[makomk]

The trouble is that, particularly with physical products, being overfunded often means the project does need more money up-front than originally estimated. For example, they may need to change their production process to one that has higher upfrunt costs but allows for higher volume.

[jdechko]

Yes. But with economies of scale each individual unit should cost less than the one before it. That's why you have minimum runs for production. Any less and the manufacturer doesn't make enough money to be worthwhile. Fixed cost vs unit cost, etc.

With things like aircraft, the first one costs billions because of the R&D having gone into it. But by the time you roll out the hundredth F-22, you've learned a lot about the process and refined the techniques. Same thing with processors.

One of the biggest downsides of kickstarter projects is that the creators have some engineering/ design skills but seem to have practically no business skills. (I'm generalizing here for the sake of argument).

[shabble]

> But with economies of scale each individual unit should cost less than the one before it.

Not really. At pretty much every step of the mass-production pyramid (higher volume, lower per-unit cost), there are significant capital costs involved. The tooling for a injection mold can be in the $10-100k+. That means the first unit in subsequent process is going to be huge, because you've got nothing to amortise it over yet. Even if the process change is free, there's almost certainly going to be some NRE work updating your designs, optimising DfM, etc.

There's an example mentioned in the excellent 'The Factory Floor'[1] where going from single-colour injection-molding plus hand-painting (with 50% defect rate) to a 'dual shot' injection-mold (with 0% defect rate) ended up costing $0.94/unit (vs $0.70 for the former; 35c each, and throwing away half)

Likewise, for small runs handing it off to Shapeways might well be the most cost-effective, but after a certain size, it would be madness. Firstcut/Protomold[2] look like they're stepping up to be the medium-scale option, which is nice.

[1] http://www.bunniestudios.com/blog/?page_id=2850

[2] http://www.wired.com/design/2013/01/protomold/