For over a decade, Amazon operated a dual role as both marketplace operator and competitor to the sellers on its platform. Internal documents obtained by Congressional investigators and journalists revealed a systematic program to harvest sales data from successful third-party products and use it to develop competing Amazon-branded alternatives. When questioned under oath in July 2020, Amazon executives including Jeff Bezos denied the practice existed as policy. The evidence showed otherwise.
Amazon operates the world's largest e-commerce marketplace, processing over $386 billion in third-party seller transactions in 2020. Simultaneously, Amazon competes directly with those sellers through its expanding private label business, which by 2019 encompassed over 158,000 products across 45 house brands. This dual role—platform operator and platform competitor—created a structural conflict that became the focus of antitrust investigations on two continents.
The conflict hinges on information asymmetry. Third-party sellers see only their own performance data. Amazon sees everything: every transaction, search query, failed purchase, and customer preference across millions of products. This God's-eye view of marketplace dynamics represents extraordinary competitive intelligence—the kind companies typically spend millions to obtain through market research. For Amazon, it was simply infrastructure.
The question was whether Amazon used this asymmetric information access to identify successful products and launch competing versions. Amazon's official position, articulated under Congressional testimony in July 2019, was unequivocal: the company had a policy prohibiting use of individual seller data to inform private label decisions. The evidence that emerged over the following year told a different story.
Internal documents obtained by House Judiciary Committee investigators revealed the existence of Project Nessie—an algorithmic system designed to identify private label opportunities by analyzing aggregated sales data across Amazon's marketplace. The system generated regular reports ranking product categories by commercial potential, focusing on items with high sales volume, healthy margins, and fragmented competition among multiple small sellers.
Former Amazon employees described Nessie as central to the private label development process. The system automated what had previously been manual market research, scanning millions of products to surface opportunities. While Amazon maintained that Nessie used only aggregated, anonymized data, the reports it generated often included specific ASINs (Amazon Standard Identification Numbers) and revealed which sellers were driving category performance.
One internal document presented during Congressional hearings showed a Nessie-generated report listing individual products with annotations like "good private label opportunity" and "high margin potential." The document included seller names, sales volumes, and profit estimates—precisely the kind of seller-specific intelligence Amazon's executives had testified the company did not use.
"We are using seller data to launch private label products."
Internal Amazon Email — February 2020, cited in House Judiciary ReportThe system's name itself—Nessie, referencing the Loch Ness Monster—suggested something whose existence was rumored but difficult to prove. Documentation of Nessie's operation was deliberately sparse in official records, appearing instead in internal communications and project files that emerged only through subpoena.
On April 23, 2020, the Wall Street Journal published an investigation that fundamentally altered the trajectory of antitrust scrutiny. Reporter Dana Mattioli had interviewed more than 20 former Amazon employees and obtained internal documents showing that Amazon employees routinely accessed sales information from individual third-party sellers to determine which products Amazon should replicate.
One former employee described accessing a report showing that a specific car trunk organizer from a third-party seller was generating significant sales. This intelligence directly informed Amazon's decision to launch its own version through Amazon Basics. Another employee described analyzing data from a successful electronics seller to reverse-engineer pricing and feature sets for competing Amazon-branded alternatives.
The investigation focused particularly on Amazon's Bangalore, India office, where a significant portion of private label development occurred. Former managers told the Journal that the India team maintained databases of "promising products" with sales data traced to specific sellers. Internal emails showed team members sharing spreadsheets with seller names, ASINs, and sales figures annotated with competitive analysis.
Amazon's initial response to the Journal's reporting was carefully calibrated. The company acknowledged that the conduct described "would be a violation of our policies" and said it was launching an internal investigation. This formulation preserved the claim that such use of seller data violated policy while implicitly confirming the conduct had occurred.
Nine months before the Journal investigation, Amazon General Counsel David Zapolsky had submitted written testimony to the House Judiciary Antitrust Subcommittee. His statement was unambiguous: "Amazon does not use individual seller data directly to compete with those sellers." This clear declarative became impossible to maintain after the Journal's reporting.
When Jeff Bezos appeared remotely before the same subcommittee on July 29, 2020, Representative Pramila Jayapal confronted him with internal emails referencing seller-specific data analysis. Bezos's response marked a careful retreat from Zapolsky's absolute denial:
"We have a policy against using seller-specific data to aid our private label business, but I can't guarantee you that that policy has never been violated."
Jeff Bezos — House Judiciary Antitrust Subcommittee Hearing, July 29, 2020The reformulation was significant. Bezos acknowledged the existence of a policy while declining to defend its enforcement. This created space to characterize documented seller data usage as individual violations rather than systematic practice—but only if the violations were genuinely isolated. The evidence suggested otherwise.
Representative David Cicilline, the subcommittee chairman, pressed further: "I want to understand whether Amazon was being truthful in the testimony when the General Counsel said it was not using third-party seller data. That's what he told us." Cicilline presented internal documents showing retail employees had routine access to seller dashboards and regularly consulted them when making private label decisions.
The House Judiciary Committee's final report, released October 6, 2020, framed Amazon's conduct within a broader structural analysis. The 449-page investigation concluded that Amazon possessed monopoly power over third-party sellers and that its dual role created "inherent conflicts of interest" that could not be resolved through policy controls alone.
The report documented multiple ways Amazon's platform advantages extended beyond data access. Amazon Basics products appeared prominently in search results—analysis found 16% appeared as the first organic search result for generic product category searches. Third-party sellers, by contrast, competed for position through a combination of pricing, fulfillment speed, and advertising spend paid to Amazon.
Economists call this vertical integration—when a company controls both the infrastructure (the marketplace) and competes on that infrastructure (as a seller). The classic concern is that the integrated firm has both the incentive and ability to disadvantage competitors. Amazon's case presented a particularly pure example because the digital nature of the platform made competitive advantages (search ranking, data access, algorithmic recommendation) less visible than they would be in physical retail.
While Congressional investigators lacked enforcement authority, the European Commission did not. On November 10, 2020—one month after the House Judiciary report—the Commission issued formal antitrust charges against Amazon, alleging systematic use of non-public seller data to benefit Amazon's retail business when competing with marketplace sellers.
The European investigation focused on Amazon's practices in Germany and France, but the Commission's findings described a global system. Amazon employees had access to granular, real-time data on competitors' transactions, customer visits, revenues, and shipped products. According to the Commission's Statement of Objections, this data directly informed decisions about which products to launch, how to price them, and which categories to prioritize.
Commissioner Margrethe Vestager's statement captured the core concern: "Data on the activity of third party sellers should not be used to the benefit of Amazon when it acts as a competitor to these sellers." The violation wasn't merely that Amazon accessed the data—it was that such access was structurally inevitable given Amazon's dual role, making policy controls insufficient protection.
In December 2022, Amazon settled the EU charges by agreeing not to use non-public marketplace seller data for retail decisions for five years and to treat all sellers equally in the algorithm determining which seller wins the "Buy Box"—the default purchase button that drives 82% of Amazon sales. The settlement imposed no fines but made Amazon's commitments legally binding with penalties up to 10% of global revenue for violations.
Amazon's Bangalore office emerged as a focal point in multiple investigations. The India-based team included data analysts and product managers explicitly tasked with identifying opportunities for Amazon's private label brands. Former employees consistently described having direct access to seller dashboards showing individual product performance.
One former manager told House investigators the team maintained a database of "promising products" with sales data traced to specific sellers. The manager described monthly meetings where team leads presented ranked lists of replication opportunities based on margin analysis derived from seller transaction data. When Amazon executives testified that such use violated policy, multiple former India team members expressed confusion—accessing seller data was explicitly part of their job responsibilities.
The geographic distribution of private label analysis created an enforcement gap. Policies issued from Seattle headquarters faced practical limitations when implementation occurred across global offices with different compliance cultures and management structures. Internal communications showed India team members sharing seller-specific reports with US-based retail decision-makers without apparent concern about policy violations.
Amazon Basics, launched in 2009, became the most visible manifestation of Amazon's private label strategy. Beginning with commodity electronics accessories—HDMI cables, batteries, phone chargers—the brand expanded to over 1,500 products across dozens of categories by 2019.
Analysis of Amazon Basics product launches showed systematic patterns. Products frequently appeared in categories where third-party sellers had demonstrated strong demand. Pricing typically matched or slightly undercut the category leader. Product features often mirrored successful third-party offerings with minor variations—enough to avoid design patent issues but insufficient to suggest independent product development.
Third-party sellers reported a consistent experience: a product would gain traction over months or years of marketing investment and inventory risk, demonstrate strong sales, then face Amazon Basics competition that appeared to replicate the successful formula while leveraging platform advantages in search placement and logistics. Many sellers found themselves competing with Amazon for sales of the same product—but only Amazon had access to complete market intelligence.
The competitive disadvantage extended beyond information. Amazon Basics products qualified for free shipping at lower thresholds, could be priced at breakeven or loss since Amazon profited from overall platform fees, and appeared in recommendation algorithms that third-party sellers could not influence. One economic analysis found that appearance of Amazon Basics competition in a category correlated with 6-8% sales decline for affected third-party sellers within six months.
Throughout investigations, Amazon maintained it had clear policies prohibiting use of individual seller data for competitive purposes. The company pointed to training materials, compliance protocols, and audit systems designed to prevent violations. Amazon executives characterized documented cases as individual employees acting contrary to policy rather than systematic practice.
This defense faced two fundamental challenges. First, the volume and geographic distribution of documented violations suggested more than isolated misconduct. House investigators found over 1,000 instances in internal records where Amazon employees accessed individual seller reports in contexts connected to private label development. The India team's practices were known to US management. Reports generated by Project Nessie circulated widely within retail divisions.
Second, the technical architecture made violations structurally easy. Amazon employees with retail responsibilities had database access to seller performance metrics. No technical controls prevented queries filtering by individual seller. Audit systems logged access but did not prevent it. In this environment, policy compliance depended entirely on individual restraint—a fragile protection given the competitive value of the intelligence.
"I think what we've seen is that what Amazon has said publicly and what Amazon does privately seem to be two different things."
Representative Pramila Jayapal — House Judiciary Hearing, July 29, 2020Former employees told investigators that pressure to identify winning products for private label development was intense. Performance reviews evaluated product managers partly on the commercial success of launches. In this context, having ready access to complete marketplace intelligence created inevitable temptation. Policy prohibitions without technical enforcement became aspirational rather than operational.
Beyond Amazon Basics, Amazon operated at least 45 private label brands by 2019. Many were not obviously identified as Amazon-owned: Pinzon (home goods), Presto (paper products), Solimo (groceries), Mama Bear (baby products), Wag (pet food), Stone & Beam (furniture), Rivet (home decor), Goodthreads (apparel), and dozens more.
The strategy was deliberately diversified. Amazon Basics competed on price and utility in commodity categories. Premium brands like Stone & Beam targeted higher-margin home goods segments. Fashion brands like Goodthreads and Amazon Essentials addressed categories where brand identity mattered. The collective footprint represented over 158,000 products—a portfolio no traditional retailer's private label approached in scale or scope.
Research firm One Click Retail estimated Amazon's fashion private labels alone generated over $450 million in sales in 2018, despite launching only two years earlier. This growth rate—achieved in competitive categories where brand matters—suggested advantages beyond typical private label dynamics. Traditional retail private labels succeed through price and placement, but rarely achieve rapid market share in brand-sensitive categories without substantial marketing investment Amazon did not make.
Third-party sellers faced an asymmetric power relationship. For many, Amazon represented 50-90% of online sales. Exiting the platform meant abandoning the majority of their business. Staying meant operating under terms that gave Amazon complete visibility into their performance while potentially training their future competitor.
The Amazon Sellers Association, formed in 2020, documented hundreds of cases where private label launches appeared to replicate successful third-party products. Sellers described a consistent pattern: invest months or years building a product's market presence, demonstrate strong sales, then face Amazon Basics competition leveraging platform advantages. Many sellers had no alternative—the concentration of e-commerce on Amazon's platform meant exit required abandoning online retail entirely.
This dynamic is what economists call "platform dependency." When a single platform controls access to customers, platform participants lose bargaining power. Amazon's terms of service gave the company broad rights to use marketplace data and prohibited sellers from taking legal action except through individual arbitration. Sellers who complained publicly about unfair treatment risked account suspension—a business death sentence for Amazon-dependent companies.
The House Judiciary Committee's report recommended Congress consider structural separations preventing platforms from competing on their own marketplaces. Representative Cicilline introduced legislation in 2021 that would prohibit Amazon from selling private label products while operating a marketplace where third parties compete—forcing a choice between being infrastructure or competitor, not both.
Amazon argued structural separation would harm consumers by eliminating popular private label products and reducing retail competition. The company commissioned economic studies claiming Amazon Basics products saved consumers money and that preventing platform operators from selling their own products would reduce innovation.
The debate centered on whether policy controls could adequately address conflicts of interest inherent in vertical integration. Amazon's European settlement demonstrated one approach: legal commitments not to use seller data, enforced through regulatory monitoring and potential fines. The House approach suggested policy controls were insufficient—that the temptation and capability to advantage Amazon's retail business were too strong when both businesses shared infrastructure and management.
As of 2024, no structural remedy had been imposed. The FTC investigation remained open. State attorneys general pursued cases focused on specific practices. The fundamental structure—Amazon as both marketplace and competitor—remained unchanged.
The documentary evidence that emerged through Congressional investigation, journalistic reporting, and European regulatory proceedings established several facts:
Amazon employees routinely accessed individual seller sales data to inform private label development. This occurred across multiple divisions and geographies, including prominently in the India-based product development team. The practice was documented in internal emails, product reports, and employee accounts.
Amazon maintained official policies prohibiting use of individual seller data for competitive purposes, but these policies lacked technical enforcement mechanisms. Access to seller data was logged but not prevented. Compliance depended on individual restraint in an environment where commercial pressure rewarded private label success.
Amazon executives testified to Congress that the company did not use individual seller data to compete with sellers. This testimony was contradicted by internal documents and employee accounts showing such usage was systematic. When confronted, Amazon retreated to acknowledging policy violations while characterizing them as not representative of company practice.
Project Nessie and other internal systems automated the process of identifying private label opportunities using marketplace data. While Amazon characterized these systems as using only aggregated data, the reports they generated included seller-specific intelligence and drove decisions about which products to replicate.
The European Commission found Amazon's practices violated EU competition law and extracted legally binding commitments not to use non-public seller data for retail decisions. These commitments acknowledged that Amazon's previous practices had been improper and that structural safeguards beyond policy statements were necessary.
The evidence established that Amazon's dual role as marketplace operator and competitor created a conflict of interest that policy controls failed to adequately manage. Whether structural remedies are warranted remains a legal and policy question. That the conflict exists and that Amazon's management of it failed to prevent systematic use of seller data for competitive advantage is documented fact.