Interplay between Amazon store and logistics

ABSTRACT

This paper examines the antitrust investigations by the European Commission (AT.40703 – Amazon Buy Box) and Italian Competition Authority (A528-FBA AMAZON), and the regulatory proceeding by the Italian Authority for Communications and Postal Services (AGCom’s resolution 94/22/CONS), surrounding Amazon's Store, and the interplay with its logistics and delivery services. While these authorities have raised concerns about Amazon's alleged “self-preferencing”, rapid growth, and its alleged ability to leverage its (upstream) e-commerce market power to strengthen its position in (downstream) logistics services, this paper highlights alternative perspectives that are often overlooked or neglected. To achieve a comprehensive understanding of Amazon's conduct related to the above-mentioned investigations, we consider its business model, monetisation strategy, and underlying incentives. By delving deeper into the Amazon Store and its connection to logistics services, we argue that the allegations do not fully capture the evidence surrounding Amazon's practices.

KEYWORDS:

• Vertical integration
• open systems
• online retailing
• logistics
• competition
• antitrust
• abuse of dominance

1. Introduction

Amazon has experienced remarkable growth since it was founded by Jeff Bezos in July 1994 as an online bookstore. Today it is a successful retailer and has launched several innovative products and services (such as Kindle, Prime, and Amazon Web Services) enhancing competition in most sectors it entered¹. Amazon grew from an online bookstore to a popular retailer active worldwide with a broad product selection, a logistics network, a Hollywood studio, a robotics manufacturer, a provider of web and data services, a retail grocery store, and much more. Its success as a retailer can be attributed to its ability to offer customers low prices, wide selection, and convenience in terms of product availability, and fast and reliable deliveries. As the Amazon Store (“Store”) grew, Amazon opened it up to third-party sellers (“3P Sellers”), which, as of 2022, accounted for almost 60% of sales in the Store.²

Many products in the Store are available from more than one seller: for a given product there may be a combination of Amazon first-party retail offers (“1P Retail”, goods purchased by Amazon directly from suppliers and sold in the Store) and offers by 3P Sellers who use the marketplace services offered by Amazon. 3P Sellers have the option to use either Amazon’s logistics services³ (Fulfilled by Amazon, “FBA”) or another delivery provider (carriers who participate in the Merchant Fulfilment Network, “MFN”, or “seller fulfilled”).⁴ This is at the core of most of the concerns against Amazon: its “dual role” both as a 1P Retailer and provider of logistics services, and as the owner of the Store.⁵ Amazon competes with 3P Sellers and other carriers in the Store, while at the same time it operates the Store.

Offers relating to the same product might be differentiated (and significantly so) in multiple dimensions including price, reliability of the seller, and speed of delivery. Because of this heterogeneity, Amazon must set the design attributes in the Store for itself as well as for both 3P Sellers and MFN. As the operator of the Store, Amazon will generally internalize negative externalities, generating significant efficiencies.⁶ As we will discuss in detail, given that 3P Sellers do not operate the Store as a whole but are instead focused on their own sales, their interests are not necessarily aligned with those of Amazon. For example, their commercial decisions will not fully take into account that a bad customer experience, such as a defective product or a missed/late delivery, will have a detrimental impact on the Store as customers will make their future purchases at one or more of the many other retailers.

Some critics argue, however, that potential misalignment of interests can also go in the other direction. They argue that Amazon’s “dual role” creates an intrinsic and unavoidable conflict of interests in which Amazon has an incentive to “self-preference” 1P Retail products over 3P Sellers, and 3P Sellers using FBA compared to those using alternative fulfilment and delivery options.

This paper addresses the main allegations made in three investigations: European Commission (AT.40703 – Amazon Buy Box),⁷ Italian Competition Authority (A528-FBA AMAZON)⁸ and the Italian Authority for Communications and Postal Services (AGCom’s resolution 94/22/CONS).⁹ In these investigations, the allegations against Amazon can be summarized as follows:

1. Amazon is “self-preferencing” its 1P Retail sales over sales by 3P Sellers; and 3P Sellers using its own logistics service (FBA) over that of rival logistics operators to the detriment of 3P carriers, 3P Sellers and competing platforms. The claims were based on (i) a preferential treatment of Amazon's 1P Retail business over 3P Sellers (e.g. 1P Retail being the Featured Offer¹⁰ more often than 3P Sellers due to the criteria selected by Amazon), and (ii) a preferential treatment (e.g. higher likelihood of becoming the Featured Offer) to 3P Sellers using Amazon’s logistics services, FBA. As a consequence, 3P Sellers would be highly incentivised to use FBA as a result of these alleged advantages;

2. Amazon has grown too much, too fast, and is “too big” – expanding significantly in e-commerce logistics and B2C parcel delivery. While this is a non-technical concern (antitrust laws address anticompetitive conduct, not size) repeatedly expressed by policymakers and the general public,¹¹ the Italian Authority for Communications and Postal Services (“AGCom”)¹² and the Italian Competition Authority (“ICA”) expressed similar allegations¹³ hinting that Amazon’s market position is per-se an issue – an argument that we label as “efficiency offence”.¹⁴ Furthermore, some commentators have argued that Amazon is “too big”, even calling it a monopoly, to justify breaking it up¹⁵; and

3. Amazon is able to leverage its market power in the (upstream) e-commerce segment to strengthen its (downstream) position in e-commerce logistics and B2C parcel delivery. According to AGCom, in Italy, Amazon is the only operator benefitting from vertical integration¹⁶ and its resulting efficiencies, including detailed customer-level information and economies of scale and scope. This, AGCom has argued, can represent a barrier to entry and expansion in the market by 3P logistics and parcel delivery operators.

This paper provides a critical assessment of these allegations, which were at the core of the three aforementioned investigations, presenting the key non-confidential economic analyses and facts. This paper aims to fill a gap between, on the one side, confidential, technical, and lengthy submissions during these investigations concerning Amazon’s allegedly abusive conduct and, on the other side, the genuine interest from the public to understand the facts and analyses in a way that includes Amazon’s perspective.

In Section 2, we provide a short overview of how Amazon achieved its current position and the role logistics played in the Store’s growth. Understanding the interplay of Amazon’s core guiding principles (most importantly “customer obsession” and “insist on the highest standards”), its business model (increasing the long-term value of the Store, including by opening it up to efficient and reliable sellers and carriers), and monetization (obtaining compensation from both sales of 1P Retail and 3P Sellers) is key to assess whether Amazon engages in conduct that harms competition. Amazon’s strategy has focused on delivering: (i) low prices to attract customers offering them the best deal; (ii) a wide selection of products from which customers can choose what fits them best; and (iii) convenience in terms of product availability, quick delivery and customer experience in the Store.

In Section 3, we provide a summary of the main arguments rebutting the key allegations put forward in the three aforementioned investigations against Amazon. Contrary to these allegations, the incentive for Amazon to self-preference 1P Retail and FBA is by no means inevitable nor likely since Amazon monetizes both 1P (product margin) and 3P (Store commission) sales. If 3P Sellers can bring a wider selection and competitive prices to the Store, or if 3P carriers are efficient (i.e. they operate at low costs and high performance), it is not in Amazon’s interest to artificially undermine them as this would reduce the value of the Store. In contrast, Amazon has always adopted an inclusive strategy, integrating efficient players in the Store. Sales made by 1P Retail reflect competitive pricing and delivery performance rather than anticompetitive self-preferencing (which would reflect exactly the opposite conduct, i.e. inferior product choices and customers abandoning the Store over time). Amazon aims at ensuring customer satisfaction and meeting their expectations, irrespective of whether logistics services are performed via FBA or (efficient) 3P carriers. Since customers can buy the same products online or in physical stores (where products are readily available), offering speedy and reliable deliveries is crucial for Amazon to better compete with brick-and-mortar retailers.

As stated by the General Court in the Google Shopping judgment, self-preferencing is a neutral concept, which can characterize both pro- and anti-competitive conduct. The judgement emphasizes that “the mere extension of an undertaking’s dominant position to a neighbouring market [generally described as “leveraging”] cannot in itself constitute proof of conduct that departs from normal competition, even if that extension leads to the disappearance or marginalization of competitors”.¹⁷ Contrary to Amazon’s conduct and incentives described below, anticompetitive self-preferencing (as affirmed in the same judgment) must be established in relation to: (i) conduct that must have actual or potential anticompetitive effects; and (ii) must depart from what would be expected under normal competition.¹⁸ For example, it could amount to anti-competitive self-preferencing if a dominant firm provides an undue advantage (e.g. premium display) to its own products or services when offering them for sale to consumers, despite their objectively inferior features compared to competing third-party products, without any reasonable justification other than its higher profitability. As we will show, this clearly does not reflect Amazon’s conduct, which includes and promotes efficient 3P Sellers and 3P carriers and aims at providing customers with the best possible purchase experience.

We conclude that once Amazon’s business model and conducts, including its positive impact on consumer welfare in what is a highly competitive retail sector, are properly considered, allegations on its position, leveraging and self-preferencing turn out to be unfounded.

2. Amazon’s value proposition and business model

Amazon’s original motto when it was founded in 1994 was to become the “Earth’s biggest bookstore”.¹⁹ Amazon’s first product and its strategy to become the biggest bookstore provide a valuable example to understand how it works.²⁰

During the “90s, internet usage was experiencing triple-digit year-to-year growth rates. Amazon wanted to ride the wave creating a new and compelling experience for book-buying customers. First, books were non-perishable, relatively lightweight and came in fairly uniform sizes, meaning they would be easy to store, pack and ship. Second, books were the product category with the most individual items. While more than 100 million books have been written and more than a million titles were in print in 1994, even the largest US bookstore²¹ at that time could not stock more than tens of thousands of titles. Amazon’s intuition, which then translated to other products, was that an online bookstore would give customers an unprecedented selection.

Fast forward to today: Amazon now sells a large variety of products, fulfilling orders rapidly in more than 100 countries, and provides advertisement, web services, data centers, streams television content and, after its acquisition of Whole Foods in 2017, it has also expanded its footprint as a physical retailer.²² To understand Amazon’s conduct and frame it correctly against the allegations²³ that will be discussed in detail in Section 3, one needs to start from Amazon’s customer-centric business model.²⁴

Amazon has always been “customer-obsessed”, continuously innovating and expanding by starting from how to best accommodate customers’ needs.²⁵ It has adopted a consistent strategy of “working backwards” from the desired customer experience to then decide what to build. This systematic way to critically examine ideas and create new products starts by defining the desired customer experience and then iteratively working backwards to the point when the product/service is released. As noted above, Amazon’s guiding principles have been price, selection, and convenience.

While 3P Sellers and 3P carriers are of great value to the Store, their incentives are not necessarily aligned with those of Amazon. Specifically, they may not take into account that a bad customer experience, such as a defective product or a missed delivery, will negatively affect customers by driving them away from the Store for their future purchases. Such poor customer experience is likely to have a significant and long-lasting negative impact on the value of the Store. For example, survey evidence confirms that late or inaccurate deliveries do have a negative effect on future purchases, as “65% of consumers polled report that they will abandon shopping with a retailer altogether after 2–3 late deliveries”.²⁶ Misaligned incentives, therefore, pose a great threat and limit to Amazon’s business model and expansion. Due to the attractiveness of the Store, which enables 3P Sellers to reach many customers worldwide instantaneously, some 3P Sellers have the incentive to free-ride on Amazon’s reputation and, driven by short-termism, make quick profits in the Store. We discuss this in more detail below.

2.1. Amazon’s store

Amazon is, first and foremost, a retailer. It operates the Store as a whole to succeed, which requires returning customers over time. The Store is successful because it is run as a single Store that offers both 1P Retail and 3P Sellers’ products. That was the central business innovation that Amazon hit on – after several failed attempts – more than 20 years ago. Over time Amazon realized that the best way to accommodate customers’ needs was to open its Store to 3P Sellers, as they increased product selection and drove down prices. This has led to a great expansion of the Store, from which Amazon benefits since it monetizes on both 3P Sellers (which pay a commission to Amazon per sale completed, irrespective of their delivery choice) and 1P Retail (for which Amazon makes a margin on the products it sells). Over the last 15 years, 3P Sellers have grown at twice the pace of 1P Retail, and, as of 2023, 3P Sellers account for 60% of the Store’s sales.²⁷ On average, sales by 3P Sellers are “more profitable for [Amazon] than an Amazon retail sale.”²⁸ As mentioned in the ICA A528 Decision, this holds true also for the profitability of fulfilment channels: “due to the significant investment and operating costs of FBA” Amazon’s operating margin is higher on seller fulfilled sales than FBA sales, despite the fee on logistics it receives on top of the Store’s commission.²⁹

The Store today is the culmination of a process of experimentation with different ways to integrate third-party offers and services into the Store. In its efforts to build stores that attract customers, Amazon has experimented with different models. To expand selection, Amazon experimented with different ways of including the offers of independent sellers in the Store. First, following the eBay model, Amazon created a separate store with auctions for products (Amazon Auctions), but the selection and user experience failed to attract customers. Amazon’s next attempt, ZShops, involved a separate online store for third-party products with set prices, as opposed to bidding. This experiment failed as well. Ultimately, Amazon found that while customers wanted increased product selection, they did not want to have to switch between two separate storefronts to access it. Such cognitive overload resulting from increased variety, which is in line with academic research,³⁰ helps explaining why Amazon has chosen not to display multiple Featured Offers next to product pages in the past.³¹ Multiple Featured Offers, in addition to being disliked by customers, could have also resulted in softening competition among 3P Sellers, who would not be as incentivised as in the baseline scenario to compete aggressively. Most importantly, over time, it became evident that, all else equal, customers would rather buy from Amazon than from 3P Sellers. Recent survey evidence has shown that more than 74% of customers prefer 1P Retail products compared to seller-fulfilled 3P Sellers’ offers,³² which has strong implications with respect to self-preferencing allegations, as it shows that it is customers who prefer Amazon’s products and delivery.

Amazon invests in supporting 3P Sellers in a variety of ways (discussed below) – which it would not do if its aim was to self-preference 1P Retail. These investments heavily support 3P Sellers’ growth with tools and resources that encourage them to list products and provide attractive pricing and services. 3P Sellers’ success is reflected, as shown in Table 1, in their worldwide share of sales in the Store, which grew from 26% in 2007–60% in 2023.

Table 1. Amazon 3P Seller Share 2007–2023 (worldwide)

Share of paid units sold by 3P Sellers in the Store
2007	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022	2023
26*	40*	43**	46	49	51	53	53	54	56	57	60

Source: Statista, “Share of paid units sold by third-party sellers on Amazon platform from 2nd quarter 2007 to 4th quarter 2022”, https://www.statista.com/statistics/259782/third-party-seller-share-of-amazon-platform/?locale=en. Note: data is obtained by taking the average of quarterly data. * 2nd quarter only; ** 3^rd and 4th quarters only.

In the EU, Amazon offers 3P Sellers access to numerous tools to manage the various aspects of their listings in the Store in a convenient manner, including creating new listings, managing inventory and orders, obtaining information on the performance of their products, and communicating with customers. Via Seller Central,³³ Amazon provides 3P Sellers with functionalities at no extra cost (i.e. included in the Store’s fee) such as:

• Inventory management: 3P Sellers can manage their listing information, inventory levels and global listings, including by receiving personalized recommendations on restocking inventory and managing excess, old, and inactive inventory.

• Pricing: 3P Sellers can access product-level pricing information (including sales rank, estimated fees, lowest price, and the price in the offer listing page) and manage their pricing by setting up pricing alerts and automated pricing rules³⁴ which take account of the Featured Offer price through built-in pricing tools.

• Information on traffic: Sellers can access a number of business reports that include product-level information on traffic in the Store such as the number of glance views or sessions count; and

• Brand Analytics tool: Brand owners can track the performance and customer engagement with their brand. 3P Sellers analytics reports provide them with access to a wealth of detailed information such as search frequency rank (which looks at the popularity of keywords), top clicked products, click shares, and conversion shares (the percentage of total sales received by a product for a given keyword). In addition, 3P Sellers receive product opportunity recommendations based on what existing business customers are looking for.

These are all services that, in the EU, enable 3P Sellers to thrive in the Store, contributing to Amazon’s long-term growth objective of providing customers with the best price, selection, and convenience. Amazon’s presence and initiatives in Italy, such as “Amazon Accelerate” and “Made in Italy” storefront, are indicative of its efforts to enhance the Store’s value through the increasing presence of 3P Sellers.³⁵ If Amazon had self-preferencing incentives, it would not provide these tools to 3P Sellers since these tools help them “take away” sales from 1P Retail.

Not only does Amazon lack self-preferencing incentives – it rather is a business enabler, facilitating operations for Sellers. Thanks to Amazon’s services (including the tools discussed above and logistics services), and similar services from other providers, firms can easily start a business without leasing a building or even having employees. Selling on Amazon and using FBA provides a way to reach consumers shopping in Amazon’s stores worldwide, a competitive logistics and delivery service provided by most efficient 3P carriers, analytics tools, as well as customer service (comprising of billing, returns, refunds and complaints handling). These figures are in line with the 2022 survey published by Nomisma,³⁶ which found that Italian businesses selling and partnering with Amazon have increased their sales thanks to more international demand for their products and an improved ability to satisfy customers’ needs.

This said, the incentives of 3P Sellers are not always aligned with Amazon’s, which threatens its business model. As the Store grew, counterfeit, unsafe and expired items made their way into the Store threatening Amazon’s reputation and threatening to deplete the trust it had cultivated with its customers. Driven by Amazon’s popularity and size, some 3P Sellers obtain short-term gains at the expense of long-term customer experience and value of the Store.³⁷ For example, in 2015, self-balancing scooters (“hoverboards”), were found to have non-compliant electrical components that could explode or catch fire. The US Consumer Product Safety Commission eventually investigated 52 fires in 24 states, causing significant property damages.³⁸ In 2021, counterfeit Salvatore Ferragamo’s fashion clothes appeared in the Store, violating Amazon’s policies and Ferragamo’s intellectual property rights.³⁹ Such misconduct constitutes a systematic strategy by some 3P Sellers rather than an occasional phenomenon. These negative actions, mostly implemented by companies that are formed and disappear overnight for this specific purpose, have a negligible impact on that specific seller (since they expect to be out of the market and then reopen with a different name, using the same strategy repeatedly), but a large negative effect in the Store as a whole – affecting all the other 3P Sellers, 3P carriers, and Amazon itself.

This is why Amazon has set policies in place to prevent bad customer experience, for example, by suspending the sale of suspect listings, blocking suspect accounts and IP infringement, removing suspect listings in response to rights owner notices or complaints from customers, and suppressing fraudulent reviews. The aim of such policies is to guarantee a continuous trustworthy shopping and selling experience. Amazon invested $1.2 billion and employed over 15,000 people in 2022 dedicated to protecting the business and end users of its Store.⁴⁰ Allegations of “self-preferencing”, discussed in Section 3.2.2, are not supported by the evidence and Amazon’s motives. The algorithm selecting the Featured Offer, for example, is designed to display the offer that customers would likely purchase if they were to compare all available offers for the same product.⁴¹ The ranking considers both price and non-price offer attributes, which are weighted based on customer purchasing patterns.⁴² To the extent that FBA offers are featured more frequently than MFN offers, this simply reflects that customers, on average, attach greater value to faster and more reliable delivery. This is not self-preferencing, but rather reflects customers’ preferences. In fact, high-quality MFN offers from reliable 3P Sellers are perfectly capable of achieving the same algorithm ratings as FBA offers, as evidenced by the significant proportion of MFN offers being selected as Featured Offers (also when FBA offers are available). On Amazon’s Italian website in 2019 about 40-50% of 3P Sellers used MFN to ship their products.⁴³ About 60-70% of 3P Seller’s sales (both in terms of volume and value) were made through the Featured Offer and 30-40% of these sales were MFN.⁴⁴ As discussed above, Amazon needs to correct for distortive effects in its Store and make sure that customers get the best experience. Since Amazon’s incentives are generally aligned with customers’ needs (also thanks to its vertical integration), which is not true for 3P Sellers or 3P carriers, Amazon will aim to select players which maximize customers’ satisfaction. Nonetheless, there are transparent policies that allow 3P Sellers to appeal against these actions if they feel that their rights have been infringed.⁴⁵

In conclusion, the Store is successful because it is run holistically as a single Store. Amazon's operations are designed to enhance the customer experience and increase the attractiveness of the Store as a whole (including offers from both 3P Sellers and 1P Retail) vis-à-vis other retail stores. Amazon has no incentive to foreclose efficient 3P Sellers or “self-preference” 1P Retail since it profits from both – but most importantly, aims to achieve long-term customer satisfaction leading to repeated sales. Self-preferencing would be a myopic and unprofitable strategy, reducing the Store’s value. Understanding the fundamental difference between aligned and misaligned incentives, along with Amazon’s holistic business strategy, its monetization, and long-term growth objective, is key to the conclusion that Amazon has no incentive to harm or disadvantage players benefitting the Store.

2.2. Amazon’s logistics is ancillary to its Store both for 3P Sellers and customers

Logistics services have been developed as ancillary to Amazon’s Store and its 1P Retail activity.⁴⁶ Logistics services increase the value of the Store for all actors involved: Amazon 1P Retail, 3P Sellers and customers. It provides customers with a reliable and fast service while 3P Sellers grow their business without needing to invest in their own logistics network, which lowers barrier to entry for 3P Sellers. While below we discuss Amazon’s logistics evolution in the US, this has relevant implications for the three investigations we address in this paper, since it shows how, over time, Amazon’s logistics evolved to better serve its customers who demanded fast and reliable deliveries. Amazon revolutionized logistics to accommodate it to the needs of e-commerce, and redesigned warehouses (named “Fulfillment Centers”) and designed state of the art software to manage inventory. The ability to efficiently and predictably fulfill customer orders allowed Amazon to expand into new product categories (including high-value goods such as jewels and perishable goods such as vegetables) and eventually to introduce Prime, an optional, paid subscription program that offers fast and reliable shipping at no extra cost.

As for the Store, this has been a long process marked by many trials and errors. In the US, before Amazon was able to develop its current network, its fulfilment centers were located in areas close to hubs of large logistics and delivery service providers, e.g. UPS and FedEx. These fulfilment centers were enormous buildings mostly located in remote areas allowing Amazon to minimize costs, but not to best serve e-commerce customers. They relied on workforce walking on average about 19km a day to find and pick the right items on shelves before these were then packed and shipped to customers. From 1999 to 2017 everything changed. As Amazon’s CEO stated in the 2021 letter to shareholders, “[i]n the early 2000s, it took us an average of 18 hours to get an item through our fulfillment centers and on the right truck for shipment. Now, it takes us two. To deliver as reliably and cost-effectively as we desire, and to serve Amazon Prime members expecting shipments in a couple of days, we spent years building out an expansive set of fulfillment centers, a substantial logistics and transportation capability, and reconfigured how we did virtually everything in our facilities⁴⁷".

