Modes of control in international digital commerce: evidence from Amazon.com

Online marketplaces such as Amazon.com represent a new channel through which multinational enterprises (MNEs) can sell their products in foreign markets, either as third-party sellers or as suppliers to the platform owner. An MNE can have better control of the marketing mix when selling directly on Amazon.com as a third-party seller, but this task entails two challenges. First is the liability of foreignness, and second is the disadvantage of competing with other products that are directly sold by the platform owner. The platform owner sets the rules of the platform and has data and algorithmic advantages, putting competitive pressure on MNEs with tighter control. Data obtained from Amazon are analyzed and reveal that maintaining control as a third-party seller predicts lower sales than being a supplier to the platform owner. However, the penalty associated with retaining control is smaller for MNEs with more host-country experience and country-of-origin advantage. These findings provide new insights into how MNEs leverage platforms such as Amazon.com to expand their global reach.


Introduction
The development of international digital commerce provides multinational enterprises (MNEs) with a new channel to access global markets.As a major force in digital commerce, Amazon stands out with a distinctive business model compared to other online marketplaces.Amazon not only serves as an intermediary facilitating transactions between consumers and sellers but is also a retailer in its own right.As the platform owner, Amazon controls critical resources such as data and information, giving it substantial power and raising concerns regarding unfair competition with third-party sellers (Etro, 2022;Hagiu et al., 2022;Li et al., 2019;Zhu & Liu, 2018).Despite the disruptive influence of Amazon's business model on international business, extant research provides limited insight or empirical evidence on how MNEs can best manage their channels on such online marketplaces in a host country (Ipsmiller et al., 2022;Samiee, 2020).
Jiatao Li and Ziyi Chen are joint first authors.
Accepted by V. Kumar, Guest Editor, 18 June 2024.This article has been with the authors for six revisions.
Our study zeroes in on Amazon to explore how MNEs navigate selling on online marketplaces where the platform owner directly competes with third-party sellers on the same platform.An MNE can sell in a foreign market through Amazon in two ways.First, they can adopt a high-control mode by operating as a third-party seller,1 as exemplified by Caudilie, a French skincare company.This approach provides good control of the marketing mix because the MNE can choose the level of advertising and other promotions, set the price, and autonomously arrange the logistics of delivery.The firm can be responsive to consumers' needs while exploiting its firm-specific advantages.Alternatively, an MNE can supply products to Amazon for resale to consumers on the platform.For example, L'Occitane adopts this mode.Here, the MNE has less control because Amazon takes charge of logistics and makes decisions related to pricing, advertising, and other marketing initiatives.The MNE can obtain only limited information concerning consumer preferences and other feedback.
The conventional wisdom in international business research is that retaining greater control of distribution leads to better performance (Giachetti et al., 2019).However, we argue that a foreign firm attempting to maintain tight control on Amazon faces two challenges which can impair sales performance.First is the liability of foreignness in a host country (Zaheer, 1995).An MNE will lack knowledge concerning the preferences and purchasing behavior of foreign customers when operating as a third-party seller.The second challenge involves competing against the platform owner (Zhu & Liu, 2018).MNEs face a significant disadvantage because Amazon, as the platform owner, holds most of the consumer-related data, sets out the rules of the platform, and has technological advantages to promote products under its own name.These factors are critical to digital success but often inaccessible to third-party sellers.
Host-country experience facilitates overcoming the liability of foreignness (Lu et al., 2022).We examine the effectiveness of such experience in mitigating the negative impact of adopting a high-control mode on Amazon.An MNE's country of origin is another potential resource (Cuervo-Cazurra & Un, 2023).Based on insights from international marketing (Magnusson et al., 2019;Samiee et al., 2024), we propose that MNEs with greater country-of-origin advantage in their products can attract and lock in consumers to the platform, improving the standing of MNEs with a high-control mode.Accordingly, these ideas are tested in this study using data from Amazon.com.
We contribute to international business scholarship by extending research on control modes to digital commerce.Retaining control of distribution has its benefits (Giachetti et al., 2019), but on exchange platforms such as Amazon, it exposes a firm to direct competition from the platform owner.This study is among the first to have examined control decisions on this new online channel, and its findings suggest methods of overcoming the difficulties involved.Our findings also contribute to research on the global digital economy, particularly within Amazon's online marketplaces.Prior research has focused on the internationalization of digital platforms (Li et al., 2019).This study takes the perspective of MNEs leveraging such platforms and provides insights into how MNEs can address the challenges in international online marketplaces (Samiee, 2020).

Theory and hypotheses
MNEs' modes of control on Amazon.comAmazon's online marketplace has become an important channel for international sales, offering detailed product information, purchasing convenience, and transaction transparency.However, this business model creates growing interdependence between the platform's owner and third-party sellers (Cutolo & Kenney, 2021).As the platform owner, Amazon can establish the rules and leverage data storage and analysis capabilities to influence seller and consumer behavior, increasing seller dependence and strengthening its power.In addition, digital platforms exhibit network effects, where the benefits to sellers or consumers using a platform increase with the size and quality of other sellers (Rochet & Tirole, 2006).Therefore, platform owners rely on sellers to bring and lock in users, motivating them to incentivize sellers of high-quality products.
An MNE attempting to expand its international sales through Amazon.comfaces a decision regarding control.The firm can adopt a high-control mode by selling directly as a third-party seller.This mode allows the firm to independently set prices, offer discounts, manage brand presence, oversee product listings, and navigate various operation aspects such as delivery and inventory level.This autonomy enables the firm to implement its own marketing strategy and develop its digital commerce capabilities.Alternatively, it can opt for a low-control mode by contracting sales to Amazon, which then resells the products to consumers.In this arrangement, Amazon controls pricing, product listings, and logistics.The firm serves the role of being Amazon's supplier, fulfilling purchase orders and shipping inventory to Amazon's warehouses.Although this process reduces the complexities of managing sales on the platform, it could limit the firm's opportunity to develop its digital commerce expertise.Table 1 details the specifics of these two control modes on Amazon, and Fig. 1 illustrates examples of each mode.
An MNE operating as a direct seller on Amazon suffers from greater liability of foreignness than if it contracted sales to the platform's owner.As a foreign actor, an MNE may not immediately understand local consumers' preferences regarding digital purchases or the sales tactics commonly used on the platform.It will also struggle to stand out as an independent seller if the platform hosts many firms offering rival products.Moreover, website traffic and consumer data are typically controlled by the platform owner.Without Amazon's support, a foreign firm will have difficulty accessing local resources and consumers through digital marketing.Thus, contracting distribution to the platform owner can minimize risks and uncertainty in a host-country online marketplace.
In addition, a high-control mode can expose a firm to direct competition from the platform owner.Such competition pressure is particularly salient when the platform acts both as an intermediary for third-party sellers and as a reseller selling directly to consumers (Hoffmann et al., 2018;Zhang et al., 2022;Zhu & Liu, 2018).Self-preferencing by the platform owner represents a major challenge.The platform owner controls the search engine, and its sorting algorithms tend to favor products with which it has a close relationship (Zennyo, 2022).By contrast, thirdparty sellers incur additional costs for advertising to gain favorable search placement on the platform.Moreover, the platform owner holds extensive data on consumer behavior and offers purchase recommendations that direct customers to products sold by the platform's owner (Chen & Tsai, 2023).Platform owners may also monitor third-party sellers' performance and can choose to offer similar products, thus putting competitive pressure on these other sellers (Zhu & Liu, 2018).Lastly, online information asymmetry and consumers' decision biases enable the platform owner to adjust platform rules in its favor (Jiang & Zou, 2020).Consumers frequently use specific signals when selecting products, such as the sellers' reputation (Reuber & Fischer, 2009).They also tend to prioritize information that appears early in the results (Narayanan & Kalyanam, 2015), allowing the platform owner to influence product selection through presentation order.Thus, contracting sales to the platform owner increases the likelihood that potential consumers will view and purchase a firm's products.

