How Amazon Rigs Its Shopping Algorithm

For companies that sell on Amazon’s platform, “winning the Buy Box”—the area on the right-hand side of the page that says “Add to Cart” or “Buy Now”—is everything. While Amazon says it’s a neutral arbiter, there is ample evidence that the company has rigged its algorithm to deliver outcomes that further its own interests at the expense of sellers and consumers.

 

 

 

More than half of all internet shopping searches start on Amazon. That staggering fact has made the tech giant an all-powerful gatekeeper for every maker and retailer of consumer products looking to reach customers online. For these companies, success or failure in the pivotal online market hinges on how Amazon positions their offerings on its platform. These sellers live and die by the tech giant’s algorithms, which function as a key mechanism of Amazon’s market power.

 

The most consequential and fear-inducing of these algorithms is the one that controls which seller gets to occupy the “Buy Box” for a given product, the area on the right-hand side of the page that contains the eye-catching orange buttons that say “Add to Cart” or “Buy Now.” When customers click one of these buttons, they’re choosing the default seller, the seller Amazon awarded the Buy Box to. 

 

The Buy Box is every bit as crucial as it sounds. For any product you might shop for on Amazon – a 55-inch Samsung television, say, or the paperback edition of Brandeis on Democracy — there are often multiple sellers offering that item, including, in many cases, Amazon itself. Amazon has an algorithm that selects one of these sellers as the default seller for that product. Sellers refer to being selected as “winning the Buy Box.”

 

Customers can opt for a different seller, but they rarely do. That’s partly owing to the one-click convenience of those orange buttons, and also because Amazon doesn’t make it very obvious that there are other options. To select a different seller, a shopper has to locate the inconspicuous text that says “Other Sellers on Amazon” or “See All Buying Options,” and then pick from the list that appears. More than 80 percent of the time, shoppers go with Amazon’s chosen seller. If they’re using a mobile device, that rises to more than 90 percent.

 

In other words, for the hundreds of thousands of companies that sell on Amazon’s platform, “winning the Buy Box” is everything.

 

“Winning,” though, is a deeply problematic way to describe this process. It brings to mind a fair contest — an image of sellers competing in a free and open market with clearly established rules that apply equally to all contenders. In reality, Amazon’s third-party sellers are pitted against one another in something more akin to a Roman arena, where, like gladiators, they are subject to the arbitrary whims of a self-serving Emperor. While some sellers may win the day, ultimately only the Emperor comes out on top. 

 

Amazon insists that its Buy Box algorithm is a neutral arbiter designed to select the seller that best meets the needs of customers. Testifying before the House Judiciary Committee in July, Amazon Associate General Counsel Nate Sutton told members of Congress, “Our algorithm such as the buy box is aimed to predict what customers want to buy, and we apply the same criteria whether you’re a third-party seller or Amazon to that because we want customers to make the right purchase regardless of whether it’s a seller or Amazon.” 

 

In fact, there’s ample evidence that Amazon has rigged its Buy Box algorithm to deliver outcomes that further its own interests at the expense of both sellers and consumers.

 

At any given moment, for millions of products, Amazon can engineer one outcome or the other, depending on which best serves its interests.”

 

A Secret Recipe With a Baked In Anti-Competitive Effect

 

Only a handful of people actually know how the Buy Box algorithm works and which factors it weighs. Even within Amazon, this is a closely guarded secret. Third-party sellers have no idea how they are being scored by Amazon’s secret formula, and the formula itself can abruptly change, upending their businesses overnight.

 

This uncertainty and the anxiety it induces has spawned a cottage industry of firms that sell insights and advice to sellers. One of the biggest is Feedvisor, which, by endeavoring to reverse-engineer the algorithm, has identified more than a dozen variables that it believes are part of the recipe. These include such things as the seller’s customer feedback rating, shipping time, in-stock availability, on-time delivery rate, and sales volume. Some of these factors count more than others. The algorithm crunches the numbers continuously, like an endless auction, and gives the Buy Box to the seller who scores the highest.

 

All of this can make the Buy Box algorithm sound like an objective standard. But it’s not. One clue that it’s not is that when Amazon itself is among the sellers offering a particular product, Amazon “wins” the Buy Box with elevated frequency — even when it’s up against highly-rated sellers offering significantly lower prices. According to Feedvisor and other firms that advise sellers, the reason for this is that Amazon has given itself a perfect score on the customer experience metrics that feed the algorithm. Amazon has, in effect, made itself immune from bad customer reviews, gaining an unfair advantage over competing sellers. This becomes all the more market-distorting when you consider that third-party sellers often rack up negative reviews for things that are actually Amazon’s fault. Amazon may, for example, co-mingle a seller’s legitimate goods with a counterfeiter’s knock-offs in its warehouses, resulting in the seller receiving negative reviews from customers who receive the shoddy fakes.

 

By artificially giving itself perfect customer scores for every product it sells, Amazon can secure the Buy Box even with an inflated price. Indeed, as the only seller that has seen the algorithm, Amazon knows exactly how much it can inflate its price and still maintain the Buy Box. According to the consulting firm Buy Box Experts, Amazon “seems to possess an uncanny knowledge of the Buy Box algorithm as it will lower its price to a few dollars more than yours, recognizing the contribution of its perfect seller performance metrics, so it still maintains Buy Box share.”

 

To be clear, Amazon’s objective isn’t to sell everything itself, but to maximize its power and earnings. In some cases, that means steering customers to its own offerings and making those sales itself. In other situations, the more profitable move for Amazon is to let a third-party have the order and collect its commission, which is 15 percent for most items, along with (in many cases) substantial fees for fulfillment, advertising, and other services. Amazon’s insider knowledge of how to trigger the Buy Box is what enables it to switch back and forth. At any given moment, for millions of products, Amazon can engineer one outcome or the other, depending on which best serves its interests.

 

Photo by Scott Lewis [CC BY 2.0], via Flickr

 

Using the Buy Box to Drive Amazon’s Growth as a Shipper 

 

Still another way that Amazon games the algorithm is by giving its own warehousing and shipping service, Fulfillment By Amazon (FBA), a perfect score on the metrics related to fulfilling an order, such as the on-time delivery rate and depth of inventory. According to Feedvisor, a seller’s fulfillment method is the single most important factor weighed by the Buy Box algorithm. By granting its own shipping service perfect scores, regardless of actual performance, Amazon has created a near-imperative for sellers to use FBA. 

 

Indeed, most of the goods sold by third-party sellers on the platform are shipped by Amazon, and the overall volume of shipments handled by Amazon has surged. With sellers forming a largely captive customer base, Amazon has repeatedly raised the fees it charges for FBA services. This stream of revenue, in turn, has helped Amazon build a vast logistics network — complete with warehouses, trucks, airplanes, cargo ships, and delivery vans — at a scale that is beginning to rival that of FedEx and UPS.

 

Amazon not only shields its own delivery performance from the algorithm; it also limits the ability of customers to publicly document its shipping failures. Customers who buy from Amazon directly, or from sellers using FBA, can post comments about the product in the customer review section of the product page, but remarks they leave about shipping and delivery are often quickly deleted by Amazon.

 

Sellers used to have another way to independently track Amazon’s fulfillment performance. Because Amazon relied heavily on other shippers, such as UPS and the US Postal Service, sellers could monitor Amazon’s on-time delivery rates through the tracking numbers generated by these companies. But Amazon’s shift to vertically integrate its fulfillment operations has left sellers with a diminished ability to assess its performance, and therefore with limited ability to compare FBA to other options.

 

This lack of accountability stifles competition in fulfilment, to the benefit of Amazon.

 

According to data from market-research firm Rakuten Intelligence, the share of Amazon packages arriving late has tripled over the last two years, rising to 15 percent of orders.

 

Right now, the market provides no discipline for Amazon, because sellers have nowhere else to go.”

 

The Policy Response

 

The many businesses, large and small, that must sell on Amazon’s platform to reach the online market are not the only ones harmed by the invisible workings of the Buy Box algorithm. Shoppers are harmed too. They assume the site is serving up the best available option, when in fact, Amazon is quietly steering their choices, while steadily eroding the viability of competitors over time. 

 

To fix this, policymakers need to do three things. First, impose a non-discrimination standard on dominant platforms that requires them to apply the same rules and terms to all participants.

 

A non-discrimination standard on its own, though, is wildly insufficient to solve the problem. That’s because it’s impossible to conceive of an oversight and adjudication process that could effectively police Amazon’s platform. This is partly owing to the vast number of transactions and sellers, and the disincentives for sellers to lodge complaints. It’s also a function of Amazon’s overwhelming incentive to find ways to subtly use algorithms and data to favor its own interests.   

 

The clean way to solve this is to require that dominant shopping platforms spin off their retail divisions as separate companies. By eliminating this conflict of interest, it then becomes possible to fashion a genuine and enforceable non-discrimination policy.

 

Breaking up Amazon would also open the way for the third prong of this solution: fostering the emergence and growth of other digital shopping platforms. Right now, the market provides no discipline for Amazon, because sellers have nowhere else to go. That would change in a world in which several shopping platforms competed to attract both buyers and sellers.

 

Until then, though, the Buy Box algorithm will continue to rule the lives of sellers as a mysterious and capricious mechanism of Amazon’s power.

 

Stacy Mitchell is co-director of the Institute for Local Self-Reliance and directs its Independent Business Initiative. She’s written extensively about Amazon’s monopoly power, including the 2016 report, “Amazon’s Stranglehold.”

 

Shaoul Sussman is a Legal Fellow at the Institute for Local Self-Reliance, a graduate of Fordham Law, and author of the 2019 paper, “Prime Predator: Amazon and the Rationale of Below Average Variable Cost Pricing Strategies Among Negative-Cash Flow Firms.”

 

The ProMarket blog is dedicated to discussing how competition tends to be subverted by special interests. The posts represent the opinions of their writers, not necessarily those of the University of Chicago, the Booth School of Business, or its faculty. For more information, please visit ProMarket Blog Policy.