Total delivery time is largely determined by two factors: the distance between the fulfilment center and the customer delivery location, and the shipment method (e.g. truck, van, rail, sea, air, express courier, etc.). Originally Amazon’s fulfilment center network in the US benefitted from workforce and land in areas with lower costs of living and close to large 3P carriers – but distant from its customers. They were placed in locations far away from city centers. It was convenient for Amazon but not in line with the growing expectations of e-commerce customers to get products delivered “fast and free”. Christmas 2013 was a turning moment for Amazon’s logistics. In the US, UPS was Amazon's primary shipping partner and was overwhelmed by the volume of last-minute e-commerce orders in addition to experiencing disruptions in its operations due to severe weather conditions. UPS failed to deliver hundreds of thousands of Amazon orders in time for Christmas – a negative externality for Amazon creating a bad customer experience and lowering the value of the Store. As an article put it, “Santa can deliver millions of packages in a single night, but this Christmas, UPS wasn't up to the task.”⁴⁸ This “Christmas fiasco” and the inability of 3P carriers to deal with peak demand led to a complete rethink of Amazon’s logistics strategy to accommodate delivery needs. This also shaped how Amazon built its logistics network in the EU by taking into account its experiences in the US.

While Amazon already ran its fulfilment centers to serve e-commerce customers seven days a week, major US carriers like UPS and FedEx did not operate on Sundays or national holidays. This was not in line with e-commerce needs and the expectations from Amazon customers. As carriers had less incentives to provide a high-quality delivery than Amazon, Amazon decided it could not solely rely on 3P shipping companies to support its growth but would also have to build an in-house logistics network facilitating the fast transportation from the fulfilment centers all the way to the customers.

To optimize delivery speed Amazon built fulfilment centers located closer to its customers. Thanks to its substantial investments in warehouse robotics⁴⁹ and logistics, Amazon permanently raised the bar for convenience in online shopping that in turn changed the types of products shoppers were willing to buy online. Logistics and delivery worked as a bridge between the online and physical worlds. Customers often purchase the same product online and in brick-and-mortar stores (where products are readily available). Therefore, offering speedy and reliable deliveries is at the core of Amazon’s strategy to be a valid alternative to physical retailers. Ultimately, after introducing the Prime⁵⁰ membership program, Amazon became an alternative to the immediacy of brick-and-mortar stores. Prime raised customer trust in e-commerce providing a compelling game-changing customer experience, resulting in a strong driver of the Store’s growth.

This was made possible by Amazon’s adoption of new technologies and constant innovation. For example, in Italy Amazon does not own or operate trucks or vans. It relies completely on 3P carriers that do the actual shipping across logistics facilities and to final customers. Amazon allocates deliveries to its logistics partners depending on their relative efficiency to meet customers’ expectations. Its main objective is to have the products sold in the Store delivered reliably and quickly, thereby fulfilling the “promise” it made to customers at the moment of their purchase. 3P Sellers make investments to expand their presence on Amazon and would not necessarily do so if they were concerned that fulfilments were of low quality, limited in choice and/or high-priced. Thus, it would represent an issue for Amazon should competing carriers experience demand contractions that reduce their sales below a minimum efficient scale. With insufficient shipping volumes, 3P carriers would not be able to generate sufficient economies of scale and thus set a competitive price. This would lead to carriers going out of the market – which, as a consequence, would discourage 3P Sellers’ investments in the first place, thereby damaging the Store’s value.

In contrast to the allegations of Amazon self-preferencing its logistics, made during the three aforementioned investigations and discussed in depth in Section 3.2.2, Amazon has strong incentives to include high-quality third-party operators in its service. And in fact, it heavily relies on 3P carriers for its own logistics operations as well as for FBA logistics services provided to 3P Sellers. However, there are important “externalities” within the Amazon Store, which means third-party logistics operators will have weaker incentives to optimize and invest in improving delivery performance (as exemplified by the “Christmas fiasco” anecdote).⁵¹ Most of the negative consequences of poor delivery performance will be felt by Amazon itself due to customers leaving the Store. Amazon thus has strong incentives to ensure that delivery performance meets customer expectations regardless of the provider of the logistics service. Considering the delivery speed and reliability when selecting the Featured Offer (both factors highly valued by customers) fundamentally differs from anti-competitive “self-preferencing” favoring 3P Sellers using FBA without justification. Amazon has no incentive to foreclose efficient third-party parcel delivery operators. On the contrary, it has all the incentives to integrate and rely on them (as it already does) to manage deliveries to customers, as long as they can provide fast and reliable deliveries at a low cost, i.e. efficiently.

Overall, as we will discuss in Section 3 in relation to the ICA and AGCom investigations in Italy, FBA deliveries are generally preferred by customers and 3P Sellers because they are faster and more reliable (more likely to arrive on time) than seller-fulfilled deliveries. Furthermore, FBA is priced competitively with the options offered by traditional carriers, despite its better quality. The great majority of respondents to the Seller survey conducted by the ICA cited competitive pricing as a reason to use FBA. And the most commonly cited reason by 87% of 3P Sellers for using FBA in the ICA Survey is “consumer satisfaction thanks to more punctual deliveries and better management of refunds”.⁵² In conclusion, Amazon logistics has clear benefits for 3P Sellers and improves the overall customer experience.

3. Commenting the allegations put forth in EC AT.40703 Amazon Buy Box, ICA A528-FBA AMAZON and AGCom 94/22/CONS

As the zeitgeist around tech companies has evolved over time, they have gone from innovative disruptors changing the way we live to “gatekeepers”.⁵³ Amazon was no exception to this criticism – its growth, success, and the disruption it brought to some sectors eventually caught the attention of policymakers and regulators.

The criticism⁵⁴ around Amazon tends to focus on its allegedly conflicting role of being at the same time “umpire and player”. As described in Section 2, Amazon is both a retailer in the Store and runs the Store setting the rules of the game. Similarly, it is both a provider of logistics services, and at the same time sets the rules in the Store. This “duality” is often depicted as creating an inherent and unavoidable conflict of interest in which Amazon might foreclose 3P Sellers and 3P carriers, as it might have an incentive to “self-preference” the products it offers directly (1P Amazon Retail) and its logistics services (FBA) over 3P alternatives.

The following discussion aims at navigating this issue by providing an overview of Amazon’s main non-confidential arguments and evidence presented during the (aforementioned) investigations by the European Commission (AT.40703 – Amazon Buy Box), Italian Competition Authority (A528-FBA AMAZON) and the Italian Authority for Communications and Postal Services (AGCom 94/22/CONS). While this paper does not provide a complete and comprehensive overview of all allegations and responses, it provides a framework for understanding Amazon’s business motives, incentives and the main arguments against these allegations.

As mentioned in the introduction, this paper addresses allegations by the EC, ICA and AGCom that Amazon: (i) is “self-preferencing” its 1P Retail arm over 3P Sellers, and 3P Sellers using FBA over MFN; (ii) has grown too much, too fast by expanding significantly in e-commerce logistics and B2C parcel delivery – an “efficiency offence”; and (iii) is able to leverage, thanks to its vertical integration, its market power in the (upstream) e-commerce segment to strengthen its (downstream) position in logistics and B2C parcel delivery. According to these allegations, Amazon, through its alleged market power and anticompetitive behaviour, has supposedly damaged its competitors, be they 3P Sellers, logistics operators, couriers, or competing e-commerce marketplaces.⁵⁵ Given Amazon’s central role in the current e-commerce landscape and the advantages of its fulfilment network, it is alleged that 3P Sellers are incentivised to offer their products on its Store and entrust their logistics to it.⁵⁶ This, it is argued, has resulted in a reduced demand for 3P logistics services for e-commerce, which has damaged both firms that, like Amazon, offer fulfilment services, and couriers that handle the delivery for third-party sellers, that are therefore prevented from achieving economies of scale. As a last stage, harm is allegedly inflicted to competing marketplaces too, as multi-homing by 3P Sellers is reduced due to higher logistics costs, e.g. double warehousing.

3.1. Competitive constraints in e-commerce and logistics

The EC, ICA and AGCom have failed to reflect market realities and most importantly the growing constraints faced by Amazon with a constantly increasing e-commerce market and parcel deliveries. As the evidence discussed below shows, these investigations define narrow antitrust markets and fail to consider the set of appropriate competitive constraints that Amazon faces. Upstream, they define the relevant product market as “the market for marketplace services for third-party sellers”.⁵⁷ This is particularly problematic since it excludes not only offline sales but also any other online sales. Downstream, the ICA defined a national market for e-commerce logistics services, while AGCom defined a national “B2C e-commerce parcel delivery” market,⁵⁸ thus, excluding all e-commerce deliveries to businesses (B2B) as well as all other parcel deliveries not related to e-commerce (B2C and C2X parcels).⁵⁹

Both the e-commerce channel (within the retail segment), and parcel deliveries and logistics have experienced tremendous growth, both worldwide, as depicted in Figure 1 below, and in Italy. E-commerce in Italy has grown substantially: its value has increased from 16.2 billion euros in 2015–35.6 in 2019, and it is projected to grow at annual rate of about 21% from 2020 onwards to exceed 100 billion euros by 2025.⁶⁰ There is room for future growth since Italy has a much lower share of e-commerce penetration (accounting in 2021 for 10.9% of retail sales vs. 26.2% in the UK and 21.9% in Germany).⁶¹ Parcel delivery has increased substantially as well, following the upstream growth of e-commerce. AGCom data show parcel delivery has grown 18.8% YoY from 2016–2020 and is expected to further grow.⁶² The investment of retailers in e-commerce and the emergence of players to facilitate e-commerce – like Shopify or Wix – mean there is an enormous addressable market for logistics operators to compete in – which goes way beyond marketplace sales encompassing all e-commerce sales.

Figure 1. On the left: global e-commerce revenues between 2017 and 2027 (in trillion US$). On the right: global parcel shipping volumes between 2017 and 2027 (in billion parcels); straight arrow on both panels reports the estimated CAGR 2017–2027 in %. Source: Authors' elaborations based on Statista, “e-commerce– Market Data Analysis & Forecast”, March 2023, and Statista, “Global parcel shipping volume between 2013 and 2027 (in billion parcels)”.

Third-party logistics operators are growing and investing in Italy and worldwide. Poste Italiane, DHL and others report growing capital investment. For example, Poste and DHL increased CapEx by over 30% in 2019 and, respectively, by 11% and 30% in 2021, after the economic slowdown in 2020 due to the COVID pandemic.⁶³ AGCom reports Bartolini, DHL, GLS, TNT, Global Express and UPS had an average annual revenue growth of 6.8% over the 2014–18 period.⁶⁴ The logistics sector has been booming, especially during Covid, both in Italy and Europe, a broader trend that reflects growth beyond just Amazon.

