea. Instead of selling crappy screwdrivers, you save time by using crappy screwdrivers internally. This is the same problem as before, but much worse; you're making bad tools with more bad tools. The bad quality slowdown becomes multiplicative: each crappy screwdriver has to be replaced more often, which means you need to make more, which puts more wear on the crappy screwdrivers you use to make them. It's a kind of endless logarithmic crap spiral.

So these two factors, the amortised cost of having to replace low quality work over time, and the multiplicative cost of bad tools, both point to a very different relationship between speed and quality. In some cases, the relationship can be positive instead of negative, and even exponentially so. That is, you can be faster and better, or slower and worse.

The stereotype is that quality is a nice ideal but that it must yield, in practice, to lower quality but faster work. However, I think that is backwards. It's actually the speed-quality tradeoff that is idealistic, and the practical reality is that cutting corners often leaves you worse off than you would be otherwise. That's not to say you can't save time by doing things more efficiently, or by doing less of them, but that you probably won't save time by doing things worse.