credit default swaps and clouds
"...far more rigorously audited than very similar processes on much the same machines that only very recently rolled up millions of sub-prime mortgages into slick collateralised debt obligations?"
I'd like to point out, I read an interview of the creator of the derivative software used for credit default swaps. He TOLD the guys at the first firm to use this software they were abusing it until they fired him. He told the NEXT place he worked at that they were also misusing it, they ignored him too. He did point out if it makes you feel any better, that most of his benefits were in the form of shares that dropped to ~$0 when these companies collapsed. The companies were flat-out told that a derivative with an average of 1% defaults per year over 100 years DOESN'T MEAN 1% defaults a year, it means near-0% defaults most of the time with high 20%+ default rates every so often. Which is exactly what happened, they laughed all the way to the bank for several years of near-0 defaults then acted like it was a surprise to everyone involved when high default rates wiped them out.
Anyway... spreadsheets pulling data in from all over really are an abomination. Too easy to change the formulas and such operating on one, and the people that make them tend to not follow any "best practices".. at the very least, the "code" and data should be widely seperated so someone doesn't accidentally alter the code, but I don't see people even doing that.
"Cloud?" No solution there --
1) Moving calculations away from local systems doesn't solve anything.
2) Additional data can be gotten from the internet (not "the cloud") but this data is unstructured so it's not going to just flow into some spreasheet, database, or application the way sales figures and such would. This can lead to severe misinterpretations of data. For example, perhaps someone decides brand awareness is good, and writes an app to find out how many times their name is mentioned per day. "FropCo" was tweeted 50,000 times? Fabulous! 49,975 of those were "FropCo sucks?" Not so good, but they won't know that if automated software does all the work. As I say, online data is unstructured.