December 13, 2016 – The text of this article is below, or click here. We will also host an Amazon conference call 12/14 at 11am ET. To RSVP, please email editorial@thecapitolforum.com

Amazon Antitrust Update

President-Elect Donald Trump has criticized Amazon for its dominance and has sparred with CEO Jeff Bezos.  One way that a Trump Administration could enforce the antitrust laws against Amazon is to sue Amazon for using its platform and algorithms in a biased manner that prioritizes Amazon’s own products and services over those of merchants that are dependent on Amazon’s platform and with whom Amazon competes.  It only takes one enforcement action covering a narrow instance of Amazon employing such bias to create precedent that could put its business model -- leveraging its platform to vertically integrate into a wide range of industries -- at risk.  This is the EU’s strategy against Google in Europe – going after narrowly defined breaches to create precedent that will cause a domino effect with wide-ranging implications.  Precedent against Google can be used against Amazon, and vice versa, as the tactics they employ are the same.

Amazon has vertically integrated into selling its own items and competing with retailers that depend on its platform. Recently, Amazon has made a push into the fashion clothing vertical, and has developed its own house brands.  In favoring its own private label clothing over other brands in prime promotional real estate, Amazon appears to utilize its platform in a non-neutral way that creates antitrust risk due to its dominance.  Such preference for the company’s own products is analogous to the Google Shopping case in Europe as well as platform behaviors that saw Google face scrutiny from the FTC in 2013.  As Amazon continues to gain commanding market share as the primary destination for online retail and to vertically integrate into fashion and other industries, any prioritization of its own products increases risk of antitrust enforcement. 

Our Findings

Through an analysis of hundreds of Amazon and non-Amazon products sold on the platform, we found the company prioritizes its own clothing brands on the promotional carousel labeled “Customers Who Bought This Item Also Bought.”  Amazon appears to relatively restrict competitors’ access to this prominent placement on product pages for its private label, directing consumers toward its own products rather than competing products.  On pages for products that are not Amazon brands, however, Amazon opens the space up much more for competing products.

Through our testing, we found that more than 71 percent of the first 20 carousel results on Amazon private label pages were products from an Amazon clothing company.  Most of the time, those results were from the same brand we were currently viewing, but a few times there were placements for other Amazon labels. Compare that to 14.9 percent of carousel results belonging to the same brand on product pages for non-Amazon marketplace brands. Further, some of the Amazon products we observed had very few results in their carousels, which was rare for non-Amazon brands.  One effect is more empty spaces in the Amazon carousels rather than potentially filling those spots with other brands. A practical analogy could be that Amazon treats the presentation of its own line of products like a brand-exclusive retail store while other lines of products are housed in a more competitive department store-type setting.

We reached out to Amazon for comment and explanation, but the company did not respond to our questions.  We think it is highly unlikely that the carousel results accurately and neutrally reflect customer shopping behavior, but without access to Amazon’s algorithm or a response from Amazon we cannot be sure.  Nonetheless, it would be difficult to envision a neutral algorithm that heavily prioritizes Amazon items or cuts off after less than 20 items for Amazon’s product set and not for other non-Amazon products on the platform.

Amazon’s Market Power In Online Retail

Amazon’s Stranglehold: How the Company’s Tightening Grip is Stifling Competition, Eroding Jobs, and Threatening Communities, a report published last month by the Institute for Self Reliance, estimates that “Amazon is now capturing nearly $1 in every $2 that Americans spend online.”  Additionally, it “sells more books, toys, and by next year, apparel and consumer electronics than any retailer online or off,” says the Institute.

The press release accompanying the report continues: “[Amazon] increasingly controls the infrastructure that rival companies depend on to reach the market, making Amazon a novel and particularly potent threat to competition. By corralling a growing share of online shopping traffic — 55% of shoppers now start their searches directly on Amazon — Amazon has left competing retailers and manufacturers with little choice but to become third-party sellers on its platform. This gives Amazon the power to dictate the terms by which its competitors and suppliers operate, and to levy a kind of tax on their sales.”  Amazon “provides essential market access for competing retailers and then uses the data it gleans from them to expand its own retail sales,” and additionally “has begun to exploit the data it’s gathered from manufacturers to compete directly with them by producing the same products,” says the report. “Amazon’s power to manipulate what products we encounter is especially concerning in the book industry,” adds the Institute. 

