Platform Owner Entry Into Complementor Spaces Under Different Governance Modes

Platform Governance on Digital Platforms

Extant platform literature suggests that platform governance can substantially shape the performance and behavior of complementors (Claussen, Kretschmer, & Mayrhofer, 2013; Rietveld, Schilling, & Bellavitis, 2019; Zhang et al., 2022). Platform governance refers to the extent to which a platform cedes decision rights regarding platform attributes and arrangements to complementors and other stakeholders in the platform (e.g., users; Boudreau, 2010; Chen et al., 2021; Chen et al., 2022; Karhu et al., 2018; Schilling, 2009; Tiwana, 2014). The goal of platform governance is to orchestrate and nurture the ecosystem of complementors and other stakeholders for the platform’s value creation and appropriation (Karhu et al., 2018; Wareham et al., 2014). A platform’s approach to governance is conceptualized as relatively stable, encompassing a wide array of platform-wide practices and strategies, such as content policies, resource distribution, terms of service, technical aspects—for example, application programming interfaces (APIs) and software development kits (SDKs), algorithms, and pricing schemes (Plantin, Lagoze, Edwards, & Sandvig, 2018; Tiwana, 2014). We therefore expect platform governance to be reflected in the owners’ mode of entry, when making inroads into their complementor spaces. While we consider a platform owner approach to governance as relatively stable, there might be cases in which platforms change their approach to governance over time. We discuss some of these cases in the discussion section of the paper.

A central distinction in platform governance exists between a decentralized, “open and laissez-faire” approach that devolves decision rights to platform stakeholders and, on the other hand, a centralized, “closed and reigning” approach that reserves most rights to the platform owner (Chen et al., 2021; Chen et al., 2022; Karhu et al., 2018; Tilson et al., 2010; Wareham et al., 2014). Extant studies show that platform governance can vary from a fully open and decentralized approach to one that is fully closed and centralized (Schilling, 2009; West, 2003). Both decentralization and centralization have their respective benefits and drawbacks (Chen et al., 2021; Chen et al., 2022). On the one hand, a centralized approach characterizing closed and reigning governance can empower platform owners, who retain tight control over platform access and usage, core interactions, and value distribution (Boudreau, 2010; Chen et al., 2021; Chen et al., 2022; Rietveld et al., 2019). This allows prioritizing the owner’s interests over those of other stakeholders and capturing a higher fraction of the value its platform creates (Chen et al., 2021; Zhu, 2019). The drawback is that such behavior could alienate and discourage other stakeholders on the platform, due to concerns about the effects of the power imbalance that favors the platform (Arrieta-Ibarra, Goff, Jiménez-Hernández, Lanier, & Weyl, 2018; Chen et al., 2021). On the other hand, a decentralized approach associated with open and laissez-faire governance devolves decision rights to platform participants, providing them with better opportunities to make decisions consistent with their preferences and interests (Chen et al., 2021). As complementors can make decisions that align better with their incentives, they have room to capture more of the value the platform creates. Therefore, such an open and laissez-faire platform will likely incentivize complementors to engage because of “enhanced incentive compatibility” (Chen et al., 2021: 1311). Thus, a decentralization approach can facilitate platform growth in terms of the number of complementors and users, which, however, may not be the best strategy for value appropriation from the owner’s perspective, at least in the short run.

The Competitive Dynamics of Platform Owner Entry Into Complementor Spaces

The theory of competitive dynamics (Chen & MacMillan, 1992; Hsieh & Vermeulen, 2014; Porter, 1980) provides a useful framework for analyzing the performance repercussions and responses of complementors upon entry of a platform owner into their spaces. Scholars have studied the impact a new entrant can have on the existing competitive dynamics, focusing on the association between how an entrant enters an existing market and the level of competitive pressure it exerts on the existing players, with the resulting impact on the players’ performance. This research primarily argues that the degree of competitive pressure the entrant exerts is a function of its level of resources (vis-à-vis the existing players) and the aggressiveness with which the entrant seeks to capture value (Chen & Miller, 1994; Ferrier et al., 1999). Even if the entrant does not immediately carry out aggressive actions to gain ground in the market, an entrant’s vast resources can influence the behavior of the existing players, who understand that the new entrant has the potential to exert competitive pressure at any moment (Karakaya & Yannopoulos, 2011). An entrant’s high level of aggressiveness on entry sends an unmistakable signal to the existing players that the new entrant came ready to put up a fight and conquer. Entry aggressiveness relates to a series of actions by the entrant designed to increase its product’s advantage and visibility, such as promotions, advertising, and price cutting. Ultimately, these actions aim to maximize the new entrant’s market share and capture value by luring customers away from the existing players in the market (Ferrier et al., 1999).

Competitive dynamics theory was formulated largely during the 1980s and 1990s before management research focused on the peculiarities of platform markets. However, today, these markets occupy a prominent place in management research and practitioner interests. Our existing understanding of competitive dynamics thus requires important adjustments in the context of entry into platform markets—particularly, the increasingly common situation of a platform owner entering its own complementor spaces. Such cases represent significant departures from the situations and studies that gave rise to competitive dynamics theory, potentially leading to different outcomes and, thus, calling for this extension of the theory. As we detail below, the entry of a platform owner into its complementor spaces is not simply the entry of a new competitor with vast resources into a market that other firms already inhabit.

