Audemars Piguet

Gary G
3734

• Show profile
• Show other posts

For folks interested in the broader topic of integrality vs. modularity...

Hey -- now there's a post title guaranteed to cure insomnia right there!

Seriously, though, for those who are interested in this topic I strongly recommend the research and writing of Clayton Christensen, the HBS professor who is best known for his work on disruptive technologies and invention of the term "the innovator's dilemma" but who has done some wonderful research on the topic of modularity.

Clay suggests that integrality (as a Purist I'll use his "correct" term rather than "integration," which to him refers to the assembly of modules into a whole) is, in the early days of many technology (or business) systems, the only approach that can yield the performance needed to meet minimum market requirements.  When there is no solution that is "good enough," taking an approach akin to carving the entire system from a single block of stone has the highest potential of squeezing out the last bits of performance that are needed to get over the line.  Think here of early personal computers, in which every piece was designed to mesh with every other piece.

Over time, as systems move up an improvement trajectory (think here of an experience curve), the fully integral systems get to a point at which they not only meet, but exceed mainstream market requirements.  At this point, other dimensions of "performance" begin to become both important and possible to various market segments.  Some folks want lower cost, others want design flexibility, the ability to repair one element without tearing out hard-wired connections to other ones and so on.  What happens is the advent of "plug and play" approaches, for instance, where you can get any of a variety of hard drive sizes, processor speeds, displays, graphic cards, and so on -- because the performance of individual modules is so good, the performance losses associated with the introduction of interfaces still don't pull the performance of the total system below mainstream market requirements, and various parts of the market value the new performance attributes (e.g., customization, cost, ease of repair) more than the ability of more and more refined integral systems to deliver marginally improved performance on the original performance attributes.

A critical point in all of this is that in a modular system, it's not just the quality of the modules that matters.  In fact, in many systems the quality of the interfaces is actually much more important, and poor interface placement or design, not the performance of the modules, is the limiting factor on performance.  Many a design has met market failure because the interfaces were put in the wrong place.  One way of thinking about this is in terms of "thick book" and "thin book" interfaces -- whenever possible, you want an interface in which the "instruction book" of rules and procedures for what goes on at the interface is a few lines on a page, not an encyclopedia.  To me the JLC Tryptique is a great example of this -- the QP complication in the back plate of the watch is actuated by a single pin connector, not by 5 or 6 levers or cogs -- that interface is both very clever and in the right place!

So what does this mean for watch design?  If watches are analogous to the many other systems that Clay has studied (an important disclaimer!), then we would expect that ultimate performance (akin to the "purity of purpose" comments by others) would in fact be higher with an integral design rather than a modular one, no matter how slick the interfaces are.  At the limit, you might be able to design the interfaces to be so perfect that there are no "frictional losses" in the system -- but then you've pretty much turned it into an integral system. 

If, however, ultimate performance on the original set of criteria is, to you, less important than other performance attributes such as cost, design flexibility, ease of repair, and so on, then modular designs are just fine  -- as long as the modules are "good enough" and the interfaces work!

Last thought -- much of the above might lead one to conclude (if you're still reading!) that systems inevitably migrate from being integral in their early days to being modular as they mature.  However, what happens more than you think is that a whole new wave of market expectations comes on the scene, and the existing modular designs cannot be made "good enough" to meet the need -- so there's a resulting new wave of integral designs created to get base performance up to the minimum spec. 

Clay eventually links all of this back to his theory of business/technology disruption -- but that's perhaps a discourse for another time!  Hope that this info was of some interest.

Best,

Gary G