We have previously written about Amazon’s conduct in books as a potential Sherman Act Section 2 violation. We wrote about how Amazon uses its retail market power to vertically integrate into and dominate self-publishing. Journalist and author Frank Foer noted in a New America event on the topic that Amazon can “crush” publishing by “deny[ing] placement in emails” and “shaft[ing] them in their algorithms if they don’t agree to Amazon’s terms.” Amazon, he said, disfavors the authors who do not agree to its terms and puts their books at the bottom of search results. Brad Stone makes similar points in his book, “The Everything Store: Jeff Bezos and the Age of Amazon”, detailing how executives play with algorithms to gain leverage over publishers and to underscore Amazon’s power during negotiations. Below we discuss in detail one way that Amazon appears to be manipulating its algorithm for featured products to favor its own private label fashion products. 

Amazon’s Vertical Integration into Private Label Clothing

Amazon launched earlier this year seven lines of private label clothing, including: Franklin & Freeman - Men's shoes; Franklin Tailored - Men's suits, jackets, pants, and accessories; James & Erin - Women's clothing; Lark & Ro - Women's clothing; North Eleven - Women's cold weather accessories; Scout + Ro - Children's clothing; Society New York - Women's clothing, purses and accessories. Amazon apparently released the brands without much notice to the greater fashion world. Industry trade publication WWD wrote in February that the “threat” from Amazon’s in-house labels “quietly arrived.”

The company has about 2,000 stock keeping units across its seven brands. At the same time, Amazon is working to gain cachet as a high-end clothing retailer. The company started sponsoring the Tokyo Fashion Show this summer with a multiyear contract. And an international ad featuring models delivering packages with the tagline “Now Delivering Fashion” ran thousands of times this fall on national television and certainly many more times for online videos across platforms. The message seems to be “you already go to Amazon for your miscellaneous merchandise, and we’re also a high fashion destination.”

In-depth Look at How Amazon Prioritizes its Private Label Clothing

To better understand Amazon’s behavior in prioritizing some of its own products, one needs to understand how the company displays merchandise across its platform, including the opportunities it offers manufacturers and resellers to advertise or market goods. When first loaded, item pages on Amazon generally follow a similar format. A large picture of the selected item with smaller thumbnail images appears in the center left of the page with item name, price, general review information, specifications, add-on products like warranties and purchasing options to the right. Most of the language and options in these first sections appears to be informational. It does not appear, for example, that on this page for a TV, Samsung could pay Amazon to include marketing language in the size category such as “Huge 65-inch TV” or something to that effect.

Directly below the informational language, Amazon does offer manufacturers the opportunity to provide more colorful descriptions. In our TV example, Samsung includes lines like “[Dimming feature] uses new technology to create a more vibrant viewing experience than you’ve ever seen.” Next there is an ad section titled “Frequently Bought Together” that appears often with electronic items. It appears that different sellers on Amazon can buy that space to bundle items with the primary product, such as HDMI cables with the Samsung example. This section does not appear on all Amazon pages and was included on none of the clothing pages that we observed.

The next section and the focus of this article is a carousel of products related to the primary product and apparently based on the behaviors of other users who visited the primary product’s page. This section is most often called “Customers Who Bought This Item Also Bought.” We have also observed instances where the section is called “Customers Who Bought This Brand Also Shopped For” or “Customers Who Viewed This Item Also Viewed.” There is no message for the user that indicates what he’s viewing is an advertisement or that sellers can pay for a placement in the carousel, which is ascribed to a similar “Sponsored Products Related To This Item” carousel or the “Frequently Bought Together” section described above. Without such a disclosure, users are likely to assume that the carousel is neutral and treats all products the same.

