Authors: Dr. Gönenç Gürkaynak [1], Harun Gündüz, Ceren Özkanlı Samlı, Göksu Kıribrahim, Gaye Vuslat Üstün
I. Introduction
The Turkish Competition Board (“Board”) concluded the investigation into the pricing algorithm practices of multi-category e-marketplaces and determined that the pricing algorithms used on the e-marketplace platform of Amazon Turkey Perakende Hizmetleri Limited Şirketi (“Amazon Turkiye”) did not violate Article 4 of Law No. 4054 on the Protection of Competition (“Law No. 4054”). [2] While the Board ultimately found no infringement, Amazon Turkiye nevertheless opted to discontinue the relevant tool, committing to its complete shutdown by April 11, 2025.
The investigation focused on whether the automatic pricing algorithms available to sellers on Amazon Turkiye’s platform infringed Article 4 of Law No. 4054 by facilitating a hub-and-spoke cartel, anti-competitive agreements or the exchange of competitively sensitive information between the sellers. Following its assessment of both the technical functioning of the mechanism and its market effects, while acknowledging that pricing algorithms may raise significant competition law concerns, the Board found that the automatic pricing tool in question is not mandatory to use, it operates in line with the parameters set by the sellers (i.e., it is rule-based, rather than learning-based) and therefore, it did not, at this stage, possess the characteristics or function necessary to be considered restrictive of competition. The Decision sheds light on the Board’s approach to theories of harm about pricing algorithms within the design and functionality nexus, where it prioritizes whether an algorithm is rule-based rather than learning-based, and whether it lacks reciprocal dynamic learning that could facilitate coordinated anti-competitive effects.
II. Background
The Turkish Competition Authority’s (“TCA”) scrutiny began with a preliminary investigation into whether D-Market Elektronik Hizmetler ve Ticaret A.Ş. (“Hepsiburada”) had infringed Law No. 4054 in relation to two key concerns: (i) alleged discriminatory practices and Most Favoured Customer (“MFC”) clauses, and (ii) the automatic pricing mechanism. During the preliminary investigation, the TCA identified that DSM Grup Danışmanlık İletişim ve Satış Ticaret AŞ (“Trendyol”) and Amazon Turkiye also offered automatic pricing mechanisms.
Accordingly, the Board expanded its investigation to include Hepsiburada, Trendyol, and Amazon Turkiye in relation to pricing algorithms, while discontinuing its inquiry into the MFC clauses.
During the investigation, Hepsiburada, Trendyol, and Amazon Turkiye initiated commitment procedures to address the TCA’s concerns. While Hepsiburada and Trendyol concluded the investigation through commitments on their part, Amazon Turkiye terminated the commitment procedure by maintaining that (i) the automatic pricing mechanism is optional and designed to enhance competition for the benefit of both customers and third-party sellers, (ii) it does not give rise to the alleged competition concerns, however (iii) it would nevertheless discontinue the mechanism in Turkiye.
III. The Board’s Approach to Theories of Harm Concerning Pricing Algorithms
According to the Decision, undertakings increasingly leverage algorithms to refine pricing models, personalized services, and forecast market trends in the digital economy. These tools serve diverse functions, including data collection, consumer-based personalization, and dynamic pricing.
Algorithms can generate pro-competitive effects by improving efficiency (e.g., fostering disruptive innovation that results in new or improved products) and by reducing transaction costs (e.g., through optimized production processes or increased worker productivity) on both the supply and demand sides. They can also lower consumer search costs and facilitate supply-demand balancing through dynamic pricing mechanisms that provide a suitable range of products and comparable information on key competitive parameters such as price, quality, and consumer preferences.
However, despite the benefits, the Decision highlights that pricing algorithms may raise significant competition concerns. By automating the decision-making process, pricing algorithms can increase market transparency, making it easier for competitors to monitor each other, align on strategic factors such as pricing, detect deviations from any implicit coordination, and ultimately respond to or discipline such deviations. Algorithms can also enable self-preferencing, predatory pricing, price discrimination, or traditional anti-competitive practices such as price-fixing or market sharing.
Algorithms may generally facilitate coordination between undertakings in three ways. First, automated pricing systems based on existing pricing data may stabilize coordination by detecting and responding to deviations (e.g., monitoring compliance with resale price maintenance or price-fixing arrangements). Second, they may enable a hub-and-spoke mechanism by facilitating information exchange (e.g., where undertakings rely on the same third-party pricing software to set prices). Third, they may lead undertakings to coordinate, or at least avoid competitive outcomes, through self-learning autonomous algorithms, even in the absence of explicit coordination or information exchange.
Against this background, the Board assessed different theories of harm scenarios centered on pricing algorithms. The Board first examined the hub-and-spoke cartel model, noting that such an arrangement requires five cumulative conditions [3] to be met, including the intentional sharing of future pricing information through a common provider and the use of that information by competitors to determine their own prices. The Board also considered the risk of tacit collusion, where the awareness that competitors are using similar algorithms allows them to more effectively predict and interpret each other’s pricing behaviors.
