Tuesday, April 01, 2008

Complexity Reduction: I've Heard This Before

I think I've heard this pretty much every year in the automotive press, especially from Ford. The domestic auto manufacturers (in general) offer far more permutations of each of their vehicles for sale -- orders of magnitude more than some of the Japanese makes. If you try to configure a Honda or a Toyota, you'll find your choices limited to a very small number of broad packages ('sport package' and so on), and that's about it. The US manufacturers offer packages, but also offer a ton of options entirely standalone (including many that are not optional in or out of a package from other OEMs), as well as many that can be obtained via standalone or package content.

The net result of this product offering strategy is that the US manufacturers offer for sale (and thus must be able to build) many more variants of their vehicles than their foreign competitors, leading to additional complexity in the manufacturing arena. Every so often (every new CEO, every new PD head, etc.), the US manufacturers say that they are going to reduce the complexity of their offerings (more closely mimicking the Asians) as a route to cost savings via reduced a engineering and manufacturing burden. The referenced link shows that they're talking about it again; they are targeting a 99% reduction in the offered permutations ('buildables') of the meaningful features on the vehicle for the 2009 Focus, for example.

I'm sure some would view this as 'bad', since it is reducing the choices available to customers of Ford/GM/Chrysler. However, just because those choices are available does not mean that people actually make use of them. Just because the F-150 is offered in a zillion ways, there really a small handful of common configurations that people ordering the F-150 as a work truck end up settling on. Similar clusters of configurations hold true for other use cases for the F-150 (and for every other such vehicle).

Back when I worked on a team dealing with Ford's global product definition (1999-2007), this was definitely the case. I was able to empirically verify this for specific vehicles, using data and configuration functionality available within our system. One quick study done out of curiosity used data for the Transit (a European market commercial van; you can see Sabine Schmidt drive one around the Nurburgring at this YouTube video). I had our system enumerate all buildable combinations of the 'defining features' (all of the important feature families in the vehicle: 'transmissions' is one such feature family, for example...'engines' is another) on the vehicle. Then, we took historical sales data (represented by a list of specific vehicle combinations for that given calendar year) and binned each of those against the theoretical possible buildables. The finding: something like 99%+ of the buildables had no sales volume. Of the tiny percent that had any actual sales associated with them, a tiny handful of that tiny handful accounted for nearly all of the sales. Being a commerical vehicle (commerical vehicles tend to have more options) offered in many markets, the Transit has more buildables and thus more complexity than most. However, the same story is true for all of the vehicles: more choices are offered than actual customer purchases seem to ever merit.

No comments: