Evolutionary Design

Jason Kottke writes about the hesitation to change a system that works. The problem with true innovation is that it requires a complete tear-down of the status-quo, and a brand new start in a completely different direction.

The best metaphor I can think of is described in the concept of a fitness landscape (specifically, this diagram). The three points A, B, and C represent three possible “ideal” designs for a given site (say, for the sake of consistency, eBay). Each is relatively ideal compared to the valleys on either side, but C is slightly “more ideal” than A, and B is clearly “more ideal” than either A or C.

If a particular design begins in the valley between A and B, there is a chance that it will “evolve” into the ideal at point B. There is also a chance that it will “evolve” into the ideal at point A. Which way it goes is heavily dependent upon some of the first choices that are made at the outset; the earliest design decisions will put a design on an irreversible course toward one or the other. If the goal of engineers and/or designers is to make the product “better” at each iteration, there is no way for the design to backtrack toward what might be a “more ideal” situation.

Once a peak is reached, there is nowhere else to go; that design has reached its absolute pinnacle. It cannot be improved. Let me repeat that, because it’s tremendously important: It cannot be improved. It is, effectively, dead.

So say a design has reached the peak at point A. In order to get from that ideal to the optimal ideal, point B, all the space between needs to be traversed. This goes against everything that engineers/designers/whoever are taught: to make it better, it’s gotta get worse. For a successful organization (eBay, DePodesta’s Cleveland Indians), it simply will not happen.

In real life, the fitness landscape isn’t even two-dimensional: it’s three-dimensional, and most possible directions do not lead to even a relative ideal, let alone the “best” ideal. Many directions are complete dead-ends, and there will probably be a period of “wandering in the wilderness.”

The only reason DePodesta was able to reshape the A’s so completely is because, when he got there, they had already basically bottomed out: there was nowhere to go but up. For the same reason, established systems (like eBay’s design) won’t ever really be able to innovate while they’re successful. This is the best argument for periodic overhauls in site design: some things will work, some things won’t, but there’s a lot to be learned from it.

February 12, 2004 - 10:59 PM