Over at the Seattle PI blogfarm, a fascinating post purports to be the inside guff on why so many XBox 360s are keeling over with the Red Ring Of Death -- the condition where the hardware refuses to wake up, flashing its circular LEDs in the eponymous signal for "I'm dead. Ship my corpse to Redmond".
If you couldn't care less if every gaming console in the world spontaneously mutated to a toad and crawled away to eat ants, stick with me here. The blog post, while being unsourced, has every sign of being correct -- I've seen exactly what's being described happen before.
It seems that nearly one in three XBox 360s is dying long before its time, and none will make old bones, due to a whole host of hardware design mistakes. There's overheating, which gradually loosens solder connections. There's no margin in the timing of some components - if one or the other is a little slower or faster than average, then there's no slack in the rest of the circuit to compensate. There's not enough testing. There are too many mistakes in production. Key components are not being screened prior to assembly. The whole thing was rushed to market for competitive reasons.
All of the above - and I mean all - happened to Sinclair Research twenty five years ago. The ZX Spectrum and the Sinclair QL between them exhibited every one of the above faults, and while I never knew the failure rate of the QL (like it sold enough to have one), the incredibly successful ZX Spectrum had a DOA (dead on arrival - the punter unpacks it and it's frobbed already) rate of 25 percent plus.
In Sinclair's case, this was because the company did not understand three important processes - designing for manufacturing, manufacturing, and management. And that boils down to one important process: management.
The engineers came up with smart circuits which worked in the lab. They then turned them into lists of parts, circuit boards, etc, for third party manufacturers - who then sourced the parts. Which, when put together, didn't work properly - because the specifications were wrong, either by design (but it had worked in the lab, due to this or that happenstance) or because the bulk supplies were different in key respects to the actual parts needed.
This sort of thing happens all the time. You solve it by keeping a very close eye on production stats, having a very close relationship (warning; may involve shouting a lot) with your suppliers, and making sure that everyone involved talks to everyone else. That doesn't just happen - you need rock-solid contracts, you need home phone numbers, you need to tell your design engineers (who can be rather odd beasts) that it's their job to talk to sales engineers (who are even odder) in the component companies. Left to their own devices, they'd throw dung at each other: management's job is to stop that, keep them talking and watch from the sidelines to make sure that the right feedback is ending up in the right ears.
None of that happened - well, nowhere near enough - as far as I could tell at Sinclair Research, and the result was hundreds of thousands of disappointed customers and a dead company.
But then, Sinclair didn't have an enormous cash machine thundering away in the corner bankrolling inept, inefficient and incompetent management. Microsoft does, which means everyone gets to pass the buck and pretend a few billion more here and there doesn't matter.
One could plausibly look at other aspects of the company and see more evidence of the same...