I really enjoyed Ian Ayres' new book, Super Crunchers: Why Thinking-by-Numbers Is the New Way to Be Smart. The book is easy to read, authentic, and thought-provoking. My only complaint is the index, which isn't comprehensive enough for such a wide-ranging book.
Ayres' topic is the growing importance of data mining and randomized testing.
On the data mining side, Ayres trots out the familiar cast of characters: CapOne and predictive modeling, NetFlix and their recommendation engine, WalMart and the pop-tarts story. He goes further, though, discussing Zillow, Inrix, Farecast, and eHarmony.
Ayres writes with calm authority -- he's an economist and law prof at Yale, rather than just a popular journalist -- covers a great deal of ground in relatively few pages.
My copy of the book is heavily marked as I found myself underlining new ideas and connections. Just a few:
- IRS. Instead of thinking as the IRS as solely a tax service, they could become an information provider, the "Information & Revenue Service". Is your small business underspending or overspending on advertising? For those who wanted to buy comparative data, the IRS could offer its opinion, benchmarking your numbers against other comparable firms, and correlating the results with growth and profitability. p34.
- 100,000 lives. Don Berwick rolls out six simple ideas from evidence-based medicine (wash hands before placing central line, elevate head of ventilator patients, etc) in 2004, gets more than half of all US hospitals to participate, and saves an estimate 122,342 lives in 18 months. p82.
- Epagogix. Dick Copaken uses regressions to help Hollywood predict (and then improve) box office revenue. p144. Malcolm Gladwell recently wrote about Epagogix in the New Yorker.
- Overconfidence. Humans regularly underestimate the uncertainties in our predictions, leading to many situations where impartial algorithms outperform experts. p112.
- Virginia's SVPA & the RRASOR. In 2003, Virginia's sexual predator act became (according to John Monahan), "the first law to specify in black letter the use of a named actuarial prediciton instrument and an exact cut-off score on that instrument." p120.
- Larry Summers. Former Harvard president Larry Summers didn't suggest women had lower average intelligence than men, he suggested that the standard deviation of intelligence was higher for men than women.
This book captures the 2007 zeitgeist, weaving together many interesting people and ideas. There are even several memes relevant online advertising.
For example, Ayres used an Adwords copy test to pick his title -- "Super Crunchers" beat "The End of Intuition" (p56).
And his notion that "experts" pulling the levers often generate sub-optimal results versus giving solid algorithms more leeway reminds me of George's post on approaches to PPC management (Navigating vs. Steering: How The Client - Agency Relationship Affects Profits.)