: New: Book Report: Super Crunchers

It's a book about working with Big Data. Considering some of the projects I've worked on, you think I'd be pretty excited. But my experience made me kind of picky about the details. At first, this book over-sells its premise: We'd make better decisions if we got out of the way and let computers do everything for us. But some self-declared "experts" don't want to give up control because then they wouldn't be special anymore. If you dig in another 100 pages, he puts in the relevant qualifications on that; so you'd better hope that readers were still paying attention that far in.

Humans aren't great at thinking about risk, about statistics. If a doctor notices that your lungs are going whooka-whooka, that doctor probably knows some of the diseases that have that symptom, might know which of those is most likely... but probably hasn't really internalized the odds of these, might not do a good job of making decisions that depend on those odds. At first, the book harps on this, pointing out that doctors may now be obsolete. Except that they're not, of course. It turns out that distinguishing between lungs that go whooka-whooka and lungs that go wheeesh ain't so easy. And if something new comes along, good luck figuring out how to deal with it without doctors, just going along based on old statistics.

Apparently, the book's author got sold a spiel from some web analytics company called Offermatica. Web analytics are a great way to make decisions... except that Offermatica doesn't seem to have had a very compelling analytics product. At least, lots of experts at web companies tried to talk the relevant purachasers out of using Offermatica. According to the book, that's because these are self-declared "experts" who clutch their decision-making power. Except that... they're not. He points out usability experts with special scorn. Usability experts I've worked with have been pretty darned glad to get analytics data and to use it to drive their design decisions. They have objected to sloppy analytics. (This is probably a good time to mention that I work for a company that has a web analytics product. I don't speak for my employers. Really, I don't speak for anyone but me. Though Offermatica appears to have disappeared in the time since this book came along, so maybe I shouldn't worry that you think I'm just trying to steer you away from my employer's competition. Anyhow.)

So... yes, trust numbers. If you have facts, please use facts. If all you have is opinions, then favor the opinions of folks who've studied their subjects.

Tags: book choice

blog comments powered by Disqus