The assessment further addressed scenarios of unintentional coordination and conscious parallelism facilitated by third-party service providers, where liability may depend on whether the parties were aware of, or could reasonably foresee, potential anti-competitive outcomes. In this context, the Board emphasized the need for a case-specific analysis, taking into account factors such as the reason for the algorithm’s application, the identity of the developer, and whether multiple competitors rely on the same third-party software.
IV. The Assessment of Amazon’s Pricing Algorithms
Amazon’s automatic pricing mechanism was designed primarily to automate how sellers compete for the Buy Box. Since multiple sellers may offer the same product on an e-marketplace, the Buy Box system groups identical offers under a single listing and highlights the seller whose offer is deemed most advantageous according to Amazon’s algorithmic criteria. Consequently, when a customer clicks “Add to Cart” or “Buy Now,” the product from the Buy Box winner is the one selected. This system allows sellers to automate price adjustments based on predefined rules to improve their chances of winning the Buy Box.
Amazon introduced the tool in April 2020 as the first platform to introduce such tools. Sellers can choose from these four options: (i) Competitive Featured Offer which aligns based on the price of the buy box product (matches, beats, or stays above the current Buy Box price), (ii) Competitive Lowest Price, which aligns based on the lowest price (matches, beats, or stays above the lowest price on the platform), (iii) External Competitive Price which allows sellers to match or stay below the lowest place observed in other channels (iv) Sales-Based Pricing which creates price changes based on sellers own sales volume rather than competitors prices. Among these tools, the competitive featured offer has a significant difference, since the Buy Box winner is determined by a dynamic and non-transparent algorithm that considers factors beyond price, such as stock availability, delivery performance, and seller reliability. Therefore, matching the Buy Box price does not guarantee winning the position, as non-price parameters play a significant role.
During the investigation, the TCA consulted the Department of Economic Analysis and Research, which assessed that the automatic pricing service in question is based on a rule-based algorithm, does not involve an in-depth analysis of competitors’ prices, and does not establish a structure based on mutual dynamic learning. It further noted that the system is not designed using machine learning; rather, it allows sellers to define parameters limited to setting prices below, above, or equal to a specified level. On this basis, the competitive concerns arising from the system were considered more limited compared to those associated with machine learningbased automated pricing.
All in all, the Board assessed whether Amazon Turkiye’s automatic pricing mechanism could give rise to concerns like a hub-and-spoke cartel, specifically where competing sellers rely on the same tool in setting their prices. That said, the Board recalled that, to mention a hub-andspoke type infringement, a number of cumulative conditions must be met (notably, the intentional sharing of strategic pricing information via the platform, its onward transmission to other sellers, awareness of its source, and reliance on such information in determining future pricing).
Further, the Board underlined that pricing algorithms are not problematic per se and may generate efficiencies and consumer benefits in certain circumstances. In this vein, the Board emphasized that the key question is whether such tools facilitate coordination in practice. To that end, the Board noted that sellers set their prices based on parameters such as the “Competitive Featured Offer,” “Competitive Lowest Price,” and “External Competitor Price,” without any direct or indirect contact between them. It also noted that the mechanism is not mandatory, that no evidence of coordination or alignment was identified, and that sellers retain discretion to define and adjust their own rule sets, including their duration. Finally, the system was found to be rule-based rather than learning-based. Considering these factors the Board concluded that, at this stage, the mechanism does not give rise to a restriction of competition under Article 4 of Law No.4054.
V. Conclusion
In conclusion, the Board determined that Amazon did not violate Article 4 of Law No. 4054 through its automatic pricing mechanism. TCA’s approach to this investigation was based on tool’s characteristics and its effects on the market. The Board took into account that the tool is based on predefined rules rather than machine learning facilities. It was a voluntary system with very low and inconsistent usage rates among sellers. Economic analysis on price movements showed dynamic pricing behaviors providing no evidence of an anti-competitive agreement or exchange of competitively sensitive strategic information which can be an indicative of a hub-and-spoke cartel.
Footnotes
[1] Attorney at Law and Founding Partner of ELIG Gürkaynak Attorneys-at-Law, Istanbul, Türkiye. Honorary Professor of Practice at University College London (UCL), Faculty of Laws and Senior Fellow at University College London, Centre for Law, Economics and Society. Member of faculty at Bilkent University, Faculty of Law, Ankara, and Bilgi University, Faculty of Law, Istanbul.
[2] Amazon Türkiye decision dated April 18, 2025, and numbered 25-15/348-164.
[3] According to Competition Law Glossary of Terms which refers to Competition Appeal Tribunal’s Tesco decision dated 20.12.2012 and numbered 1188/1/1/11.