Amazon and Algorithmic Pricing

Recently an interesting incidence happened in E-Commerce giant Amazon.

As reported by Michael Eisen, 17 copies of the biology book “The Making of a Fly” by Peter Lawrence published in 1992 (not printed anymore) was for sale on Amazon. 15 second hand copies started from $35.54, whereas 2 new copies were sold starting from exactly 1730045.91$.

More interestingly, the two sellers of the new copies have both profiles with high customer satisfaction feedback and high transaction volumes and they both ask for more than million $s . This abnormal situation lead to a much more interesting discovery. Eisen followed the behavior of the prices, both rising steadily and discovered a very simple relationship between the two sellers prices.

Profnath, strategically was positioning itself “slightly” below the market price to be price competitive and draw more customers to its e-store, so it was setting its price exactly 0.9983 times the lowest market price. Bordeebook, on the other hand, was relying on its high sales volume and excellent customer feedback to increase its profit margin, thus setting the price to exactly 1.270589 times the lowest price. As there are no other new copies of the book, these sellers run into an infinite loop, where they rise the prices to the roof (million $ level). Furthermore, this kind of responsive behavior raises the concerns, that Bordeebook maybe does not even posses the actual copy of the book and acts as a reseller.

Today, while firms essentially consider the production and market data for pricing, it is not possible to overlook the competitors data and strategy. Especially, e-commerce enables possible ways of automation and leads to new, emerging situations. The main reason for these kind situations is that it is impossible to manually maintain such a diverse inventory of enormous volume. Therefore, as in the case of Amazon doing business with countless suppliers, for the sustainability of the system automation is inevitable. Thus, interesting price tags with .01$ precision are mostly outputs of smartly prepared scripts. In this regards, process control is obviously necessary, as I doubt someone would be interested in a regular book sold for million $s.

– this post has been published in dijitolog