Platform Desertion by App Developers

Abstract

Platform desertion, or a developer’s stopping the development of an app for a platform, is a widespread phenomenon to the detriment of platforms. However, the extant literature focuses primarily on why app developers join—not leave—a platform. This app-level study develops two ideas: (a) coordination costs borne by an app’s developer are associated with platform desertion, and (b) these costs are, in turn, shaped by a nuanced interplay between app decision rights and app “microarchitecture” introduced here. We use survey and snapshot archival data spanning 2009–2014 on over 300 apps in the Mozilla Firefox ecosystem to test these ideas. Our novel contribution shows how, by influencing coordination costs, the previously invisible interplay between app decision rights and app microarchitecture shapes an app’s platform desertion. We find that delegating app decision rights to its developer weakens the coordination cost-reducing benefits of decoupling an app from the platform but strengthens those of standardizing its interfaces to the platform. The key theoretical implication is that app decision rights and app microarchitecture symbiotically influence the coordination costs borne by an app’s developer. The key practical implication for platform designers is that the choices about who ought to make what decisions are intertwined with the architecture of the governed information technology artifact.

Key words and phrases:

• app developers
• apps
• architecture
• coordination cost
• decision rights
• ecosystems
• mobile apps
• modularity
• platforms

Notes

¹ Firefox, for example, adopted a rapid release philosophy in 2011, whereby it releases new version every few weeks; any such change can jeopardize an app’s interoperability with it.

² A contrast of Apple’s iOS versus BlackBerry software platforms illustrates this point. Apple provides a variety of developer-friendly tools including simulators, technical guidelines, APIs, and programming tools to make it easier for iOS app developers to take advantage of the platform’s evolving capabilities and to ensure their apps’ interoperability [16, 17]. In contrast, such integration is much more difficult for programmers developing apps for BlackBerry, which in turn has an active app developer community about one-tenth the size of Apple’s. This however is a cross-platform-level difference whereas our focus is zoomed in on differences among apps within a single platform.

³ Prior empirical studies almost always aggregate decoupling and interface standardization with little theoretical rationalization (e.g., [47, 66, 75]). For example, Tiwana and Konsynski [69] ironically show these facets to be empirically distinct yet aggregate them formatively and Nambisan [47] did not measure the underlying dimensions.

⁴ Fama and Jensen [27] refer to these collectively as decision management rights.

⁵ DR delegation at the app level represents organizational modularization, which Sanchez [55] defines as the degree to which a party can function autonomously and concurrently vis-à-vis another. This conceptualization using decision right structure to represent organizational structure is consistent with others’ definitions of organizational modularity such as unbundling of decision rights [38] and a “quasi independent” organizational structure [56].

⁶ For example, Apple emphasizes intuitive usability and minimal replication of functionality provided by the iOS platform’s APIs and native iOS apps. Similarly, BlackBerry emphasizes tight integration of BlackBerryOS with apps targeted primarily to business users; Firefox prioritizes end-user security and privacy above all other demands. An app must be sufficiently consistent with the platform’s vision.

⁷ An example of this is the fragmentation of the Android platform into many inconsistent and incompatible derivative versions of Android (“forks”) by Amazon, Samsung, LG, and Sony [50].

⁸ We thank an anonymous reviewer for suggesting this point.

⁹ This is akin to an astronomer concluding that Halley’s comet does not exist after fruitlessly trying to observe it for 20 years (the observation interval), although it appears once every 75 years (the phenomenon’s existence interval, in these authors’ language).

¹⁰ Using t1 ratings instead of t2 ratings produced consistent results.

¹¹ Developers’ perceptions about the controlling nature of the platform owner are likely heterogeneous, as its high variance also suggests.

¹² Using these controls on coordination costs misspecifies the model [12]; the results, however, were robust when controls were used in this manner.

¹³ A multistep test for mediated moderation met the conditions for interface standardization, that is, (1) a significant relationship between the mediator and platform desertion (β = .41; p < .05), (2) moderation of its effect on coordination costs (β = –.15; p < .01), (3) mediation by coordination costs of this interaction term’s effect on platform desertion (mediation test significant at p < .05), and (4) a drop in direct effect path from the moderator when the interaction term is introduced. However, the corresponding test for decoupling failed the third mediation-test step, providing no evidence to support mediated moderation for it. The (app interface standardization * app decoupling) interaction term included in Model 1.4 in Table 4 to avoid model underspecification was nonsignificant (β = .08, T = 1.46; ns), indicating that they play independent roles in reducing coordination costs.

Additional information

Notes on contributors

Amrit Tiwana

Amrit Tiwana is a professor of management information systems in the Terry College of Business at the University of Georgia. His research focuses on systems development and IT governance at the firm, interfirm, and ecosystem levels. His work has appeared in Information Systems Research, Journal of Management Information Systems, MIS Quarterly, Decision Sciences, and IEEE Transactions on Engineering Management, as well as in strategy, software engineering, marketing, and finance journals such as Strategic Management Journal, California Management Review, ACM Transactions on Software Engineering, Journal of Academy of Marketing Science, Journal of Financial Intermediation, and others. He is a senior editor at Information Systems Research and is on the boards of Journal of Management Information Systems and Strategic Management Journal. He also served as the conference program cochair for the 2014 Americas Conference on Information Systems (AMCIS). His research has been supported by various government agencies in the United States, Europe, and Asia, as well as by firms including UPS, IBM, Hitachi, Mitsubishi, Toshiba, Sumitomo Steel, Kansai Electric, and Nippon Telegraph and Telephone.