Digital platforms present unique characteristics that may exhibit different market dynamics from those in traditional industries. A platform owner tends to be larger and more resource-rich than most complementors, and the level of power and influence that the owner exerts over its complementor markets is far greater than that of a typical new market entrant (Boudreau & Hagiu, 2009; Claussen et al., 2013). The owner’s power on a digital platform has been compared to the “bouncer’s right” (Strahilovetz, 2006). In traditional settings, a new entrant can influence the behavior of the existing players mainly through adjusting prices or promoting its product—for instance, it can lower its price, which will likely produce some reaction from the incumbents. Therefore, the effect is indirect; the existing players would react to the entrant’s success in attracting customers with its price-setting and promotions. However, when a platform owner enters its complementor spaces, the options do not just shrink to investments in attracting users. A platform owner also has a direct influence on complementor behavior because the owner sets the rules by which complementors must play (Boudreau & Hagiu, 2009). For instance, a platform owner may require complementors to adopt specific new technology standards or protocols when releasing its product, as often happens when mobile apps must adapt their products to the new changes introduced when the owner launches a new operating system. This change can directly shape the operations and technological priorities of complementors on digital platforms. A platform owner entering its complementor spaces can influence the existing complementors not only through its effect on users but also directly, by modifying the technologies and the set of rules that affect the operations of incumbent complementors. Thus, new possibilities open, beyond those that extant competitive dynamics theory contemplates.

In addition, on digital platforms, a close association exists among entry aggressiveness, the level of competitive pressure the owner exerts, and an owner’s approach to platform governance. We argue and expect that platforms that pursue a closed, reigning approach to governance will enter complementor spaces deploying more aggressive strategies and exerting more competitive pressure than those that pursue an open, laissez-faire approach to governance.

In their attempts to enter complementor markets, platforms with a closed and reigning approach to governance will likely tightly control usage of and access to the owner’s product and its underpinning technologies, exercising significant control and restricting how users and complementors can use the product and related technologies. For example, an owner could promote its product prominently on the landing page of the platform over those of complementors and, in addition, make technological or operational changes in the platform that require complementors to comply. Owners may also tie platform-level services and programs to the use of the owner’s product (Zhu & Liu, 2018) while constraining complementors from leveraging or integrating with the owner product or its underlying technologies. For instance, when entering its complementor spaces, Amazon has favored its own products (e.g., offering free shipping through its platform-wide service, Prime, or giving them priority and prominent space on its web pages or in search results) (Zhu & Liu, 2018). Such practices can siphon consumer attention away from the affected complementors’ product and toward the owner’s. Further, this approach can curb opportunities for complementors to benefit from the positive spillovers that result from owner entry, such as enhanced visibility and attention to that domain, which can lead to market expansion. By enacting tight controls over the usage and access to its product and technology, such owner can capture much of its entry’s newly created value, thus restricting complementors’ residual value capture. Hence, in entries carried out by platforms with a closed and reigning approach to governance, competitive forces are likely to dominate cooperative forces, intensifying the competitive pressure on complementors. Thus, such entry is likely to be negatively associated with the product performance of the affected complementors. The level of entry aggressiveness and competitive pressure on the existing firms can be, at least theoretically, far greater in the case of platform owner entry than an entry in a nonplatform setting, given that a platform owner can directly influence the rules of the game that both users and complementors must follow, in addition to having much greater resources and valuable information.

Platforms with an open and laissez-faire approach to governance are likely to give ample discretion to users and complementors in their incursions into complementor spaces, exercising limited control over how platform participants can use the owner’s product and its underlying technologies. This can beget significant possibilities for complementors to capture a part of the value that the owner’s product introduction creates. For instance, an owner could release a product in a complementor market that serves as a “hub,” to which complementors can easily connect their products, amplifying their offerings and allowing them to create new services. Generous access to the owner’s product and its underpinning technologies can allow high and customizable levels of integration that complementors can use to enhance their own products, to their advantage. This integration can benefit complementors because the owner’s product likely leverages the owner’s built-in asymmetric advantage and deeper knowledge of the platform. For instance, the owner’s product could feature a tighter and smoother interface and compatibility with new or upcoming features in the platform’s main operating system. These improvements and opportunities can increase consumer satisfaction and enhance demand for all complementor products in the affected markets, producing positive spillovers from the owner’s entry. Moreover, owners with an open and laissez-faire approach to governance would also refrain from giving their products unfair advantages over those of complementors, making siphoning user attention away from the complementor products less likely. Hence, in entries carried out by platforms with an open and laissez-faire approach to governance, the cooperative forces will likely predominate over the competitive forces, resulting in relatively low competitive pressure on the affected complementors. Such an entry is thus likely to be positively associated with the product performance of the affected complementors, unlike what happens with a new market player entering a traditional industry, in which case the effect on the existing players is bound to be negative or, at best, nil. The entry by an owner with an open and laissez-faire approach creates a new range of possible behaviors and outcomes that can affect complementors positively.

Taken together, a platform owner’s mode of entry into its complementor markets—which its approach to platform governance largely determines—will drive the competitive pressure the entry exerts on the affected complementors, tilting the balance between competitive and cooperative forces of platform owner entry and ultimately determining the valence of such entry’s impact. It then follows:

Hypothesis 1a: The entry of a platform owner with a closed and reigning approach to governance is negatively associated with the affected complementors’ product performance.

Competitive dynamics theory also informs the existing market players’ response to the entry of a new player in their market (Bengtsson & Marell, 2006; Gatignon, Anderson, & Helsen, 1989; Karakaya & Yannopoulos, 2011). This is of utmost importance in platform settings due to the symbiotic relationship between a platform owner and its complementors. Complementors play a key role in platform growth when they engage with and invest in the platform, focusing their resources and continued efforts on producing and providing value-adding complementary products through the platform (Boudreau, 2012; Jacobides, Cennamo, & Gawer, 2018; Saadatmand, Lindgren, & Schultze, 2019; Wang & Miller, 2020). Firms are constantly scouting their focal and adjacent markets and competitors, and the observed and expected market dynamics inform their actions (Smith et al., 2001). The conceptualization of adjacent markets depicts them as the set of markets that firms could feasibly consider when trying to grow outside of their focal market (Zook & Allen, 2003). While early conceptualizations of adjacent markets include a geographical dimension, this has become less relevant for today’s digital markets. More recent conceptualizations emphasize the closeness of required technological capabilities and customer needs (Lehmberg, Dhanaraj, & White, 2019). In the context of digital platforms, the specific segment of a product category in which the firm competes (e.g., sports games) represents the core, focal market of a complementor firm while adjacent markets are related focal markets in that product category that present demand and capability requirements overlapping those of the firm’s focal market (e.g., in the example above, the rest of the market segments in the broader “games” category, such as action games or adventure games). This conceptualization, in line with recent work, emphasizes the need for adjacent markets to be feasible for a firm given its capabilities and experience in the focal market. As Lehmberg et al. (2019: 1430) note, adjacent markets “are adjacent in the sense that the technological requirements to enter a new adjacent market build on the requirements for previous markets, and the additional technological challenges required for entry into this new adjacent market are relatively modest.”

This conceptualization leads us to expect a complementor’s decision to continue investing in or to divest from its focal and adjacent markets after an owner’s entry to be associated with the level of competitive pressure the entry exerts on the complementor; the level of competitive pressure can affect the complementor’s expected market position and value appropriation potential (e.g., Bracha & Fershtman, 2013). A new entrant in the focal market, particularly if large and/or well-endowed, always adds uncertainty and brings the possibility of a change in competitive dynamics and market positions in the focal and adjacent markets, which, due to their closeness in terms of demand and technical requirements, should also be within the reach of the new entrant. When the entry of a new player poses substantial competitive pressure, the existing players may either confront it directly, as a preemptive measure or to regain lost ground, or avoid direct competition, such as by shifting their investments toward more distant markets or unrelated product categories (Barnett, 1993; Baum & Korn, 1996; Edwards, 1955). Competitive dynamics theory suggests that when the existing players believe that they are stronger than the new entrant, they will likely go on the offensive and confront the entrant directly with a “tit-for-tat” strategy, aiming at surpassing the entrant’s offerings to neutralize the threat. However, when they perceive the level of resources and visibility of the new entrant to be much larger than their own, the existing firms are unlikely to fight head-to-head against the aggressive entrant. As the magnitude of the threat that the new entrant represents increases, particularly in situations where the entrant benefits from large power and resource asymmetry, the existing firms will likely find it increasingly difficult to respond (Dasgupta & Stiglitz, 1980; Smith et al., 2001). Existing players cannot easily find effective competitive actions against the entrant (Chen & Hambrick, 1995; Young, Smith, & Grimm, 1996) and, thus, face the prospects of a deteriorating market position and increased risk of market viability (Hannan & Carroll, 1992). Such a situation discourages the existing players from investing in the focal market and adjacent markets that are likely targets of a resource-rich entrant. Doing so becomes increasingly risky and appears to be a “lost battle.” Retreats from the market will likely intensify among the existing players (Baum & Korn, 1996; van Witteloostuijn & van Wegberg, 1992).

The case of a platform owner with a closed and reigning approach to governance entering a complementor market resembles the scenario above. As noted earlier, a platform owner typically has a much larger resource base than its complementors and significant power and influence over them (Boudreau & Hagiu, 2009). A reigning platform owner will enter the market decisively and aggressively, intensifying competitive pressure on complementors and, thus, lowering their value-appropriation potential. The disadvantageous asymmetry they face will cause complementors to find continuing to invest in the focal and adjacent markets less attractive. They will likely substantially reduce their level of investment in these markets (Boone, 2000). The owner’s forceful exercise of power and control will reinforce the complementors’ perception of the closed and reigning governance approach as detrimental to their prospects, leading them to reduce their level of engagement in markets they consider at risk.

However, we expect a different situation when the entrant into complementor spaces is a platform owner with an open and laissez-faire approach to governance. In this case, the owner’s entry will be perceived as nonaggressive and even offering possibilities to “enlarge the pie,” bringing opportunities to all players. Competitive dynamics theory posits that the firms’ incentives to further invest in the focal and adjacent markets will likely increase when new conditions enhance their positions and increase chances to appropriate value (e.g., Boone, 2000; Nalebuff & Brandenburger, 1996). Pursuing an open, laissez-faire approach will lead the owner to exert mild or minimal competitive pressure on existing complementors and to provide them with new opportunities to grow their business through exploiting technological synergies and benefiting from the market space’s enhanced visibility. Thus, complementors’ engagement in the focal and adjacent markets will likely not diminish; indeed, given complementors expectation of positive spillover benefits from the owner’s entry, it could even increase.

