Dispatches from the front line of the internet

RSS

Sellers on Amazon use automatic tools which pushed a book on flies up to $23,698,655.93

Robo-pricing gone wrong

New Statesman
A really expensive book? Or a really well-designed fly?

Barney Jopson for the Financial Times:

Amazon sellers – using third-party software – can set rules to ensure their prices are always, for example, $1 lower than their rivals’. More complex algorithms can analyse data to set prices most likely to secure a prominent position on the site.

Last year, out-of-control algorithms inflated the price of The Making of a Fly, a genetics book, to more than $23m, according to Michael Eisen, a biologist who blogged about it.

Eisen's blog is a great case study. My favourite excerpt:

The behavior of [one of the sellers,] profnath is easy to deconstruct. They presumably have a new copy of the book, and want to make sure theirs is the lowest priced – but only by a tiny bit ($9.98 compared to $10.00). Why though would  [the second seller] bordeebook want to make sure theirs is always more expensive? Since the prices of all the sellers are posted, this would seem to guarantee they would get no sales. But maybe this isn’t right – they have a huge volume of positive feedback – far more than most others. And some buyers might choose to pay a few extra dollars for the level of confidence in the transaction this might impart. Nonetheless this seems like a fairly risky thing to rely on – most people probably don’t behave that way – and meanwhile you’ve got a book sitting on the shelf collecting dust. Unless, of course, you don’t actually have the book….

My preferred explanation for bordeebook’s pricing is that they do not actually possess the book. Rather, they noticed that someone else listed a copy for sale, and so they put it up as well – relying on their better feedback record to attract buyers. But, of course, if someone actually orders the book, they have to get it – so they have to set their price significantly higher – say 1.27059 times higher – than the price they’d have to pay to get the book elsewhere.

Of course, realistically this sort of thing isn't the natural result of computers entering the fray; if the two booksellers had used the tool at all well, they would presumeably have entered some sort of upper bound, where an actual human person would be notified.

Really, this is just another example of how the internet makes economics textbooks come to life. This is exactly the sort of situation which sounds like it might happen in theory, but never could in practice, and then, all of a sudden, it does. Amazon has already been known to practice dynamic pricing, charging customers the most it thinks they will pay. What next?