On market definition, the EC, ICA and AGCom have failed to engage with the arguments and competitive constraints raised by Amazon in its submissions, summarized below. Before discussing the competitive constraints Amazon faces as a retailer and logistics operator, we briefly address two key concerns of AGCom’s market share calculations.⁶⁵

First, in relation to Amazon’s B2C parcel delivery shares, a major shortcoming in AGCom’s calculations is the inclusion of Amazon’s 1P Retail (“captive”) sales, i.e. sales made by Amazon on its Store and fulfilled through Amazon’s fulfilment network. Captive sales generated within Amazon’s online store for products sold directly by Amazon are not contestable by third-party logistics and postal operators. Since the ICA has assessed the ability of 3P carriers to compete with Amazon’s logistics services for 3P Sellers, it would make sense to consider only contestable logistics sales by firms other than Amazon, i.e. to exclude Amazon’s captive sales. To the extent that shares of supply are used as a proxy/indicator for Amazon’s power in logistics, shares that are calculated to include captive sales do not meaningfully reflect market power in the segment under consideration, i.e. Amazon has market power “over itself” rather than with respect to customers and competitors. This approach by AGCom is in contrast with the EC’s approach in similar past cases, which did not involve Amazon. The presumption is to generally exclude captive sales from the relevant market because such sales are less likely to be an option for customers following a price increase for the non-captive business. This rule is commonly known as the “merchant market rule” and has been applied by the EC in numerous cases.⁶⁶ Given a large share of sales in the Store in Italy are generated from 1P Retail, removing these captive sales results in a lower share.

Second, AGCom acknowledges that the e-commerce channel in Italy is “extremely fragmented” and dynamic. Within this level of fragmentation, based on third-party evidence, AGCom estimates Amazon accounts for (only) 17% of the Italian e-commerce market.⁶⁷ Such a share is clearly inconsistent with Amazon holding market power in e-commerce, and consequently with market power that could be “leveraged” into B2C parcel delivery. If Amazon were to attempt any “leverage” strategy, 83% of the (upstream) e-commerce deliveries – not intermediated by Amazon – would remain contestable for other players or new entrants in B2C parcel delivery. Yet, AGCom refers repeatedly to a “leveraging” theory of harm whereby Amazon would transfer its market power “upstream” in e-commerce to the downstream delivery market.

Since, according to AGCom, Amazon’s increasing importance in logistics is a direct consequence of its leverage due to its upstream market power in marketplace services, below we first review the main constraints Amazon faces upstream in marketplace services. All vertical theories of harm fail if Amazon is not dominant upstream (as it cannot leverage any market power). For an in-depth review of competitive constraints in logistics, we refer to Copenhagen Economics (2021).⁶⁸

3.1.1 Brief overview of Amazon’s significant competitive constraints

Amazon faces competition from a vast and growing number of players. Competitors include a wide range of brands, brick-and-mortar and omnichannel retailers, online retailers, as well as other marketplaces which allow a broad range of vendors to sell via their sites. On each of Amazon’s main product categories, brick-and-mortar and e-commerce retailers, marketplaces, and omnichannel retailers are also active and compete.

The market definition adopted by the EC, ICA and AGCom fails to capture the constraints from the fiercely competitive retail landscape which all retailers face, irrespective of their business models and their sales channel (online, offline, or both). The evidence shows that the product market is wider than online marketplaces, as marketplaces face significant constraints both from Sellers and customers.⁶⁹

First, the retail industry is characterized by smaller net margins than other industries – with a net margin of just 0.64% in e-commerce retail (compared to a general retail margin of 2.35%).⁷⁰ For example, in 2022, Walmart and Amazon had rather small net margins. Walmart had a net margin of just 2.39% while it generated an income of $13.6 billion on revenues of $573 billion⁷¹ by being one of the largest retailers in the world. Amazon had a net margin of less than 2% for several years prior to 2018, and a net margin of less than 2.5% in 2022.⁷²

Second, 3P Sellers can constrain Amazon. For example, 3P Sellers sell through multiple, alternative channels and can substitute away from online marketplaces, both offline and online. They can also start selling through new non-marketplace channels (e.g. their own webpage) quickly and at low cost. Results of the Amazon Seller survey, cited below, show that almost half of sellers in online marketplaces in Italy also use four or more distribution channels. Finally, sellers can substitute partially as well as entirely between online marketplaces and alternative channels (multi-homing). For example, based on the EC’s e-commerce sector enquiry, “more than 90% of respondent retailers use their own online shop, with 61% of respondent retailers using their online shops as their sole online selling channel.”⁷³ The growth in own online stores is, in part, due to entry and expansion from innovative e-commerce solutions providers, website builders, and third-party providers of retail information which help any business to create its own online store at a low cost. Providers like Shopify or Wix have helped a vast number of businesses to set up their own stores quickly, easily, and conveniently.

The findings contained in the Amazon Buyer and Seller Survey (2020) confirm these arguments.⁷⁴ Indeed, demand and supply-side constraints including ease of switching and multi-homing by both consumers and sellers is widespread and indisputable:

1. More than half of consumers reported checking non-marketplace channels (like the brand’s own webstore or brick-and-mortar stores) as well as other marketplaces before purchasing a product on a marketplace;

2. Consumers who purchased a product on a marketplace had used multiple channels and several marketplaces to buy similar products over the preceding year;

3. The majority of marketplace consumers use more than one marketplace for making online purchases;

4. About one-third of consumers reported that they would look elsewhere if prices increased by a small amount (5%) on marketplaces, which is evidence of substitution between channels by consumers; and

5. The great majority of sellers on marketplaces, including a majority of small sellers, were not dependent on a single marketplace for their sales but used other marketplaces, e-commerce channels (such as their own website) or brick-and-mortar channels (this was also confirmed by the ICA survey).⁷⁵

Additional evidence pointing to a wider product market than just online marketplaces comes from the (indirect) constraints imposed by consumers. The Notice on Market Definition acknowledges the relevance of indirect constraints through its assessment of chains of substitution, as well as by referring to “taking into account as well, constraints on substitution imposed by conditions in the connected markets”.⁷⁶ These additional “feedback effects” resulting from competition for consumers should properly be accounted for. The interactions between online marketplaces and Sellers are further constrained by the fact that consumers can and do shop across a variety of sales channels with negligible switching costs, and are highly sensitive to the selection of products offered by different retailers and their prices.

Most consumers use both in-store and online channels when purchasing a product: a majority of customers compare prices over the various channels prior to making a purchase. Many browse the internet in search of available products and then purchase in-store. As further explained below, these (indirect) constraints via consumer behaviour create significant competitive pressure on both online and offline sellers, in that they face competition from both physical and online stores. Amazon faces significant competitive pressure from (physical) stores as consumers gather information on the various different offers available to them and choose where to purchase. Most people carry out “omnishopping” which involves combining visits to stores and online research.

These additional (indirect) constraints make it even clearer that the relevant product market definition cannot be limited to online marketplaces. The market definition assessment by the EC and ICA fails to capture these considerations. The evidence shows that retail consumers are very savvy and shop around to find the products they are looking for by multi-homing and switching between the many distribution channels available. For instance, the results of Amazon's Buyer Survey cited above find that in Italy, 57% of consumers that had made multiple purchases within a given product category had purchased across two or more different sales channels.⁷⁷ As already mentioned, 30% of all consumers reported that they would look elsewhere if prices went up by as little as 5% on online marketplaces. This shows that restricting the analysis on marketplace services for 3P Sellers was too narrow. All online and offline constraints Amazon faces should be appropriately considered.⁷⁸

3.2. An assessment of the theories put forward by the EC, ICA and AGCom

3.2.1. Allegations of Amazon being “too big and too efficient” – an “efficiency offence” and leveraging

Many critics of Amazon assert that it “grew too much too fast, and is too big and too efficient”. For instance, in its resolution 94/22/CONS, AGCom follows this flow stating that: “[..] in only five years of operation it has become the number one player in the B2C delivery market [..]”.⁷⁹

In relation to this argument, which has been referred to as “efficiency offense”, we note that:

1. If Amazon is preferred by customers because it offers a more reliable and efficient logistics service than competing firms, it will naturally grow more rapidly – as a result of competition on the merits. This is perfectly in line with the standard process of competition, which benefits customers, and the empirical evidence;

2. There is evidence that competitive pressure from Amazon has led to improvements in the performance of incumbent postal operators. All the most important multinational parcel delivery and logistics operators are growing, expanding their global revenues and their investments.⁸⁰ AGCom itself indicated how Italian parcel delivery operators have modified their infrastructures to improve their e-commerce deliveries.⁸¹ As described above, global revenues and the capital expenditures for some of the largest logistics operators (UPS, DHL, FedEx, and Poste Italiane)⁸² have increased over time for all operators, apart from 2020 due to the Covid pandemic. Many operators (such as DHL, GLS, the French La Poste, FedEx-TNT) are planning to heavily invest in Italy.⁸³ This shows that the industry is very dynamic and also traditional operators are expanding – in direct contrast to the claim that Amazon is hindering competition; and

3. An “efficiency offence” ignores – and effectively penalises – the fact that Amazon has invested significantly in Italy, taking on a material amount of risk. Since 2010, Amazon has invested over €16.9 billion in its Italian operations between capital and operating expenditure.⁸⁴ These investments have generated beneficial spillover effects on third-party parcel delivery and logistics operators, expanding demand for online shopping and creating infrastructure that can be used by third-party operators in addition to Amazon itself.

Both the ICA and AGCom acknowledge that vertical integration between the Store and logistics generates efficiencies.⁸⁵ Most importantly, it aligns incentives towards long-term Store growth and eliminates “double margins” – i.e. the need to pay a margin to external suppliers to procure a service, instead of producing it “at cost” in-house – which generally lowers the cost of supply, hence lowering prices.⁸⁶ These cost savings, due to strong and sustained competition, are then passed-on to 3P Sellers, which in turn provide final customers with a better service creating efficiencies throughout the entire value chain. Furthermore, AGCom also recognises the benefits that 3P Sellers receive from working with Amazon in the form of lower delivery costs: “[Amazon’s service] allows third-party sellers not only to increase their sales, but also to obtain more advantageous delivery terms than they could negotiate independently with delivery operators”⁸⁷, which, given the competitiveness of the retail industry will ultimately be passed-on to final customers. This is a broader result, recognised, for example, by the EC Non-Horizontal Merger Guidelines, which states that “the integration of complementary activities or products within a single firm may produce significant efficiencies and be pro-competitive” and accepting that such combinations are “generally less likely” to raise competition concerns.⁸⁸ As we explained in Section 2.2 Amazon has shaped logistics in a way to accommodate the newly emerging needs of e-commerce customers, as an ancillary service to its Store. Being an e-commerce retailer provided Amazon with the specific knowledge to design and develop a fast and reliable service to the benefit of customers and 3P Sellers. Amazon’s entry and expansion into logistics is therefore consistent with legitimate, efficiency-enhancing vertical integration.

3.2.2. The EC’s and ICA’s allegations of Amazon anticompetitively “self-preferencing” its 1P Retail sales and FBA

In the Store, Amazon’s 1P Retail arm competes with 3P Sellers. At the same time, some 3P Sellers use Amazon’s logistics (FBA) for all or just a share of their offers, whereas others opt for alternative, seller fulfilled deliveries using their in-house carrier or 3P parcel delivery and logistics operators. As explained above, this allegedly creates a “dual role” and concerns that Amazon “self-preferences”, favouring (i) its own 1P Retail offers at the expense of 3P Sellers and (ii) 3P Sellers that use Amazon logistics (FBA) compared to 3P carriers.

Product offers by different sellers differ with respect to many characteristics (price, Prime eligibility, seller rating, defect rates, rate of returns, delivery speed, etc.). Amazon must trade-off between observable features to identify and select the offer that is likely to be most appealing to the customer (and thereby generate a higher probability of sales conversion). Some of these features (e.g. rates of returns of a specific seller and product) are only observable to Amazon. Anticompetitive self-preferencing cannot be defined simply as an absence of perfect “symmetry” in the treatment of 1P and 3P offers. Rather, it must involve Amazon systematically showing objectively inferior offers (such as, ceteris paribus, 1P Retail or FBA offers with a higher price or lower delivery speed than, otherwise equivalent, 3P Seller offers or MFN offers) to consumers as a result of anticompetitive incentives (e.g. to remove competitive threats and to increase its profits).