Taken together, we argue that the owner’s mode of entry into complementor spaces, a reflection of the approach to platform governance, will result in markedly different engagement strategies from the affected complementors, thereby leading to different levels of investment into the markets they consider at risk of being affected by owner entry. This will particularly apply to complementors that have first-hand experience with both platforms—that is, those whose apps have multihomed. Multihoming would allow complementors to experience or witness owner entry and its consequences in each case and convey a much clearer perception of and specific reference points for the risks and benefits of entry by each type of platform owner. It then follows:

Hypothesis 2a: The entry of a platform owner with a closed and reigning approach to governance to a focal complementor market is negatively associated with the affected complementors’ investment into the focal and adjacent markets.

Empirical Context: H&F Mobile Ecosystem

Our empirical context comprises two leading mobile platforms in the United States: the Apple iOS and Google Play app stores. Apple and Google launched their app stores in July 2008 and October 2008, respectively. These two players dominate the market, fiercely competing against each other while jointly capturing over 90% of the market share on mobile platforms. In 2020, Apple and Google had around 2 million apps for download in their respective stores.

Our paper focuses on a single mobile app category, H&F, to mitigate concerns over possible confounding effects that could come from combining different categories (e.g., Foerderer et al., 2018; Shaheer & Li, 2020). The H&F category offers an ideal setting to explore our research questions, for several reasons. First, it is an established category. The H&F category ranks 8th in size out of 25 categories in the Apple iOS app store and 11th out of 35 categories in the Google Play app store. Second, the H&F category is relatively fragmented and fiercely contested. No single firm or even a small number of firms dominates the market, making it suitable for exploring within-platform dynamics. Third, platform owner entry into complementor markets took place in H&F in the two app stores just 1 month apart. Apple and Google launched their own mobile H&F apps, respectively labeled Apple Health and Google Fit, with functionality overlapping that of some third-party apps in their stores. However, due to their contrasting approaches to platform governance, the specific mode of entry that each platform owner employed in its incursion into its respective complementor market starkly contrasts with the other’s. Thus, we can examine the role of entry mode heterogeneity in platform owner entry. (The main differences appear in the section “Apple Health and Google Fit Release” below.) Finally, the launch of Apple Health and Google Fit did not affect all H&F apps equally. The H&F category can be divided into subcategories, based upon a focal area of business and its primary purpose (e.g., running apps, nutrition apps, mental apps). At their initial launches in 2014, Apple Health and Google Fit did not cover all the H&F subcategories; thus, their effect varied by subcategory. This heterogeneous subcategory effect enables us to use some H&F subcategories as a control to explore the consequences of platform owner entry.

Data

Given the emerging state of research on platform owner entry into complementor spaces, we conducted a large-scale quantitative analysis together with qualitative examination, to elucidate the underlying mechanisms and to better interpret our quantitative results.

Apple Health and Google Fit Release

Apple and Google launched their apps in their respective stores on September 17 and October 28, 2014, respectively. Both Apple Health and Google Fit provided functionality already available from some third-party H&F apps on their platform. Both Apple Health and Google Fit could work as stand-alone apps. Users could now store, record, and track a range of H&F data directly in Apple Health or Google Fit without relying on third-party apps. Simultaneously, Apple and Google shared their product’s underpinning technologies with their complementors, releasing corresponding APIs, through which third-party H&F apps could access the key functions of each owner’s app. For example, third-party H&F apps could directly embed the tracking functionality of Apple Health and Google Fit into their apps or import/export H&F user data from/to Apple Health and Google Fit. Thus, while the owner’s products had functionality that overlapped those of complementor products, they also provided a mechanism for the complementors to take advantage of the possibilities the owner’s new product represented.

Table 1 provides qualitative evidence of the industry stakeholders’ early reactions. Ambivalent sentiments had prevailed surrounding the potential impacts on the third-party H&F app ecosystems of Apple Health and Google Fit releases. The opinions mixed positive and negative predictions, hinting at considerable uncertainties among industry participants and analysts regarding the ultimate effect that the two apps would have on complementors. By and large, industry participants perceived the launches of Apple Health and Google Fit as equivalent, simply occurring on competing platforms. These ambiguities, also present in extant studies of platform owner entry, highlight the importance of our approach to parse out the role of heterogeneity in owner’s “mode of entry” on the relationship between owner entry and complementor performance.

OPEN IN VIEWER

Table 1

Ambivalent Early Perspectives on the Impact of Apple Health and Google Fit Release on Third Party H&F Apps