Hypothesis 1 (H1)
There is a negative relationship between the degree of control an MNE retains on a host-country online marketplace and its sales performance.

Moderating effect of host-country experience and country-of-origin advantage
Given the challenges faced by foreign sellers on Amazon, we examine what resources an MNE can leverage to mitigate the negative impact of retaining high control on the platform.Previous research has demonstrated that host-country experience helps reduce the liability of foreignness (Lu et al., 2022).First, host-country experience allows a foreign firm to better understand local customers.A lack of local knowledge represents a major challenge for foreign firms independently managing the marketing mix on a platform (Gaur & Lu, 2007).Prior interactions with host-market customers provide the firms with a nuanced understanding of their preferences, which can be applied to digital commerce.Moreover, a foreign firm selling in a high-control mode can gain easier access to local customers when it has more experience in the host country.Downstream capabilities are often developed through extensive local experience because of the tacit Table 1 MNEs' modes of control on Amazon.comLow-control mode (MNE as a supplier to Amazon) High-control mode (MNE as a third-party seller on Amazon)

Pricing
Amazon sets the price and discount on its platform based on internal algorithms.
The third-party seller keeps control of retail price and determines discounts on Amazon.com.Marketing Amazon determines product listing content and can change the descriptions without the supplier's approval.
The third-party seller has full control over its brand and the product listings, including descriptions, key features, images, and promotions, etc.Some third-party sellers can redirect consumers to their own website.Operation Amazon handles operation and delivery.The supplier only fulfils purchase orders from Amazon and ships inventory to Amazon's warehouses via portal, Vendor Central.
The third-party seller independently manages operation on the portal, Seller Central, overseeing tasks such as inventory forecast and control, product delivery, and overall portal management.

Digital commerce capability
The supplier has limited opportunities to develop its digital expertise.
The third-party seller can implement its digital commerce strategies with greater autonomy and develop its digital commerce capabilities.
knowledge and social complexity involved.Greater local experience leads to higher customer trust even without local distributor affiliation.The established customer base in the host market can then become an important source of revenue for firms operating their own online channels.Together, these factors make tight control more feasible for a firm with more host-country experience.
Hypothesis 2 (H2) The negative effect of using a high-control mode on a host-country online marketplace is weaker for MNEs with greater host-country experience.
Apart from host-country experience, a product's country of origin can be another useful resource in competing within a host-country online marketplace (Cuervo-Cazurra & Un, 2023).International marketing research has demonstrated that a product's country of origin, often highlighted by the "Made in" label, has a substantial influence on consumer evaluations and sales abroad (Samiee et al., 2024).Certain products from particular countries are viewed favorably owing to their superior quality, advanced technology, rich historical tradition, or strong cultural authenticity (Magnusson et al., 2019;Verlegh & Steenkamp, 1999).Examples include German cars, French wine, Italian fashion goods, and Swiss mechanical watches, among others.Products from these countries benefit from a thriving home-country industry environment, providing their MNEs with a distinct country-of-origin advantage over international rivals.
We argue that MNEs with a country-of-origin advantage in their industry can mitigate the negative effect of using the high-control mode.First, such advantages are considered valuable assets by platforms, as they enhance consumer perceptions regarding the variety and quality of products on the platform.This advantage, in turn, gives MNEs leverage over the platform owner.In fact, the platform owner may promote them in the marketplace to enhance the platform's own standing (Rietveld et al., 2019;Taeuscher & Rothe, 2021).Second, MNEs from countries renowned for specific superior products have a deep understanding of the capabilities and cultural nuances in their national contexts.This insight enables the MNEs to manage their branding and marketing more effectively when retaining greater control on the platform.For example, marketers can craft unique and engaging narratives using multimedia tools tailored to their brands.In this case, a product's foreignness becomes an asset that enhances its appeal and attracts consumer purchases.
Hypothesis 3 (H3) The negative effect of using a high-control mode on a host-country online marketplace is weaker for MNEs with a greater country-of-origin advantage.

Data and sample
We focused this study on the beauty and personal care category because it is the second most popular category on Amazon.com and features a substantial number of foreign brands (Jungle Scout, 2023).To construct the sample, we started with the 2418 luxury beauty and personal care products that had received at least one review on Amazon's U.S. platform between January 2017 and August 2018 (Ni et al., 2019).Starting from August 2018, Amazon implemented a brand gating policy in this category to attract premium beauty brands.Under this policy, if a beauty brand's owner collaborated with Amazon, only the brand owner or its authorized distributor could sell the brand's products.Other sellers' offers for that brand were removed from the platform.We focused on the 1665 (or roughly 69%) products that collaborated with Amazon, as only the brand owners of these products faced control mode decisions.We then used the U.S. Patent and Trademark Office's trademark database to identify each brand's country of origin and earliest registration date.Brands without a registered trademark in a beautyrelated product category were excluded, which left a total of 1555 products.We focused entirely on products from outside the U.S., which accounted for 30.8% of the sample.We also collected information on the products' sales ranks, ratings, prices, and other variables up to 2021 using an application programming interface from Keepa.com, a commercial data provider that tracks sellers and products on Amazon on a near-real-time basis.After eliminating observations with missing values, the final sample consisted of 479 products with 7460 product-year-month observations.

Variables
The dependent variable was sales performance, which was quantified as the percentile rank of a product's sales volume (in units) within the population of all products on Amazon.com in a given month.2A higher percentile rank indicates better sales performance.
The predictor MNEs' high-control mode was a binary variable that takes a value of 1 if the MNE brand owner or authorized distributor acted as the seller on Amazon.com, and 0 if Amazon was the seller.An MNE can still maintain good control over its products by using a non-Amazon distributor, which is smaller than Amazon and works closely with the MNE.
Host-country experience was measured as the logarithm of the number of years that the firm's trademark had been registered in the U.S.This information was also collected from the USPTO Trademark database.
Country-of-origin advantage was a binary variable that takes a value of 1 for French and Italian brands, and 0 otherwise.In the luxury fashion and beauty industry, France and Italy are historically renowned for their elegance, specialized craftmanship, and unique cultural identity (Guercini & Milanesi, 2017).The prestige and cultural capital of these countries shape consumer perceptions, associating French and Italian brands with "aesthetic sensitivity, refined taste, and sensory pleasure," qualities highly valued in the beauty sector (Leclerc et al., 1994: 264;Melnyk et al., 2012).This perception grants French and Italian brands a distinctive country-of-origin advantage.Our analysis of the 100 most valuable global brands reveals that among those offering beauty products, nine out of the 11 originate from France and Italy (Interbrand, 2023), demonstrating their strong market presence.This binary measure does not imply that only French or Italian brands have country-of-origin advantage.Instead, it captures the relatively more favorable perception of brands from these countries in the beauty sector than those from other countries.
Several other variables were included in the analyses.One was the logarithm of the price, which should be negatively related to demand and has been commonly used as a control when studying sales performance (e.g., Bahadir et al., 2015).Another factor was product quality, represented by rating scores and the logarithm of the number of consumer reviews the product had received (Zhu & Liu, 2018).Amazon possibly favors trendy products, and consumers prefer newly listed ones too.Therefore, we controlled for product age using the logarithm of the number of months since its listing on Amazon.com.We also included regional fixed effects by creating a series of dummy variables to indicate each firm's continent of origin (e.g., North America, Europe, and Asia).We therefore considered customers' preferences for products from different geographic regions.Finally, category and year-month fixed effects were included to account for differences in product subcategories (e.g., fragrance, skin care, hair care, etc.) and changes in the macroeconomic environment.