Below we summarise the main results of the theoretical and non-confidential empirical analysis performed during the EC’s and ICA’s investigations concluding that Amazon is not self-preferencing. First, there is no logical economic incentive for Amazon to behave in this way, since it would not be in its own (nor its customers’) best interest in the long run. Second, its actual behaviour is inconsistent with self-preferencing. Any deviation from “symmetry” in treatment of 1P Retail and 3P Sellers can be explained by objective quality differences. During the investigations, Amazon brought forward evidence showing that once every observed feature has been accounted for in the selection mechanism, thereby also internalizing any negative externalities 3P Sellers and 3P carriers might exert in the Store and related delivery options, any putative additional advantage favouring Amazon 1P Retail or FBA disappears. Thus, the selection process done by Amazon in the Store between 1P Retail and 3P Sellers’ products, and between FBA and MFN logistics services does not display traits of anticompetitive self-preferencing but is rather in the best interest of consumers.

Amazon’s business model, discussed in detail in Section 2, and monetization strategy (obtaining compensation from both 1P Retail and 3P Sellers’ sales), mean it does not have an obvious incentive to self-preference. The general assumption underpinning claims that Amazon preferences its own 1P Retail offers over those of 3P Sellers is that Amazon has an economic incentive to do so that is counter to consumer interests (as it damages competition). This is in stark contrast with Amazon’s long-term growth goals and customer-centric approach, the key drivers of Amazon’s strategy.

Since, as discussed above, Amazon can monetize the sales made by 3P Sellers through the Store’s commissions earning a 1P Retail margin, there is no clear reason to expect Amazon to self-preference. In fact, as discussed above, Amazon’s profitability data show that, on average, Amazon’s operating margin is higher on 3P sales than on 1P Retail sales.

3.2.2.1. Theoretical analysis of Amazon’s monetization and incentives to self-preference

The key academic reference to understand Amazon’s monetization incentives is Etro (2021).⁸⁹ Amazon’s ability to monetize both 3P Sellers and 1P Retail sales encourages it to prefer models which are cost efficient, as doing so makes the platform more competitive, generating more value for consumers, which can then be appropriately monetized over the long term. This is very different from past antitrust cases involving self-preferencing where certain platforms have either been constrained in their ability to monetize third-party complements (e.g. because of zero price constraints inherent in a “free”, ad-funded model) or had dynamic incentives to snuff out complements which could evolve into competitive threats to the platforms themselves.

The model also shows that to address negative externalities Amazon has an incentive to monitor the quality of 3P Sellers: when Amazon is seeking to maximize long-run profits, it factors in the negative impact of poor service quality driven by the short-term profitability of (some) 3P Sellers compared to what a vertically integrated seller (like Amazon) would do. According to the model, in these circumstances, there are legitimate reasons why a platform might place quality requirements on third parties that would be superfluous in respect of its own operations (since its incentives are generally aligned with maximizing the long-term value of the platform). Let us consider two examples. First, Amazon 1P Retail would never sell counterfeit products. To do so would be both unlawful and undermine the trust that customers have when buying in the Store, thereby jeopardizing its growth. The same is not true for 3P Sellers, some of which sold such products at a cheap price to increase sales in the short-term knowing they would eventually be forced out of the Store. Second, Amazon would never overpromise on delivery speed in order to sell more to customers and then systematically fail to deliver on time. Amazon optimizes to meet the promised delivery date and would rather give a less ambitious delivery promise – and achieving it – than promise to be faster and deliver late. In contrast, MFN Sellers might benefit from overpromising on delivery speed as this makes it more likely for them to make the sale and, if it is late, the primary negative effect is felt across the Store. These findings are relevant to motivate the factual and empirical analysis below. Most importantly, they provide a theoretical underpinning for why there can be no presumption that Amazon will want to self-preference and that alternative, pro-competitive justifications for the empirical finding that, e.g. 1P Retail and FBA offers are more often the Featured Offer, should be taken seriously.

3.2.2.2. Review of empirical findings assessing Amazon’s alleged self-preferencing

We now investigate whether a seemingly “worse” Amazon 1P Retail offer (for example in terms of price and/or delivery speed) being displayed as the Featured Offer in the Buy Box when a (seemingly) “better” 3P Seller offer is available is consistent with self-preferencing.

In the EU, most sales in the Store are made (as we would expect) through the Featured Offer, which presents the “best” of the competing offers depending on variables such as price, quality, seller features, and delivery speed. From the perspective of a customer, the Featured Offer is the most appropriate offer (although all other offers remain available, just one click away).

The fact that 1P Retail compared to 3P Sellers, or FBA offers compared to MFN offers are more often displayed as the Featured Offer is both unsurprising and unrelated to self-preferencing.⁹⁰ A direct, unconditional, comparison of the propensity of 1P Retail, FBA – 3P Seller, and seller-fulfilled 3P Seller offers as the Featured Offer is not meaningful: 3P Seller offers (with or without FBA) differ significantly from 1P Retail offers in many ways. For example, 1P Retail offers are competitively priced since Amazon “constantly compares [its] prices to [its] competitors’ prices to make sure that [Amazon’s] prices are as low or lower than all relevant competitors”.⁹¹ This is not always true for 3P Sellers. In fact, there is a long tail of 3P Sellers with high prices and low delivery speed that seldom make it to the Featured Offer – but precisely because their offer is expected to be of less value for customers. 1P Retail and FBA offers have (on average) better quality (lower price and/or faster delivery) and are thus preferred by customers – hence they are displayed more often as the Featured Offer.⁹²

Amazon’s algorithm determining the Featured Offer in the EU follows a holistic approach, weighing different characteristics (not just limited to price and delivery speed) to identify the offer that will provide customers with the best choice and customer experience. The algorithm selects the offer that customers would likely purchase if they were to compare all available offers for the same product.⁹³ Generally, customers value low prices, good quality, and fast and reliable delivery – 1P Retail incentives are generally aligned to deliver this desired outcome. Therefore, the algorithm is directed at selecting the most relevant offer, not to (unduly) advantage 1P Retail or FBA. In fact, high-quality MFN offers from reliable 3P Sellers are perfectly capable of achieving the same algorithm ratings as FBA offers, as evidenced by the significant proportion of MFN offers being selected as Featured Offers (also when FBA offers are available). In fact, as reported by the ICA, “about [70-80%] of the top-100 3P Sellers on Amazon.it use MFN, and about [40-50%] of the 3P Sellers use MFN only”.⁹⁴ This shows that 3P Sellers using MFN can be highly successful in the Store. The key challenge with assessing the extent of any self-preferencing is that offer quality is multi-dimensional and often difficult to observe. For example, a data-scraping analysis comparing Featured Offers with other offers in the offer listing page is meaningless since it omits fundamental, unobserved variables affecting customers’ satisfaction with the product such as rate of returns and late deliveries. For example, cheaper 3P Seller offers might seem “better” at first, but that could be associated with a 3P Seller with a higher rate of defective products – and thus with costly returns and an unsatisfactory customer experience. Therefore, such analysis would be subject to omitted variable bias and would not yield meaningful results. Thus, an Amazon 1P Retail offer being displayed as the Featured Offer when a cheaper 3P Seller offer is available does not necessarily imply self-preferencing. There may be (many) other unobserved offsetting factors, such as delivery speed or defect rates, that make the Amazon 1P Retail offer the one a typical consumer would prefer.

In conclusion, Amazon does not set the algorithm determining the Featured Offer to maximize short-term profitability or to provide an advantage to 1P Retail or FBA. In contrast, the algorithm is set to maximize long-term sales by providing the customer with the best possible experience and product. In fact, Amazon’s business model is based on providing customers with low prices, wide selection, and convenience.

3.2.2.3. Assessing whether Amazon self-preferences its logistics and delivery service FBA

The key to understanding allegations of Amazon favouring its logistics is to establish factually the price and performance of FBA relative to third-party logistics operators. Market and Amazon data reported in ICA A528 clearly show a stark difference in performance between FBA and third-party logistics. FBA prices are competitive with those of third-party logistics operators, while delivering a better-quality service to final customers. Survey evidence shows that the most common reason given by 3P Sellers to use FBA is its high-quality performance.⁹⁵ The empirical evidence and delivery performance metrics show that FBA is objectively superior to the alternatives, and allegations that Sellers only use FBA due to alleged preferential treatment in the Store are unfounded.

In line with the arguments set out above, there can be no anticompetitive self-preferencing if price and quality differences are not taken into account. Claims about FBA being overpriced⁹⁶ and only used by Sellers because of advantages in the Store (while being unrelated to its delivery performance) are not supported by the evidence. By sanctioning Amazon for better service and quality (an “efficiency offence”), the ICA has abandoned the consumer welfare standard, possibly in favour of an interest in punishing efficiency to advantage large incumbent postal operators.

Specifically, the empirical analysis performed during the ICA investigation showed that:

1. FBA deliveries are faster and more reliable than MFN deliveries. The average MFN delivery takes 5–6 days vs. less than 2 days for FBA. Despite having faster delivery times, FBA deliveries are more likely to arrive on time than MFN deliveries (ICA, A528, para. 396).

2. During peak times, MFN underperforms significantly compared to FBA. It has transit times⁹⁷ 300–400% higher than FBA deliveries (ICA, A528, para. 396).

3. FBA is priced competitively. FBA fees are generally comparable to delivery options offered by 3P carriers (ICA, A528, para. 397). FBA is not a standalone profit centre for Amazon (ICA, A528, para. 400) and operating margins on MFN orders are comparable or higher (ICA, A528, para. 391). Survey evidence shows that 64% of respondents to the survey of 3P Sellers conducted by the ICA cited competitive pricing as a reason to use FBA (ICA, A528, para. 400); and

4. Sellers use FBA because it is high-quality. The most cited reason for using FBA (by 87% of Sellers) in the ICA Survey is “consumer satisfaction thanks to more punctual deliveries and better management of refunds”. Amazon’s internal surveys and evidence show a similar picture – 3P Sellers like using FBA because it is of better quality (ICA, A528, para. 398).

As explained in Section 2, being vertically integrated, Amazon has strong incentives to protect the customer experience and ensure high-quality delivery performance. Offering high-quality delivery is important to Amazon, as Amazon cares about consumers returning to the Store for their future purchases – thereby maximizing long-term growth – and not being turned off by a bad experience.

This also results from Amazon’s strategy of including an increasing number of efficient third-parties in the Store. Amazon’s fulfilment network is composed of many high-quality 3P carriers, since Amazon monetizes all sales, whether by Amazon’s 1P Retail or by 3P Sellers (irrespective of their chosen delivery method). If third-party logistics operators are efficient, i.e. they can operate at low cost and high performances, it would not make sense for Amazon to exclude them as this would reduce its sales and profits. This said, there are important “externalities” within the Amazon Store, which means 3P carriers will have weaker incentives to optimize and invest in improving delivery performance than Amazon. Poor delivery (such as missed delivery or broken parcel) performance will marginally harm the 3P logistics operator on the delivery in question (as well as the 3P Seller), but – most importantly – will make the customer less likely to return to and shop with Amazon in the future.