Perspectives	Excerpts From Business Press Releases and Interviews With Industry Experts and Industry Participants
Platform owner’s product perceived as competitive to third-party H&F apps	“Apple Health and Google Fit work directly with health and fitness wearables” [no need for third-party apps] “Google Fit uses sensors in activity trackers [as Fitbit, a complementor, does] or mobile devices to record physical fitness activities” “In iOS 8.0, the Apple system can automatically save data from compatible M7 motion processor, if available” [similar to what other apps do] “Apple Health is already logging steps (if you have an iPhone 5S or newer) and flights of stairs (if you have iPhone 6 or 6 plus)” [a central feature of many apps] “Apple Health and google Fit can keep track of your steps without help from an external app or device” [a central feature of many apps] “Truth be told, you don’t even need a wearable or additional device to get moderate use from Apple Health/Google Fit” “Unless you have tech that goes beyond what Apple or Google can offer in a smartphone, you’ll be eaten alive” [serious competition for complementors]
Platform owner’s product perceived as cooperative to third-party H&F apps	“Apple Health allows third-party apps to focus on implementing only the aspects they are most interested in, leaving other tasks to more specialized apps” [suggest complementarity] “The main aim of Apple Health is to provide a single point of reference for all of your aggregated health and fitness data … and the more third-party services you plug into it, the better …” [presents owner’s apps as hubs for complementors to connect to] “For a more complete picture, Apple Health works best when synced to another fitness tracker or smartwatch you are wearing … via that device’s companion app” [suggests complementarity] “Having a company like Apple and Google educating everyone on why this stuff matters can help the whole category become mainstream faster” [suggests positive spillovers from owner’s apps] “Apple Health and Google Fit can help reach a broader audience and forge partnerships with the traditional healthcare industry … that would be hard for start-ups to accomplish alone” [suggest positive spillovers from owner’s apps] “So, for Apple Health and google Fit app to be useful, you’ll need other health and fitness apps” [suggest complementarity]

Source: Business press releases, official app store sites, developer forums, and interviews with industry experts and industry participants.

While Apple and Google released their apps almost simultaneously, the specific mode of entry into their respective app stores markedly differed. Apple and Google have pursued a starkly contrasting approach to platform governance (Eaton, Elaluf-Calderwood, Sørensen, & Yoo, 2015; Parker, van Alstyne, & Jiang, 2017). Their respective entry mode in releasing Apple Health and Google Fit reflected this different approach to governance. On the one hand, Apple has been known to pursue a closed and reigning governance approach with tight central control. For instance, Apple has a rigorous quality-review system in place to gatekeep those who wish to publish their apps in iOS, with Apple fully controlling the distribution channel (Benlian, Hilkert, & Hess, 2015; Schaarschmidt, Homscheid, & Kilian, 2019). Apple also reserves most of the rights in appropriating developer innovation in its platform ecosystem, curating the admission of codes by third parties and controlling a set of boundary resources, such as APIs and SDKs, which Apple approves and distributes (Eaton et al., 2015; Parker et al., 2017). Such a closed approach to governance has allowed Apple to claim more of the value created within its platform—for example, Apple charges higher prices than competing platforms, thus obtaining higher margins (Parker et al., 2017). On the other hand, Google has taken a contrasting approach to governance, relatively open and laissez-faire with modest central control (Karhu et al., 2018; Parker et al., 2017). For instance, there is no formal approval process for developers to publish their apps. Furthermore, Google allows the sale of Android apps in app stores other than its own, such as the Amazon app store (Remneland-Wikhamn, Ljungberg, Bergquist, & Kuschel, 2011; Schaarschmidt et al., 2019). Also, Google provides ample access for developers to not only use platform-wide technical interfaces, such as APIs and SDKs, but also integrate or extend them or replace existing components in the Android codebase (Schlagwein, Schoder, & Fischbach, 2010). Google’s open approach, while resulting in lower revenue for the platform compared to Apple’s, has fueled platform growth by attracting a larger number of apps and developers (Parker et al., 2017).

The companies’ distinctive approach to platform governance was visible in the way each entered its complementor space when releasing its H&F app. The differences were striking. Apple tightly controlled access and imposed restrictions on user and developer use of both Apple Health and its related technologies. For example, Apple Health came as a default program on iOS devices, pre-installed in all new iPhones and automatically installed in all existing iPhones (without user permission) during iOS operating system upgrades. Apple did not even allow users to delete the app, effectively resulting in forcing users to have the app on their devices. Apple also kept Apple Health compatibility relatively closed with limited device support, making its Health app available only to iPhones. The company also tightly controlled the Apple Health-related APIs, distributing them only in a standardized format that provided little room for developers to customize or modify. Apple controlled and structured the data types and gave developers no discretion to create custom data types or units, constraining them to use only Apple-approved data types on an official list. Furthermore, Apple did not allow third-party apps to use Apple Health APIs without integrating their products with Apple Health. Apple asked developers to explicitly mention the use of Apple Health APIs in describing their app; failure to comply led to barring their apps from the store. Apple also restricted the data stored using Apple Health APIs. Data exported from Apple Health had to conform to strict Apple specifications and reside only on iPhone devices. Developers could not use in their advertising any information about users’ H&F activities coming from Apple Health APIs. Apple reserved the right to earn on ads that used information coming from Apple Health. It also required developers to share their complete dataset with Apple Health to utilize the Apple Health APIs.