Through an analysis of hundreds of Amazon and non-Amazon products sold on the platform, we found that the company prioritizes its own clothing brands through the carousel. Amazon appears to relatively restrict competitors from this advertising space on its private label pages, potentially keeping consumers from viewing alternative products. On competitors’ pages, however, Amazon opens the space up much more for alternative products. Further, it appears partners in Amazon’s “Top Brands” program are afforded more representation in the carousel than the average marketplace seller.

Tests of Amazon’s Carousels

In order to test the neutrality of Amazon’s carousels, we identified 10 of the top results when searching for each of Amazon’s private labels. We recorded whether the first 20 results in the carousel were from one of Amazon’s private labels or a different brand. (Our monitor and browser sizes allowed us to see 10 products per carousel cycle. It appears results are uniform regardless of how many products appear on a page, and the website will adjust based on the size of the browser window.). To compare to those results, we searched Amazon for products in a category similar to each of the private label brands. Some Amazon partners receive a top banner ad reading “Shop Our Top Brands” that appears at the top of search results. Clicking that link presents a separate list of results, the top 10 of which we used to collect data on Amazon partners. For data on the average marketplace seller, we used the top 10 results from a regular search.

The search terms we used for each of the corresponding private labels include: Franklin & Freeman – “Men’s dress boot”; Franklin Tailored – “Men's suits”; James & Erin – “Women's tops and blouses”; Lark & Ro – “Women's dresses”; North Eleven – “Women’s winter”; Scout + Ro – “Children's clothes” and “Boys activewear” and “Girls dresses”; Society New York – “Women's skirts”. Considering the overlap in product set for James & Erin, Lark & Ro, and Society New York, we used specific types of women’s clothing in our searches. Further, we needed to use more specific search terms for comparison products in children’s clothing because no “Shop Our Top Brands” banner appeared for a generic “Children’s clothes” search. We used the top 5 results of each from “Boys activewear” and “Girls Dresses.” We also used incognito mode in our Chrome browser to prevent Amazon from accessing our cookie store and previous browsing history in case that had an effect on results.

Results

Through our testing, we found that more than 71 percent of the first 20 carousel results on Amazon private label pages were products from an Amazon clothing company. Most of the time, those results were from the same brand we were currently viewing, though a few times there were cross-brand placements. Compare that to 47.4 percent of the results in “Top Brands” carousels belonging to the same brands and 14.9 percent of carousel results belonging to the same brand for the brands found through general search results. There were also 17 instances when all 20 carousel items on an Amazon page were Amazon products, which occurred two times for “Top Brands” and zero times for marketplace search items.

The first set of items in the carousel have prime placement to attract the user. It also appears that additional carousels can appear on product pages further down the page, though for our survey we only gathered results from the first carousel visible on each page. The first carousels are directly below the primary item on the page and one of the first alternative, add-on, or supplemental products that the user will see when scrolling. We found that Amazon’s priority placement on its own private label clothing was relatively higher for the first set of 10 items than the first set of 20, with 80 percent of products shown in that first set coming from an Amazon private label.

Of the 70 Amazon clothing product pages we observed, 33 of them displayed a first set of carousel items made entirely of Amazon private label clothing. Compare that to 7 instances among the 70 “Top Brands” pages we observed and once among the marketplace results. Below are a few samples of carousels from Amazon private labels compared to their non-Amazon alternatives, in order of an Amazon private label, a “Top Brand,” and a marketplace result. Checkmarks indicate the carousel item is from the same brand or brand owner as the primary item on the page:

Additional Observations

We also identified some characteristics that only applied to Amazon’s private label brands. It is unclear why only the carousels for Amazon brands behaved in these ways. For example, some of the Amazon products we observed had very few results in its carousels, a few times less than the 20 needed for a full survey. That was rare for non-Amazon brands.

One effect, whether purposeful or not, seems to allow for more empty spaces in the Amazon carousels rather than filling those spots with potentially other brands. Further, we previously mentioned that the primary name for the carousel was “Customers Who Bought This Item Also Bought” and sometimes “Customers Who Viewed This Item Also Viewed.” However, only on some Amazon brand product pages were there carousels called “Customers Who Bought This Brand Also Shopped For.”