Questioning Amazon’s algorithms. Amazon famously puts customers first. But a new report by ProPublica suggests that the company’s code pushes its own merchandise, even when other listings on the site offer better deals. The algorithm that chooses which seller to feature in the rectangular orange “buy box” seems to favor Amazon itself, or the “Fulfilled by Amazon” partners who pay the company to handle inventory and shipping. Read closely, though, and ProPublica’s argument turns out to be almost entirely about shipping costs. For Amazon’s favored Prime customers (who pay $100 a year for free shipping and other perks), and for anyone whose order tops $50, the shipping costs nothing, and the deal Amazon highlights really is the cheapest one. That’s still arguably a problem, but hardly the capital offense the story implies. Watchdogging algorithms is the investigative journalism of the future, and ProPublica does great work. But in its eagerness to tar Amazon it has obscured the real lessons here: Platform owners are always going to give themselves an edge. Amazon deserves credit for running an open platform that lets alert consumers find good deals and gives outside merchants access to its vast market. It has also earned a rap on the knuckles for tilting its listings to goose Prime signups. Instructive, for sure. Scandalous? Probably not.
Nuggets of gold in piles of user comments. Of course you care about user feedback and customer reviews! But who has time to read them all? Now there’s a machine-learning-style data analytics tool that will read them for you and tell you what to do (Buzzfeed). This service is called Metis, and for the moment its eyes are trained on the products of luxury merchants. For instance, it told a high-end hotelier that its guests really, really cared about customizing their breakfasts. We can assume this sort of analysis will quickly move down market as well, where the volume of feedback is even more overwhelming, and the potential payoff for small incremental improvements is that much higher. Anything that helps businesses listen better is valuable — as long as the process still allows individual human voices to make themselves heard.
Read More