There is wide evidence consistent with such a benevolent view of Amazon’s conduct. First, there is a significant role for 3P logistics operators within and around the Amazon Store: about 60% of 3P Sellers⁹⁸ use MFN and even Sellers using FBA make also non-FBA deliveries. Third-party logistics operators saw their order volumes on Amazon grow over time as they became more efficient. Second, this logic is best illustrated by the mechanisms for granting Prime status to products delivered by 3P logistics operators, which was at the core of ICA A528. The Prime status for 3P logistics operators under the “Seller Fulfilled Prime” (SFP) program is solely granted in recognition of high-performance standards. The Prime label is granted to FBA offers because of (objective) high quality standards, in line with FBA. 3P Sellers can get the Prime label as long as they choose logistic operators that conform to its standards. Therefore, conferring the Prime status to Sellers’ offers using FBA or SFP is based on objective performance metrics, and solely rationalized by Amazon’s long-term growth strategy and customer focus. Specifically, Amazon has strong incentives to expand Prime-eligible selection in order to increase the value of Prime memberships for customers. It makes economic sense to rely on high-performing 3P logistics operators to do this, including them in Amazon’s logistics network or allowing their deliveries to be considered eligible for Prime. Consistent with this, Amazon does heavily rely on efficient parcel delivery and logistics operators for fulfilling FBA orders, in line with its need to maintain high-quality standards. As mentioned above, in 2019, Amazon launched SFP to allow MFN offers to receive Prime status.⁹⁹

In addition to evidence of Amazon including (rather than excluding) efficient third-parties from the Store, there is also evidence that parcel delivery and logistics operators worldwide and in Italy are expanding at a fast pace and heavily investing. These companies are growing and investing in a large and increasingly addressable market. First, e-commerce in Italy has enormous room for future growth, and parcel delivery and third-party logistics operators are growing and investing in Italy and worldwide (as described in Section 3.1). Second, the investment of retailers in e-commerce and the emergence of players to facilitate e-commerce mean there is an enormous addressable market for logistics operators to compete in. Finally, Sellers use multiple sales channels and delivery providers (multi-homing) which remains true also of Sellers using FBA. Several retailers are increasingly expanding their e-commerce and delivery capabilities. Supermarket chains like Esselunga and Carrefour in Italy are examples of integrated retailers that now provide both the possibility to shop online and delivery services. While Esselunga and Carrefour¹⁰⁰ have developed their own carriers, other retailers rely on third-party services, such as UPS and TNT.

In conclusion, there is no evidence of foreclosing effects on traditional parcel delivery and logistics operators. The evidence (including from the ICA’s own survey) does not support this hypothesis and, consequently, the theory of harm fails.

4. Conclusion

Amazon has a “dual role”: it both operates the Store (determining its policies to make sure the Store is providing a good experience for customers) and, at the same time, competes in the Store (with 3P Sellers’ offers, as well as 3P logistics operators serving those Sellers). It is competing in the Store as well as operating it. This is at the core of the three investigations by the European Commission (AT.40703 – Amazon Buy Box), Italian Competition Authority (A528-FBA AMAZON) and the Italian Authority for Communications and Postal Services (AGCom 94/22/CONS). These allegations against Amazon include that it (i) is “self-preferencing” its 1P Retail operations and 3P Sellers using FBA, (ii) has grown too much, too fast and is “too big”, and (iii) is able to leverage its alleged market power in the (upstream) e-commerce segment to strengthen its (downstream) position in e-commerce logistics.

As we have argued, Amazon generally internalizes negative externalities in the Store – thus generating efficiencies. 3P Sellers and 3P carriers, by contrast, do not operate the Store as a whole and therefore do not always have the same incentives as Amazon. They may not fully take into account that a bad customer experience, such as a defective product or a missed/late delivery, will negatively impact all the sellers and carriers in the Store, as well as Amazon itself. Amazon’s customer-centric business model aims at providing customers with competitive prices, wide selection, and convenience (ample availability of products and fast delivery) – irrespective of whether these are 1P Retail or 3P Sellers offers or whether the fulfilment of the order is performed by Amazon or by 3P carriers. This is true because Amazon monetizes both 1P and 3P sales and has a long-term incentive to increase the Store’s value, which is best achieved by widening the selection and competitive prices via 3P Sellers and using efficient 3P carriers.

We conclude that the EC, ICA, and AGCom should have assessed Amazon’s business model more accurately and considered the holistic picture of its conduct. Amazon has been a procompetitive force both in retail and logistics – creating significant benefits for customers as well as 3P Sellers. Once this is properly accounted for, allegations on Amazon’s market position, leveraging and self-preferencing turn out to be unfounded.

Acknowledgement

The authors acknowledge Amazon’s financial support of this research. This paper only contains public information. The arguments, data and evidence discussed in this paper are only in relation to the three investigations mentioned in the abstract, namely: European Commission (AT.40703 – Amazon Buy Box), Italian Competition Authority (A528-FBA AMAZON) and the Italian Authority for Communications and Postal Services (AGCom 94/22/CONS – “Obblighi regolamentari nel mercato dei servizi di consegna dei pacchi”). All opinions expressed in this paper are those of the authors and not necessarily reflect those of Amazon or of Charles River Associates. We thank Dr Oliver Latham, Dr Valter Sorana, Dr Sam Marden, Dr Dominik Fischer, and Dr Francesca Garbin for valuable comments and suggestions.

Disclosure statement

The authors declare that while this research received financial support from Amazon, the contributions, analyses, interpretations, and conclusions articulated in this paper are solely those of the authors. Amazon had no role in the study design, data collection and analysis.

Additional information

Funding

This work was supported by Amazon.

Notes

1 There have also been failures in Amazon's history. The biggest one was Amazon Fire Phone, a smartphone announced in June 2014. In October of the same year, Amazon announced a $170 million write-down "primarily related to Fire phone inventory valuation and supplier commitment costs" in its quarterly earnings report. For a list of Amazon failures see: https://www.businessinsider.com/amazon-products-services-failed-discontinued-2019-3?r=US&IR=T#fire-phone-21.

2 Statista, “Share of paid units sold by third-party sellers on Amazon platform from 2nd quarter 2007 to 4th quarter 2022”, https://www.statista.com/statistics/259782/third-party-seller-share-of-amazon-platform/?locale=en.

3 FBA allow sellers to use Amazon’s logistics centres and services to store, package, and ship their products to consumers. Henceforth, we will refer to these services as “logistics services”.

4 MFN is further divided into two parts: “Seller Fulfilled Prime” (“SFP”) and “MFN Core”. SFP is the Prime program that allows 3P Sellers to deliver customers from their own warehouse and being able to display the Prime badge on their offers in the Store. Throughout this paper we refer to MFN as “MFN Core”, i.e. the logistics and delivery is entirely performed by 3P carriers without Amazon’s involvement. At the time the Italian Competition Authority assessed SFP during the A528 investigation, SFP allowed 3P Sellers to obtain the Prime badge on their products conditional on using 3P carriers that were pre-authorised by Amazon and met certain quality standards.

5 For a high-profile example see US Senator Elizabeth Warren’s statements that “you can’t be an umpire and own one of the teams that’s in the game”. See: The New York Times, “Elizabeth Warren sticks her message in Big Tech’s Face”, June 3, 2019, https://www.nytimes.com/2019/06/03/technology/elizabeth-warren-big-tech-break-up.html.

6 An externality is a secondary effect (positive or negative) of the action of one party experienced by another party. In this case, the poor performance of a Seller or logistics operator will have negative externalities on other Sellers, other logistics operators, and on Amazon itself.

7 European Commission (“EC”), AT.40703 – Amazon Buy Box (December 2022, available at https://ec.europa.eu/competition/antitrust/cases1/202310/AT_40462_8990760_8322_4.pdf).

8 ICA, “A528 – FBA AMAZON, Provvedimento n. 29925” (December 2021, available at https://www.agcm.it/dotcmsCustom/getDominoAttach?urlStr=192.168.14.10:8080/41256297003874BD/0/801201274D8FDD40C12587AA0056B614/$File/p29925.pdf).

9 AGCom, “Delibera n. 94/22/CONS – Obblighi regolamentari nel mercato dei servizi di consegna dei pacchi” (March 2022, available at https://www.agcom.it/documentazione/documento?p_p_auth=fLw7zRht&p_p_id=101_INSTANCE_FnOw5lVOIXoE&p_p_lifecycle=0&p_p_col_id=column-1&p_p_col_count=1&_101_INSTANCE_FnOw5lVOIXoE_struts_action=%2Fasset_publisher%2Fview_content&_101_INSTANCE_FnOw5lVOIXoE_assetEntryId=26689634&_101_INSTANCE_FnOw5lVOIXoE_type=document).

10 At the time of the ICA and EC investigations, Amazon’s Buy Box displayed the “Featured Offer” of one single seller and allowed products to be easily purchased by directly clicking on a “buy” button. The Featured Offer has been designed by Amazon to simplify the shopping experience for customers when there are multiple merchants offering the same product. It does so by featuring the offer that Amazon’s algorithm determines customer are most likely to prefer based on a combination of objective parameters such as price and delivery speed. Becoming the “Featured Offer” (i.e. winning the Buy Box) provides sellers (1P Retail or 3P Sellers) with more visibility to consumers and higher likelihood of converting their offers into actual sales offers. Offers that are not surfaced to customers as Featured Offers are still accessible (one click away) in the Offer Listing Page and can be purchased.

11 For example, US Senator Elizabeth Warren’s stated that “Amazon has the power and uses it to smash little businesses before they ever get a foothold,”. See: CNN Business, “Elizabeth Warren says Amazon is ‘like a monster’ that must be fed every minute”, by Matt Egan, October 15, 2021, https://edition.cnn.com/2021/10/15/business/amazon-elizabeth-warren/index.html.

12 Specifically, AGCom 94/22/CONS argues that the “significant market power” Amazon allegedly holds in the “Italian B2C e-commerce parcel delivery market”, in connection with its vertical integration, is the main ground for why intervention in B2C parcel delivery in Italy is needed.

13 Amazon claimed that given the absence of a coherent and well-defined theory of harm by the ICA, it “would violate competition law because of its superior efficiency and ability to exploit economies of scale, which other logistics operators would not be able to replicate”, i.e. an “efficiency offence” (ICA A528, paragraph 404-405).

14 Commenting ICA A528, Sabino Cassese, Justice Emeritus of the Italian Constitutional Court and former Minister for Public Function, argued “we have stepped back more than a century, we have gone back to the earliest steps of antitrust. In essence, the mere motivation [of the decision] is the most basic one from the earliest history of antitrust legislation in the world. That is, that being big is a bad thing that needs to be sanctioned.” (emphasis added). See: Huffington Post, “Antitrust contro Amazon. Chi vince secondo Sabino Cassese (e i tempi moderni)”, by Fabio Insenga, February 8, 2022, https://www.huffingtonpost.it/economia/2022/02/08/news/antitrust_contro_amazon_chi_vince_secondo_sabino_cassese_e_il_buon_senso_-8689862/.

15 See for example, the May 2022 bill “Competition and Transparency in Digital Advertising Act” (CTDA), available here: https://www.lee.senate.gov/services/files/7384B096-04C3-4A3A-9796-80D22483026F. See also The Wall Street Journal, “The Case for Splitting Amazon in Two”, by Joe Lonsdale, February 7, 2022, https://www.wsj.com/articles/split-amazon-in-two-prime-web-services-aws-logistics-third-party-earnings-report-consumers-antitrust-11644249482?mod=article_inline.

