The Apotheosis of Capitalism: Amazon’s Billion Dollar Buy Box and the Algorithms Behind It

When in the course of buying books made from tumblr blogs, or DVD box sets, or 12 packs of Pop Tarts on Amazon, while marveling at their ability to provide two day free shipping (how do they do it?), you probably don’t pay much attention those other prices listed below the main one, the “More Buying Choices.” After all, who would pay an extra $8.28 just to order the second season of Breaking Bad from a place called Brandon’s House of Fun? However, a whole cottage industry has sprung up around getting third-party sellers into the “More Buying Choices” box, or “Buy Box” as Amazon calls it. Increasingly, the answer seems to be algorithms.

Although anecdotally the lowest prices always seem to come from Amazon directly, third-party sellers now make up over one third, or around $6 billion, of all Amazon sales. Their strategies for competing, absent hiring a 16-year old intern to refresh the “Used & New” sales page every five minutes for all of their products, seem to center on algorithms, or robo-pricing. Most of the articles about robo-pricing imply that these algorithms come from Wall Street, but there’s actually an entire body of academic work focused specifically on pricing algorithms for e-commerce. Some of them are on that Bayesian tip, relying on models of probability. Others prefer game theory or machine learning. In all of them, the goal is to form a competitive price while also maximizing profit through rules based on transaction costs, production costs, and consumer preferences. Here, they differ from Wall Street algorithms, which are mainly about price alone.

They also differ from Wall Street algorithms in how they go awry. Much was made of the 2010 Flash Crash when a stock market heavily dependent on algorithmic trading dropped by 1000 points on May 6 before rebounding within twenty minutes. On Amazon though, despite what one might infer from the nature of competitive markets, the opposite happens – witness: the $23 million biology textbook. In Amazon’s case, the profit maximization elements take over. Now, the price has returned to sanity, but the same company that was selling the book for $23 million, bordeebooks, still has a listing for the book in “Used – Good” condition for $900. Indeed, these price absurdities are not lost on the authors, many of whom complain to the third-party sellers, only to find the responsibility constantly shifting from the seller to Amazon and back.

The robo-pricing of Amazon, though, is just a symptom of a rapid trend over the past ten years of turning more and more of our decisions over to algorithms. This TED talk is good at explaining the implications of this surrender in broad terms, but specific implications can be found in the algorithms themselves. Take, for example, the Markov decision-making process, on which this paper on pricing algorithms relies heavily. The key feature of the Markov process is that it is memory-less; it pays no attention to the situation that preceded it nor does it look to the future. It only focuses on the present to make chains of decisions. This myopia toward context is clearly on display in the case of the $23 million textbook, and and the ramifications loom larger as Amazon expands in its quest to be the conduit for all Internet commerce. What does myopic pricing mean for inflation? Additionally, these Markovian games largely just depend on the price of the competitor, but as these vendors are locked in feedback loops, how are they supposed to respond to consumer demand?

Ever since the demise of the agora, examples of a perfectly competitive marketplace have proved elusive for the economist-minded among us. Amazon represents just such a place, albeit one where Amazon gets to write the rules to its own advantage. Here, in the battle for the buy box, we get to see the principles of the free market at work, and where do they lead us in the 21^st century? To an automated system. Robo-pricing may be able to bring lower prices for some products, but for others, instead of competing to drive the price down, the price goes all over the place until humans intervene. Let’s just hope the market doesn’t become so big that these decisions come down to the microsecond. One can only imagine the horrors of a country where the bulk price of Pop-Tarts starts to destabilize.