e.g. When I edited a photo-book, sure, the formats and printing were negotiated in advance, but truthfully if you really needed to change it, you could, quite close up to the moment of final production.

So out of 2 months work, things are malleable right up to 5 days from the end. It sucks for everyone who has to adjust to the adjustments, but it just takes time and effort.

Mechanical things aren't always so flexible. Machining large important parts is 10 step process, spread over 2 months. If you change anything at the 1 month point, you're going to have to throw away a large chunk of aluminium and start again. You might even have to rearrange your whole machine setup as well. And everyone else whose parts fit on your part might have to do the same.

This is not to say it cannot work at all, but that (contra the wikinomics and other boosters) a lot of these models are a lot less impressive once you get out of producing software, media or other "knowledge products" which are mutable all the time.

Of course, if you're engaged in bespoke/small scale industrial mechanical production, things can be more flexible, but that model is energy intensive and costly.

As a related point, I recall a discussion from a random economist about how Chrysler could reinvent itself as the "Apple of car companies." I had to point out that while it is theoretically possible a lot of the conditions present in the computer hardware market (massive standardised production chain, open standards for a large number of components) just aren't present in the car market. And they aren't present in part because bending metal is slower and more expensive and less flexible than stamping out silicon.

Point being the whole thing is a spectrum and the transaction costs in production of physical goods just don't fall like the ones in software do.

Ag/Food processing is likewise all about creating very standard raw materials, then combining them in very simple ways, industrially speaking.