16 AGCom 94/22/CONS, paragraph 45: “As of today in Italy Amazon is the only platform that is vertically integrated in the delivery sector to distribute the products sold online”. See also paragraphs 71–73.

17 General Court Google Shopping judgement, November 10, 2021, available at: https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:62017TJ0612, paragraph 162.

18 General Court Google Shopping judgment, paragraphs 195, 437 and 438. Also, the EC Digital Markets Act (“DMA”) defines self-preferencing in the following way: “The gatekeeper shall not treat more favourably, in ranking and related indexing and crawling, services and products offered by the gatekeeper itself than similar services or products of a third party. The gatekeeper shall apply transparent, fair and non-discriminatory conditions to such ranking.” (DMA Art. 6(5)).

19 The Robin Report, “Amazon … From Earth’s Biggest Bookstore To The Biggest Store on Earth?”, by Robin Lewis, January 24, 2012, https://www.therobinreport.com/amazon-from-earths-biggest-bookstore-to-the-biggest-store-on-earth/.

20 For a thorough overview of Amazon’s initial years and the evolution of the Store see B Stone, The Everything Store (Little, Brown & Company, 2013).

21 Barnes & Noble, as reported in B. Stone, Amazon Unbound: Jeff Bezos and the Invention of a Global Empire (Simon & Schuster 2021).

22 For an overview of Amazon’s work style and strategy: C Bryar and B Carr, Working Backwards: Insights, Stories, and Secrets from Inside Amazon (St. Martin’s Press 2021).

23 Throughout the paper we always refer to the allegations and investigations by the European Commission (AT.40703 – Amazon Buy Box), Italian Competition Authority (A528-FBA AMAZON) and the Italian Authority for Communications and Postal Services (AGCom 94/22/CONS).

24 The need to understand how tech companies work has been stressed out repeatedly by tech observers and commentators. For example, Ben Thompson of Stratechery and Ben Evans have remarked how the regulatory discourse around large platforms is naive and uninformed: “We have locked in the tech giants for a generation by not understanding how their businesses work”; “What I see again and again is an insistence on a framework of understanding that is obsolete. And then people wonder why regulation is so ineffective, or why things like GDPR make these companies stronger. This outcome was knowable!”. See Vox EU CEPR, “Designing regulation for digital platforms: Why economists need to work on business models”, by Scott Morton, F. et al., June 4, 2020, https://cepr.org/voxeu/columns/designing-regulation-digital-platforms-why-economists-need-work-business-models.

25 Amazon has been a continual innovator, launching many new products and services over the last three decades, including Prime, Kindle and Amazon Web Services.

26 VoxWare, “Voxware Biennial Holiday Survey Indicates Consumers Less Tolerant for Delivery Delays and Inaccurate Order Fulfilment”, November 22, 2022, https://www.voxware.com/press-releases/voxware-biennial-holiday-survey-indicates-consumers-less-tolerant-for-delivery-delays-and-inaccurate-order-fulfilment/.

27 See Amazon Investor Relations, Third Quarter Results, “WW seller unit mix – % of WW paid units (7)”, https://ir.aboutamazon.com/news-release/news-release-details/2023/Amazon.com-Announces-Third-Quarter-Results/default.aspx

28 Amazon, “Come Amazon e i partner di vendita, insieme, offrono ai clienti una maggiore libertà di scelta”, October 2021, https://www.aboutamazon.it/notizie/piccole-e-medie-imprese/come-amazon-e-i-partner-di-vendita-insieme-offrono-ai-clienti-una-maggiore-liberta-di-scelta.

29 ICA A528, paragraph 391.

30 See, for example, A Chernev et al., ‘Choice Overload: A Conceptual Review and Meta-Analysis’ (2015) Journal of Consumer Psychology; and B McShane and U Böckenholt, ‘Multilevel multivariate meta-analysis with application to choice overload’ (2018) Psychometrika.

31 In December 2022, Amazon agreed to a set of commitments, to address the EC’s Buy Box concern, including “treat all sellers equally when ranking the offers for the purposes of the selection of the Buy Box winner; display a second competing offer to the Buy Box winner if there is a second offer from a different seller that is sufficiently differentiated from the first one on price and/or delivery. Both offers will display the same descriptive information and provide the same purchasing experience.” (https://ec.europa.eu/commission/presscorner/detail/en/ip_22_7777).

32 Econcrew, “From Checking Out to Checkout: What Makes Amazon Customers Buy?”, January 2023, https://www.ecomcrew.com/consumer-behavior-review-2022/#:~:text=More%20than%2074%25%20of%20customers,ergo%20third%2Dparty%20marketplace%20sellers.

33 Further information on Amazon Seller Central is available here: https://sell.amazon.com/tools/seller-central.

34 The Competitive Price Match Rule updates a 3P Seller’s price to “help stay eligible for Featured Offer status and ensure that [the 3P Seller] always matches the competitive price (when there is one), to increase [its] chance of becoming the Featured Offer.” See: https://sell.amazon.com/tools/seller-central.

35 For example, “Amazon Accelerate”, devised to address small and medium enterprises (SMEs), offers free “learning and consulting tools for anyone looking to start a new online business or accelerate an existing one, regardless of whether they intend to do it with Amazon or another operator.” See: https://sell.amazon.it/accelera-con-amazon. Another initiative by Amazon specifically designed to promote Italian 3P Sellers abroad is the “Made in Italy” storefront, which Amazon introduced in 2015 to highlight a selection of Italian products to some of its international customers. As of 2023, the storefront hosts more than 1 million products from more than 5,500 artisans and SMEs in 10 different foreign Amazon Stores. Amazon has helped the 21,000 Italian SMEs it hosts on its Store to create and maintain 60,000 jobs and reach €950 million of sales. Amazon, “Report sull’Impatto delle Piccole e Medie Imprese Italiane”, 2023, https://assets.aboutamazon.com/59/a1/50b4ad36474387647d046e58086d/rapporto-impatto-amazon-pmi-italiane-2022.pdf.

36 Nomisma, “Il contributo di Amazon nell’innovazione di impresa: la ricerca Nomisma”, March 17, 2022, https://www.nomisma.it/il-contributo-di-amazon-nellinnovazione-di-impresa/.

37 3P Sellers can damage Amazon’s reputation through multiple strategies, including: (i) offering counterfeit products, thereby infringing brand owners’ intellectual property, as well as selling defective or non-compliant products to customers; (ii) abusing product reviews, thereby attempting to gain unfair competitive advantages by purchasing positive reviews from customers or third-party service providers, creating fake customer accounts to write positive reviews, or leaving negative reviews on competitors’ products; (iii) selling prohibited products, in violation of the law and Amazon’s publicly available policies; and (iv) committing “hit and run” frauds, which consist in creating product listings and accepting payment from customers without fulfilling customer orders.

38 Amazon sold about half of these products. It notified its customers to who bought the hoverboards to dispose them. See Vox, “The long-awaited hoverboard safety crackdown has finally arrived”, by Libby Nelson, February 19, 2016, https://www.vox.com/2015/12/30/10690720/hoverboard-fires-safety.

39 This led Amazon and the Italian luxury brand to jointly file two lawsuits against 3P Sellers that attempted to offer the infringing products on Amazon’s store. See: Amazon, “Amazon and Salvatore Ferragamo file two joint lawsuits against counterfeiters”, 2021, https://www.aboutamazon.eu/news/policy/amazon-and-salvatore-ferragamo-file-two-joint-lawsuits-against-counterfeiters.

40 Amazon, “Brand Protection Report”, April 2023, https://storage.pardot.com/326621/1680574783Y5dqjaG2/Amazon_BPR_English_2023.pdf.

41 ICA A528, paragraph 257.

42 ICA A528, paragraph 258-271.

43 ICA A528, Figure 20.

44 ICA A528, paragraph 271.

45 These rules are driven by the platform-to-business (P2B) Regulation (EU 2019/1150) that aims at providing businesses (e.g. 3P Sellers) with instruments to challenge the decision of the Store, present complaints, and require the activation of a mediation mechanism. Information regarding these procedures is also publicly available on Seller Central. See, for example, the rules regarding the mediation of vendor disputes (https://sellercentral.amazon.it/help/hub/reference/external/G67ETGRC3ZJQBTVT?locale=it-IT).

46 The evolution of Amazon’s logistics in this section is based on: Stone, B. (2021), “Amazon Unbound: Jeff Bezos and the Invention of a Global Empire” Simon & Schuster, Chapter 9: The Last Mile.

47 Amazon, "Annual Report", 2021, https://s2.q4cdn.com/299287126/files/doc_financials/2022/ar/Amazon-2021-Annual-Report.pdf, page 3.

48 Wired, “Christmas Delivery Fiasco Shows Why Amazon Wants Its Own UPS”, by Klint Finley, December 30, 2013, https://www.wired.com/2013/12/amazon-ups/.

49 Robotics and automatization were two crucial innovations that improved Amazon’s logistics. Robots increased worker productivity and decreased the rate of growth of Amazon seasonal labor needs relative to its sales. They also allowed Amazon to build denser fulfilment centers. In 2014 Amazon estimated it was able to get 50% more products per square meter into new fulfilment centers than the previous generation. The robots also transformed labor that was physically exhausting. Instead of pickers walking about 19km a day to select items from shelves spread out over giant warehouses, robots maneuvered portable containers of merchandise around the building. See Stone, B. (2021), “Amazon Unbound: Jeff Bezos and the Invention of a Global Empire” Simon & Schuster.

50 Vox, “The making of Amazon Prime, the internet’s most successful and devastating membership program”, by Jason Del Rey, May 3, 2019, https://www.vox.com/recode/2019/5/3/18511544/amazon-prime-oral-history-jeff-bezos-one-day-shipping.

51 This is a phenomenon widely recognised by the EC, ICA, and AGCom. For example, the EC vertical guidelines (paragraph 13) note that: “Vertical externalities arise because the decisions and actions taken at different levels of the production or distribution chain determine aspects of the sale of goods or services, such as price, quality, related services, and marketing, which affect not only the undertaking making the decisions but also other undertakings at other levels of the production or distribution chain. For instance, a distributor may not gain all the benefits of its efforts to increase sales, as some of those benefits may go to the supplier.” See: EC, “Guidelines on vertical restraints”, May 10, 2022, https://competition-policy.ec.europa.eu/system/files/2022-05/20220510_guidelines_vertical_restraints_art101_TFEU_.pdf.

52 ICA A528, paragraphs 392-400.

53 See for example: The Furman, Stigler, and Cremer Reports, ‘Unlocking Digital Competition - Report of the Digital Competition Expert Panel’ (2019) HM Treasury; F Scott Morton, et al., ‘Report of the Committee for the Study of Digital Platforms’ (2019) Stigler Center for the Study of the Economy and the State; J Crémer et al., ‘Competition policy for the digital era’ (2019) European Commission. Multiple agencies’ studies have set out how internet aggregators, online marketplaces and social networks benefit from direct or indirect network effects and scale. These agencies have evaluated how these forces have led to high concentration and winner-takes-all dynamics, and how the ability to collect and exploit data at scale plays a critical role; see e.g. CMA (2019), “Online platforms and digital advertising. Market study interim report”, or the Australian Competition and Consumer Commission (ACCC) (2019), “Digital platforms inquiry – final report”, or Bundeskartellamt (2019), “Algorithms and Competition”.

54 See L Khan, ‘Amazon's Antitrust Paradox (2016) The Yale Law Journal, and D Klingler et al., ‘Amazon’s Theory of Harm’ (2020) Yale University, for a comprehensive review of allegations against Amazon.