Analyses
Firms make control decisions on the platform based on their attributes and environmental conditions, thereby rendering such strategy choices endogenous (Shaver, 1998).To address this concern, we used two-stage least squares (2SLS) regressions (Wooldridge, 2010).The instrumental variable in the first-stage model was weighted peers' choice of the highcontrol mode (WPHC) (Lim et al., 2020).Products from the same home country were considered peers (Li et al., 2007).The weight on each peer product was defined on the basis of its similarity to the focal product, with more similar products receiving greater weight.This instrument satisfies the relevance criterion because a firm's control decisions can be influenced by its peers' choices due to uncertainty and normative pressure.Meanwhile, no clear correlation exists between the peers' decisions and the unobservable factors in the error term that might affect the focal product's future performance (Wolfolds & Siegel, 2019).Further details on the construction of weights and validity of this instrument are discussed in the Online Appendix.

Results
Table 2 presents the summary statistics and pairwise correlations for the variables.The variance inflation factors of the variables ranged from 1.07 and 1.42, well below 5, the threshold beyond which multicollinearity concerns may arise (Ryan, 2008).
Table 3 reports the coefficients of the 2SLS regressions.Column 1 presents the first-stage estimation results and shows that the coefficient of the weighted peers' choice of the high-control mode term was positive and significant at the 0.1% level (b = 0.629, p = 0.000).An MNE was more likely to apply a high-control mode on Amazon.com if more of its peers also chose that mode.Column 2 shows the second-stage result from the instrumental variable regression.Consistent with H1, the coefficient of the MNEs' high-control mode term was negative and significant at the 0.1% level (b = − 0.101, p = 0.000), indicating that this control mode impairs sales performance.
We then investigated the moderating effect of hostcountry experience and country-of-origin advantage.Column 3 of Table 3 shows that consistent with H2, the interaction between MNEs' high-control mode and hostcountry experience was positive and significant at the 0.1% level (b = 0.073, p = 0.000).Adopting the highcontrol mode on Amazon was less of an issue for firms with more experience in the U.S. market.Furthermore, consistent with H3, Column 4 shows that the interaction between MNEs' high-control mode and country-oforigin advantage was positive and significant at the 1% level (b = 0.116, p = 0.000).The finding suggests that the negative effect of retaining greater control on the platform is weaker for MNEs with greater country-of-origin advantage (Fig. 2).We conducted several robustness checks and additional analysis.First, we split the sample into high and low groups at the median of the host-country experience and estimated a regression for each group.The negative effect of MNEs' high-control mode was 1.6 times stronger for MNEs with low host-country experience, supporting H2 (low experience: b = − 0.201, p = 0.000 vs. high experience: b = − 0.076, p = 0.000; the difference of the two coefficients is significant at the 0.1% level, p = 0.000).Similarly, the effect of using a high-control mode was only negative for products with less country-of-origin advantage, supporting H3 (less country-of-origin advantage: b = − 0.204, p = 0.000 vs. stronger country-of-origin advantage: b = 0.054, p = 0.004; the difference of the two coefficients is significant at the 0.1% level, p = 0.000).Second, to test the robustness of our country-of-origin advantage measure, we used only French brands in defining this variable, given France's distinct cultural traits (Leclerc et al., 1994).The results remained consistent (b = 0.148, p = 0.000).Third, an alternative explanation for our finding could be the presence of MNEs' other sales channel outside Amazon.com.For example, MNEs with alternative channels in a host market might choose a high-control mode on Amazon.com to complement their existing sales strategies.We collected data on whether the brands in our study also sold directly through their own official websites and included them as a control variable. 3oughly 80% brands had such alternative online channels.With this additional control, the effect of MNEs' high-control mode remained negative and significant (b = − 0.110, p = 0.000).The effect was also consistent across samples with or without sales on the brands' own websites (with: b = − 0.089, p = 0.000; without: b = − 0.227, p = 0.000).competitive landscape and provide a framework for using such platforms effectively.While many MNEs already distribute their products and have gained local experience in host countries, Amazon's digital prowess compels MNEs to reevaluate their experience and enhance their brand positioning in this online marketplace.We also extend the scholarly understanding of the global digital economy.Research into global online platforms has focused on how platform owners coordinate participants' activities and build ecosystem-specific advantages in a host country (e.g., Brouthers et al., 2016;Li et al., 2019).Less is known on how traditional MNEs can better engage in international digital commerce.The limited research is generally conceptual or focuses on B2B markets, overlooking B2C markets which include many structural and consumer hurdles (Samiee, 2020).Focusing on Amazon's marketplace, our findings show that MNEs can address the power imbalance by leveraging their host-and home-country advantages, providing implications to MNEs to operate on similar online marketplaces.
Lastly, our study contributes to the research on the relationship between Amazon as a platform owner and its sellers (e.g., Wang & Miller, 2020).It aligns with previous studies in highlighting that the degree of control is a crucial strategic decision for managing competition and cooperation within platform ecosystems (Altman et al., 2022;Hoffmann et al., 2018).Prior research has explored how platform owners leverage seller-level heterogeneity, such as product quality and competitive positioning, to design governance mechanisms that maximize total value creation (Agarwal et al., 2023;Rietveld et al., 2019).Our study suggests that sellers' experience and country-of-origin advantage are also important factors for Amazon when managing relationships with its sellers in international marketplaces.