Google took a very different approach, giving users and developers ample freedom in the use of the Google Fit app and its underpinning technologies. Google Fit was not pre-installed, requiring a manual download by users who wanted the app. Those wishing to install Google Fit had to search for the app in the Google Play store and install it on their smartphones as they would any third-party app. Moreover, users could choose to delete the app anytime. Google kept Google Fit relatively open by supporting a wide list of compatible devices across Android-based phones, tablets, wearable gadgets, and the web, as long as the app was signed into a Google account. Google also offered third-party developers ample choice in how to access and use the Google Fit APIs, comprising multiple separate subsets that allowed developers to mix and match for their specific needs. For instance, some of the API subsets focused on sensors, which let developers read raw sensor data on devices or wearables; others focused on recording, which let developers gather, store, or sync data of a user’s physical activities; another focused on history, giving developers access to historical fitness and wellness data and enabling data management to insert, delete, or read the data. Furthermore, developers could use Google Fit technologies in their products without having to ask consumers to install the Google Fit app. In addition, developers were free to customize the data structure and format as they wished or even to share these customized pieces of software with other developers on the platform. Google accepted many types of data from third-party apps, including those data types it did not directly support. Google also did not place specific constraints on using data stored in Google Fit APIs, except to forbid its use for questionable advertising, such as to promote the development of gambling. Moreover, Google did not set any strict requirements for third-party apps to share the data they collected through the Google Fit APIs. In practice, some apps shared their data with Google, while other apps did not share data with Google at all.

 Table 2 summarizes these key differences in owner entry mode between Apple Health and Google Fit, coming from our qualitative examination of the official app stores, news articles, and press releases. Industry participants acknowledged that Google Fit offered more flexibility and compatibility than Apple Health, which placed much emphasis on data security. However, the difference in the mode of entry between Apple Health and Google Fit was much more profound than that, and industry participants did not fully grasp how these differences would shape the H&F app ecosystem differently. Industry participants had largely perceived Google Fit as the counterpart to Apple Health and simply a reflection of the strong rivalry between these platform competitors. The following quote represents the prevalent views of participants:

There is no clear winner here. Google has a wide list of compatible devices and partners … Google Fit offers more flexibility; Healthkit offers more security of the sensitive content. But one thing we can surely say is that with Health and Fit, we have officially entered the commercial phase of digital health revolution.

OPEN IN VIEWER

Table 2

Differences in Entry Mode Between Apple Health and Google Fit

	Apple Health	Google Fit
	Owner entry under a “closed and reigning” approach to platform governance	Owner entry under an “open and laissez-faire” approach to platform governance
Product access	Apple Health was pre-installed for all new iPhones with no right to waive installation. For existing iPhones, it was automatically installed, when upgrading the iOS system (without requiring user permission). Users could not remove the app from the iPhone.	Google Fit required manual search and download from the app store, as done with third-party apps. Users were given the choice of whether to have it or not. Users could remove the app from the phones anytime.
Device compatibility	Available only on iPhones (not even available for use on iPad).	Support multiple devices and platforms, running across Android-based phones, tablets, and on the web.
Data types covered	Restricted to a list of Apple Health APIs pre-defined by Apple. Customization was not allowed.	Developers were allowed ample customization: allowed to create their own custom data types/units of Google Fit APIs, and even share with others.
Data storage	Apple Health APIs required all data to be centrally collected and stored only through Apple Health.	Google Fit APIs accepted data from third-party apps. Third-party apps could collect and share data directly without intervention of Google Fit.
API structure	One type of framework was offered.	Separate subsets of frameworks were offered (e.g., Sensors API; History API; Recording API), allowing mix & match.
Data usage	Developers could not use data from Apple Health APIs in their app advertisements. Apple reserved the right to earn on ads from those data.	Almost no constraints for developers to use the data stored in Google Fit APIs for their app advertisements.
Data share structure	Data drawn from Apple Health APIs were required to conform to specifications dictated by Apple. Apps were required to share their complete dataset with Apple Health, when using Apple Health APIs.	Data drawn or collected through Google Fit APIs were not required to be shared with Google and could be collected and stored in whatever format was preferred by developers.

Note: The description here refers to the initial release of Apple Health and Google Fit (late 2014), aligned with the period of our empirical analysis.

Source: Business press releases, official app store sites, developer forums, interviews with industry experts and industry participants.

This sentiment among industry practitioners substantiates the importance of our study’s systematic look at the difference between Apple Health and Google Fit, drawing on a careful quantitative examination using a large-scale panel dataset.

Types of Apps Affected by Apple Health and Google Fit Release

As noted earlier, the apps that belong to the H&F category can be divided into subcategories based upon their primary focus of activity and functionality. Using the descriptions both app stores provided and after consultation with our industry informants, we distinguished nine subcategories within H&F: running/walking (pedometer)/cycling, physical workouts/training, food/nutrition, sleep, women’s health, hydration, medical, body measurement, and others. Table 3 shows a brief description of each subcategory.

OPEN IN VIEWER

Table 3

Subcategorization Within the Health and Fitness Category

H&F Subcategory	Apple iOS (Apple Health)		Google Play (Google Fit)		Key Criteria
	Treatment	(% in H&F)	Treatment	(% in H&F)
1. Running/walking/cycling	Affected	13.66	Affected	13.02	GPS tracking-activities related
2. Physical workouts/ training	Affected	27.27	Unaffected	29.17	Physical fitness class; training sessions; workout enhancing activities related
3. Food/nutrition	Affected	13.50	Unaffected	15.01	Calorie or macro-nutrition related; meal-planning; diet program
4. Sleep well	Affected	6.14	Unaffected	5.36	Sleep cycle; sleep inducing activities related
5. Women’s health	Unaffected	2.01	Unaffected	1.91	Fertility or ovulation tracking related
6. Hydration	Unaffected	1.11	Unaffected	0.54	Hydration level tracking; hydration inducing activities related
7. Medical	Affected	3.90	Unaffected	4.44	Vital signs (e.g., blood pressure, pulse rate, glucose, blood alcohol content, body temperate); medical related
8. Body measurement	Affected	3.22	Affected	2.53	Weights; heights; BMI; body-fat measurement related
9. Others	Unaffected	29.19	Unaffected	28.02	Offline business support; baby & pregnancy; beauty; mental health; e-magazine & book; smoking & alcohol