55 The description regarding how this harm was allegedly produced is contained in ICA A528, paragraphs 700-703, and reiterated by P Bougette et al., ‘Self-Preferencing and Competitive Damages: A Focus on Exploitative Abuses’ (2022) The Antitrust Bulletin; and M Motta, ‘Self-Preferencing and Foreclosure in Digital Markets: Theories of Harm for Abuse Cases’ (2022) BSE Working Paper. For a discussion of Amazon’s conduct and how it affects several interlinked markets, see D Geradin and T Smith, ‘Spinning Amazon’s Flywheel: How Amazon's Business Model Harms Competition -- A View from Europe’ (2023).

56 According to the ICA, multi-homing of third-party sellers is further discouraged by the alleged duplication of warehousing costs that these must face after joining FBA (ICA A528, paragraphs 826-831).

57 EC, Decision on case AT.40462 and AT.40703, section 4.2.2, and ICA A528, section V.3.

58 AGCom 94/22/CONS, section 2.2.

59 While the report by Copenhagen Economics provides a full review of market definition, we note that Amazon has a large number of deliveries to businesses and most (non-bulky) B2B and B2C deliveries are indistinguishable from each other.

60 The European House – Ambrosetti, “Italia cashless: cambiamenti in atto e prospettive future”, 2021, https://www.sipotra.it/wp-content/uploads/2021/05/Italia-cashless-cambiamenti-in-atto-e-prospettive-future.pdf, pages 25–26.

61 Statista, Online share of retail trade in selected European countries 2014–2021.

62 AGCom 94/22/CONS, section 2.

63 Poste Italiane 2019 and 2021 data in the respective Financial Reports (https://www.posteitaliane.it/files/1476515475256/Annual-Report-2019.pdf, https://www.posteitaliane.it/files/1476557513302/Annual-Financial-Report-2021-PDF.pdf). Worldwide figures for DHL available at https://www.dpdhl.com/content/dam/dpdhl/en/media-center/investors/documents/annual-reports/DPDHL-2021-Annual-Report.pdf.

64 AGCom, “Principali evidenze contabili delle maggiori società del settore dei servizi postali (2014 – 2018)”, October 2019, https://www.agcom.it/documents/10179/4290580/Documento±generico±29-10-2019/abc889c4-c54c-4184-96b8-9c0d4129fbd0?version=1.0, page 2.

65 While Amazon’s delivery share calculation by AGCom is problematic and flawed, this goes beyond the scope of this paper. For example, AGCom computes Amazon’s B2C delivery shares using Amazon’s B2X revenues in the numerator, while using B2C market totals in the denominator, thus overstating Amazon’s B2C shares.

66 See, e.g. Case COMP/M.1693 – Alcoa/Reynolds; Case COMP/M.6611 – Arla Foods/Milk Link; Case COMP/M.1574 – Kirch/Mediaset; CASE M.8306 – Qualcomm/ NXP Semiconductors; Case COMP/M.6872 – Barry Callebaut/Petra Foods; Goetyn, G. & Ashall, S. (2015), “EU Merger Control: The Relevance of Captive Sales for the Purpose of Market Definition and Competition Assessment”, Journal of European Competition Law & Practice; Case COMP/M.3056 – Celanese/Degussa/European OXO Chemicals. This said, we recognise that ultimately this is a case-by-case decision.

67 AGCom 94/22/CONS, paragraph 39.

68 Copenhagen Economics (November 2020), “The economic rationale for vertical integration in the tech sector”, available at https://copenhageneconomics.com/wp-content/uploads/2021/12/copenhagen-economics-the-economic-rationale-for-vertical-integration-in-tech.pdf.

69 ICA A528, paragraphs 406-418.

70 Stern School of Business, Margins by sector (US) https://pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/margin.html.

71 Macrotrends. Walmart Net Profit Margin 2010–2023 | WMT, https://www.macrotrends.net/stocks/charts/WMT/walmart/net-profit-margin.

72 Macrotrends. Amazon Net Profit Margin 2010–2023 | AMZN, https://www.macrotrends.net/stocks/charts/AMZN/amazon/net-profit-margin.

73 EC (2017), “E-commerce sector inquiry”, Staff Working Document, page 134.

74 Amazon Buyer and Seller Survey (2020), available at “Survey evidence on user multi-homing in online retail businesses”, https://www.compasslexecon.com/wp-content/uploads/2022/07/Survey-Evidence-on-User-Multi-Homing-in-Online-Retail-Businesses.pdf. The survey was conducted in Italy, Germany, France, Spain, and UK.

75 The ICA survey shows that 53% of interviewed retailers were selling on Amazon and only 24.5% were “single-homing” in the Store. See: ICA A528, paragraph 70.

76 EC (2022), “Notice on Market Definition”, paragraphs 56–57, https://ec.europa.eu/commission/presscorner/detail/en/ip_22_6528.

77 See Table 5, Amazon Buyer and Seller Survey (2020) cited above.

78 For a summary of the evidence, see ICA A528, paragraphs 406-418.

79 AGCom 94/22/CONS, paragraph 46. See also AGCom 94/22/CONS, paragraphs 68-69: “Amazon. […] has become the leading player in the e-commerce delivery market due to the competitive advantage it derives […] from vertical integration. There is, therefore, the possibility that Amazon may in the future be the only player able to take advantage of the growth potential of the parcel delivery market resulting from the development of e-commerce, with a negative effect on the competitiveness of the market and, therefore, on the quality of services. [..]”.

80 Among other innovations, Amazon’s pioneering use of robotics in its warehouse and its use of regional fulfilment centers have been replicated by other players, DHL and eBay among others. See: ZDNet, “DHL expands robotic footprint with 1000 autonomous robots”, by Greg Nichols, March 12, 2020, https://www.zdnet.com/article/dhl-expands-robotic-footprint-with-1000-autonomous-robots/; The Verge, “eBay to take a shot at Amazon with end-to-end fulfilment service next year”, July 25, 2019, https://www.theverge.com/2019/7/25/20727220/ebay-managed-delivery-fulfilment-service-amazon.

81 See, among others, AGCom, “Analisi del mercato dei servizi di consegna dei pacchi”, Interim Report, Allegato A alla Delibera n. 212/20/CONS, available at https://www.agcom.it/documents/10179/18568747/Allegato±1-7-2020/0b7aead1-68e3-4716-822d-f07c9db8aba8?version=1.0, paragraphs 28–30 and paragraphs 79-81.

82 For example, in 2017, FedEx launched a multi-channel ecommerce fulfilment service targeting SMEs. The service includes warehousing, fulfilment, packaging, transportation, and reverse logistics. UPS offers an eFulfilment programme, which allows SMEs to manage sales across multiple marketplaces and web stores using a single login, and DHL offers the DHL e-commerce shipping solution. GLS has also developed an e-commerce management platform and has signed a partnership with Prestashop in Italy, an opensource e-commerce platform, to launch a free module designed for SMEs, start-ups, and artisans selling products online.

83 See Ansa, “DHL: in meno di 3 anni 350 milioni di investimenti in Italia”, January 21, 2021, https://www.ansa.it/emiliaromagna/notizie/2021/01/21/dhl-in-meno-di-3-anni-350-milioni-di-investimenti-in-italia_fa73e8ff-9e4f-4fc3-90ee-74efe57c9b8a.html; GLS Newsroom, “VGP Italy e GLS Italy insieme per il nuovo polo logistico internazionale di Sordio”, July 9, 2021, https://www.gls-newsroom.it/news/all/vgp-italy-e-gls-italy-insieme-per-il-nuovo-polo-logistico-internazionale-di-sordio/s/3ca92726-b736-48fd-8ddd-c75f69c1982e; TrasportoEuropa, “La Poste acquisirà controllo di Bartolini“, August 206, 2019, https://www.trasportoeuropa.it/notizie/logistica/la-poste-acquisira-controllo-di-bartolini/; Piacenza24, “Fedex-TNT, investimenti in tutta Italia ma chiude l’hub di Piacenza: “Non svolge più un ruolo centrale””, March 29, 2021, https://www.piacenza24.eu/fedex-tnt/.

84 See “Amazon’s economic impact on Italy: 2022 progress”, October 17, 2023, https://www.aboutamazon.eu/news/empowering-small-business/amazons-economic-impact-on-italy-2021-progress.

85 See AGCom Allegato A alla Delibera n. 212/20/CONS, paragraph 44: “Vertical integration [is] a strategy to exploit economies of scope to enter new markets and achieve greater production efficiency in the initial market”.

86 Double marginalisation “occurs when one or more firms selling to each other along a vertical chain have market power” and therefore apply a mark-up to their prices, as reported by Pindyck, R. (2011), “Lecture Notes on Vertical Structure”, Sloan School of Management – Massachusetts Institute of Technology, https://web.mit.edu/rpindyck/www/Courses/VS_11.pdf.

87 AGCom 94/22/CONS, paragraph 49. See also AGCom, Allegato A alla Delibera n. 212/20/CONS.

88 Non-Horizontal Merger Guidelines, EC, pages 6–25, paragraphs 11 and 13.

89 F Etro, ‘Product Selection in Online Marketplaces’ (2021) Journal of Economics & Management Strategy. See also: F Etro, ‘Device-Funded vs Ad-funded Platforms’ (2021) International Journal of Industrial Organization; A Hagiu, ‘Merchant or two-sided platform?’ (2007) Review of Network Economics; B Jiang, K Jerath and K Srinivasan, ‘Firm strategies in the “Mid Tail” of Platform-Based Retailing’ (2011) Marketing Science; A Hagiu, and J Wright, ‘Marketplace or Reseller?’ (2015), Management Science; A Hagiu and J Wright, ‘Multi-sided Platforms’ (2015) International Journal of Industrial Organization. For a non-technical summary of monetisation in tech see Vox EU CEPR, “Designing regulation for digital platforms: Why economists need to work on business models”, by Scott Morton, F. et al., June 4, 2020, https://cepr.org/voxeu/columns/designing-regulation-digital-platforms-why-economists-need-work-business-models.

90 ICA A528, paragraph 22 and Section III.

91 See: “All You Need to Know About the Amazon Pricing Strategy in 2023”, Intelligence Node, available at: https://www.intelligencenode.com/blog/amazon-pricing-strategies-model/. See also Amazon’s Help & Customer Service on price matching, available here: https://www.amazon.com/gp/help/customer/display.html?nodeId=G9EAYKPV5YYDB8P7

92 ICA A528, paragraphs 369-368.

93 ICA A528, paragraph 144.

94 ICA A528, paragraph 463.

95 ICA A528, paragraph 316.

96 This was also recognised by ICA in A528, stating that pricing of FBA deliveries was in line with competitors and can be “higher or lower than those charged by couriers to individual sellers” (paragraph 184).

97 The transit time is the delivery length, from the moment of collection (i.e. the pick-up at the point of departure) to its delivery. In the comparison of delivery speed across carriers, this metric ensures that upstream logistics is not included.

98 Report sull’Impatto delle Piccole e Medie Imprese Italiane, Amazon, 2022.

99 ICA A528, paragraph 211.

100 See Joinrs, “La logistica in Esselunga”, September 9, 2022, https://www.joinrs.com/it/posts/24788/ and Experience Retail, “Carrefour punta sulla logistica automatizzata”, February 14, 2018, https://www.experienceretail.it/it/news/carrefour-punta-sulla-logistica-automatizzata/.