Wait, How Much?
I just came across a fascinating blog post by Dr. Michael Eisen, a biologist at UC Berkely, about a copy of a book on flies that was offered for sale at Amazon.com for $23,698,655.93 (plus $3.99 shipping). A quick look at Wikipedia’s list of most expensive books shows that this comes somewhere in between Leonardo Da Vinci’s Codex Leicester (bought by Bill Gates for $30.8m) and the St Cuthbert Gospel (bought by the British Museum for $14.3m).
With all due respect to Professor Peter Lawrence, the author of The Making of a Fly, it’s hard to believe that the book is worth more than the combined sum of money paid for Chaucer’s Canterbury Tales ($7m), Shakespeare’s First Folio ($5.6m), the Gutenberg Bible (4.9m) and Ptolemy’s Geographia ($3.5m). You could buy a Sebastian Errazuriz bookcase ($75K) for each of them, and still have plenty of change left over.
In a nutshell, what happened was this: Amazon listed two copies of the book for sale, one offered by bordeebook and the other by profnath. It seems that they were both using automatic algorithms to determine the optimum price by comparing other vendors’ prices and modifying their own price every day by a chosen factor. Profnath was multiplying the previous bordeebook price by a factor of 0.99830, while bordeebook was multiplying the resulting profnath price by 1.27059. Since the product of the factors was greater than one, the price exploded way beyond anything that might be called reasonable.
And it seems that this went on for a while. I ran a quick simulation in Excel, starting from a (reasonable) price of $101.15, and repeatedly multiplying by Dr Eisen’s factors of 0.9983 and 1.27059. After around 11 iterations – or 11 days, since the prices were updated each day – both sellers’ prices reached $1000. 40 days in, they were at 1 million dollars. After 53 days, they hit $23m, when someone finally noticed what was happening. That’s almost two months. Out of curiosity, had it not been caught then, the price would have reached $1 billion only two weeks later.
This is my Excel simulation. The key lines are highlighted in yellow:
Dr Eisen’s story is a nice illustration of why it is important to think about what you’re testing, and to ensure that the behavior of the system is not just correct, but also reasonable. A business process might produce results that are technically accurate, but you still need to think about whether these results make sense in the context of that business process.
After all, you don’t want to wake up one day to find that no one wants to buy your books…