We set out to identify these subcategories by considering how the app stores listed those apps on their H&F page. While neither Apple nor Google used a specific set of subcategories for H&F, they sometimes listed apps under themes such as Sleep Well, Nutrition, and Workouts. We carefully reviewed titles, app descriptions, and snapshots of those apps, to understand how the app stores sorted them into subgroups. We then looked at how leading mobile analytic sites grouped H&F apps and how H&F journals referred to groups of apps when they reviewed them. We focused particularly on the core functionality of each app and used that together with the groupings the platforms, mobile analytic companies, and H&F journals employed, to produce a comprehensive but parsimonious list of subcategories. For example, we grouped running/walking/cycling apps as one subcategory because they all use the GPS tracker for their core functionality. Also, as was often the case, most apps in this subcategory offer the three capabilities together (running, walking, and cycling), further suggesting that these activities can be grouped as one H&F domain. Last, we polished this subcategorization through iterative discussions with industry experts and informants, looking for mistakes or omissions. We then carefully coded our data by assigning each app to a corresponding subcategory, based on its title, description, and snapshots determining its capabilities. We checked the accuracy of this coding task by drawing a random sample of apps and asking our research assistant to replicate the task. The interrater reliability was over 86%.

The impact of Apple Health and Google Fit varied by subcategory. For instance, at their initial launch, Apple Health and Google Fit both offered tracking as a key functionality, and this functionality covered running activities, thereby directly affecting running apps. However, neither Apple Health nor Google Fit initially included tracking functionality for activities related to women’s health (e.g., tracking fertility or ovulation cycles). Therefore, unlike running apps, the initial release of Apple Health and Google Fit did not directly affect apps focusing on women’s health. We exploited this heterogeneity by subcategory in the extent and timing of each platform owner entry. Our treated group comprises apps whose focal subcategory was directly affected by the owner’s product. Our control group comprises the apps whose primary subcategory was not directly affected by the owner’s product. Apple Health and Google Fit affected different subcategories in H&F within their respective platforms, as Table 3 shows. The initial release of Apple Health in 2014 affected the H&F apps in the subcategories running/walking/cycling, physical workouts/training, food/nutrition, sleep well, medical, and body measurement. It did not affect the remaining three subcategories, women’s health, hydration, and others. The initial release of Google Fit in 2014 affected the H&F apps in subcategories running/walking/cycling and body measurement, while the other seven subcategories remained unaffected.

Based upon this classification, as Table 3 illustrates, we could have run our analyses drawing on affected (unaffected) groups. However, to further account for any potential unobserved heterogeneities that may be associated with assignment to a set of subcategories across these two platforms, we kept our analysis to a common set of affected (and unaffected) subcategories across the two platforms. ² Therefore, our analyses only consider the subcategories that were commonly affected or unaffected on both platforms. This led us to consider the subcategories running/walking/cycling and body measurement as affected in the analyses of both platforms and the subcategories of women’s health, hydration, and several of the subcategories included in “others” as unaffected apps in both platforms. The subcategory “others,” as seen in Table 3, comprises multiple subgroups, such as baby and pregnancy, mental health, beauty, quitting tobacco and alcohol, and miscellaneous groupings. We thus further scrutinized the others subcategory to make sure that they were classified in the same way in each platform, either affected or unaffected when they corresponded. We excluded from the analysis those that did not meet this condition—these accounted for only 2% or less of the total apps in the “others” subcategory.

Variables

Control variables

We include a set of control variables that can affect app performance and complementors’ engagement decisions. The variable app age and its square term account for an app’s exponential downward-sloping curve (operationalized as time elapsed since an app was first launched; a proxy for an app’s vintage effect). The variable app reviews is a proxy for an app’s visibility and popularity (measured by the number of reviews an app received in a given week, using a logarithm form to account for skewness). The variable app rating is a proxy for an app’s perceived quality, operationalized as the app’s average review score in a given week on a 0–5 scale, where 0 indicates an app with no rating, as in previous studies (e.g., Zhu & Liu, 2018). Table 4 reports summary statistics and correlations for our variables.

OPEN IN VIEWER

Table 4

Descriptive Statistics and Correlation

	Hypothesis 1
Apple iOS Platform (App-Week Observations)
Variable	M	SD	1	2	3	4	5
1. log(Product performance)	2.43	2.43
2. Affected apps by platform owner entry	0.35	0.48	.04
3. After platform owner entry	0.80	0.40	–.01	.00
4. log(App reviews)	0.42	0.95	.55	.11	–.02
5. App rating	0.87	1.68	.47	.06	–.03	.76
6. App age	944.11	540.87	.08	.00	.18	.03	.01

Note: The table reports means, standard deviations, and Pearson's correlation coefficients.