Limitations and future research
This study focuses on Amazon's platform, and its findings may not be generalizable to digital commerce platforms with a different business model.Our measure of country-of-origin advantage is simplified and only captures one aspect of MNEs' strengths.This measure is tailored specifically for the beauty sector and may not apply to other industries.Future researchers could devise more refined measures to capture MNEs' other advantages relative to the platform.In addition, we used sales performance rather than profitability as the dependent variable given the lack of information on profitability.Contracting sales to Amazon could garner larger sales volume but smaller profits due to weaker bargaining power relative to the platform's owner.By collecting more detailed financial data and investigating other online marketplaces, future studies could offer deeper insights of the tradeoff between value creation and capture in international digital commerce.
The proliferation of digital channels allows an MNE to adopt multi-channel strategies in host markets.An MNE might simultaneously launch a localized direct-sales website, distribute products through various online marketplaces, and sell offline in its own or partnered retail stores.Our study focuses on Amazon.com,examining the implications of different control modes within this platform.Experience with other channels can influence control mode choice on Amazon.For example, direct sales through the brand's own website might enhance digital commerce expertise, thereby increasing the likelihood of a high-control mode on Amazon.Alternatively, a firm might select a control mode that complements its existing channel strategies.Moreover, access to alternative channels can decrease an MNE's dependence on Amazon, mitigating concerns regarding the adverse effects of retaining high control over their Amazon operations.These factors could bias the relationship between MNEs' control mode on Amazon and their sales performance.Although we controlled for direct sales through the brand's website, this step does not account for all possible channels, consequently limiting our understanding of control modes in broader contexts.Future research could conduct a more comprehensive analysis to explore how MNEs select control modes across various channels and how strategies in one channel affect the choice and performance on other channels.
The change in control modes on a digital platform over time is also worthy of further research.Our sample contains no instances of control mode changes owing to the short duration.In traditional contexts, some MNEs enter a foreign market with low control and shift toward greater control as they gain experience (Chang, 2019).Scholars can examine the extent to which this scenario applies in the digital commerce context.Firms switching from a low-control to a high-control mode could risk losing support and damaging existing collaborative ties with Amazon, unlike those making the opposite switch.Even if an MNE has gained online experience in the host market, it may not have full access to relevant consumer data from Amazon and thus remains at a disadvantage.The shift of control modes in digital commerce and the dynamic capabilities that enable MNEs to better compete with the platform owner present an intriguing topic for future research.

Fig. 1
Fig. 1 Examples of control modes on Amazon.com.A High-control mode.B Lowcontrol mode

Table 2
Summary statistics