OPEN IN VIEWER

Google Play Platform (App-Week Observations)
Variable	M	SD	1	2	3	4	5
1. log(Product performance)	2.34	2.79
2. Affected apps by platform owner entry	0.34	0.47	.09
3. After platform owner entry	0.76	0.42	–.08	–.00
4. log(App reviews)	0.82	1.30	.71	.19	–.01
5. App rating	1.52	1.98	.57	.11	–.03	.74
6. App age	834.67	397.06	.05	–.04	.27	.04	.03

Note: The table reports means, standard deviations, and Pearson's correlation coefficients.

OPEN IN VIEWER

	Hypothesis 2
Apple iOS Platform (Developer-Week Observations)
Variable	M	SD	1	2	3	4	5
1. log(Platform investment)	0.01	0.09
2. Affected apps by platform owner entry	0.35	0.48	.01
3. After platform owner entry	0.81	0.39	–.01	.00
4. log(App reviews)	0.24	0.63	–.00	.05	.01
5. App rating	0.57	1.34	.00	.03	–.00	.78
6. App age	937.28	549.91	–.05	.02	.16	–.02	–.03

Note: The table reports means, standard deviations, and Pearson's correlation coefficients.

OPEN IN VIEWER

Google Play Platform (Developer-Week Observations)
Variable	M	SD	1	2	3	4	5
1. log(Platform investment)	0.01	0.07
2. Affected apps by platform owner entry	0.33	0.47	–.01
3. After platform owner entry	0.76	0.42	–.00	–.00
4. log(App reviews)	0.75	1.17	–.00	.18	–.01
5. App rating	1.43	1.85	–.01	.10	–.04	.74
6. App age	801.18	375.68	–.03	–.04	.28	.03	–.00

Note: The table reports means, standard deviations, and Pearson’s correlation coefficients.

Our final sample includes, for Apple iOS app store, 13,696 observations at the app-week level and 8,192 observations at the developer-week level and for Google Play app store, 8,640 observations at the app-week level and 5,112 observations at the developer-week level.

Methodology

We use a fixed-effects ordinary least squares (OLS) specification to examine changes associated with a platform owner entry on the performance of affected complementor products and the level of complementor engagement with the focal platform. Previous studies on the topic used a similar research design (e.g., Foerderer et al., 2018; Wen & Zhu, 2019; Zhu & Liu, 2018). We employed fixed effects at the subcategory level and platform level to control for possible unobserved subcategory-specific and platform-week-specific heterogeneities. ⁴ We report robust standard errors clustered around the subcategory level to account for serial correlation of the error term. Our Hypothesis 1 explores the product performance of the affected complementors after platform owner entry; therefore, our analysis for H1 occurs at the app level. Our Hypothesis 2 explores the new product development activities of the affected complementors after platform owner entry, so our analysis for H2 occurs at the developer level. The following equations represent our model of choice, in which i indexes subcategories, j indexes apps, k indexes developers, l indexes platforms, t indexes time (weeks), and u is the error term:

log(Product performance _jilt ) = β0 + Affected apps by platform owner entry _jilt β1 + After platform entry _jilt β2 + Affected apps by platform owner entry _jilt × After platform owner entry _jilt β3 + α _i × α _l  + u _jilt

Ideally, we would randomly assign complementors to affected and unaffected domains by platform owner entry. Our qualitative evidence (as Table 1 shows) suggests that complementors in our sample might have been unable to anticipate the extent and directionality (positive, negative) of the impact of Apple’s and Google’s entry into the H&F area. As detailed earlier, industry participants were uncertain about the impact that Apple Health and Google Fit would have on complementors, largely considering Apple Health and Google Fit as equivalent offerings happening on competing platforms. There was also scant information or anticipation regarding the precise scope of the impact (i.e., which H&F subcategories would be most affected). For example, the subcategory “women’s health” was not part of the initial 2014 release of Apple Health or Google Fit. However, Apple included it 1 year later when it released an Apple Health update. Industry informants concurred that the primary reason for Apple not including this subcategory in the original release was that most Apple developers were male and inadvertently omitted this subcategory in its initial launch.

Despite this, we conducted additional analyses to further account for potential unobserved heterogeneities that may be associated with belonging to a certain H&F subcategorical domain on each platform, which could potentially create omitted variable bias and selection bias. First, we employed a coarsened exact matching (CEM) analysis (Blackwell, Iacus, King, & Porro, 2009; Iacus, King, & Porro, 2012). The CEM analysis allowed us to pair unaffected apps with affected apps that are similar except for the fact that the latter experienced platform owner entry. We matched affected and unaffected apps on the variables app age, app reviews, app rating, a set of key dimensions that could result in confounding effects. The degree of imbalance is reduced for both samples after matching. For Apple, L1 statistics before matching is 0.14, and it becomes 0.06 after matching (57% reduction in imbalance). For Google, L1 statistics before matching is 0.46, and it becomes 0.16 after matching (65% reduction in imbalance). After applying CEM, we observe no statistically significant differences between the elements of the two groups, suggesting that the two samples are quite similar (see Appendix A for details).

Moreover, we further restrict our analyses to the set of apps that multihome—that is, apps available on both Apple and Google platforms. ⁵ Multihoming apps that share identical characteristics at the product and developer levels can present a unique setting in the context of a mobile app market wherein comparisons between an identical set of apps are possible (Chen, Zhang, Li, & Turner, 2022). This allows us to better control for any potential differences in product or complementor characteristics that could be associated with the dynamics on different platforms. The resulting comparison groups become thus as similar as possible given our dataset.