Platform Ecosystems:

How Developers Invert the Firm

Geo↵rey Parker Marshall Van Alstyne Xiaoyue Jiang
Dartmouth College/MIT Boston University Quinnipiac University

17-August-2016

For a period starting in 2015, Apple, Google, and Microsoft became the most
valuable companies in the world. Each was marked by an external developer
ecosystem. Anecdotally, at least, developers matter. Using a formal model of
code spillovers, we show how a rising number of developers can invert the firm.
That is, firms will choose to innovate using open external contracts in preference
to closed vertical integration. The locus of value creation moves from inside the
firm to outside. Distinct from physical goods, digital goods a↵ord firms the chance
to optimize spillovers. Further, firms that pursue high risk innovations with more
developers can be more profitable than firms that pursue low risk innovations with
fewer developers. More developers give platform firms more chances at success.
Our contribution is to show why developers might cause a shift in organizational
form and to provide a theory of how platform firms optimize their own intellectual
property regimes in order to maximize growth. We use stylized facts from multiple
platform firms to illustrate our theory and results.

Keywords: Open Innovation, Sequential Innovation, Platforms, R&D Spillovers, Intel-

lectual Property, Network E↵ects, Network Externalities, Bundling, Two-Sided Net-
works, Two-Sided Markets, Vertical Integration, Standard Setting Organizations, Plat-
form.

JEL Codes: D02, D21, D62, D85, L00, L17, L22, L24, L5, O3, O31, O32, O34.

Electronic copy available at: https://ssrn.com/abstract=2861574

1 Introduction and Background

One of the most important questions a firm can ask is how to efficiently create value: should
it produce its own output or should it orchestrate the output of others. In the case of code,
the choice increasingly favors orchestration over production. Apple, Google, and Microsoft,
for example, became the three most valuable companies in the world in 2015, having passed
energy and investment firms Exxon-Mobile and Berkshire-Hathaway.1
We argue that developer communities are inverting the firm. That is, firms must now
manage value creation that occurs externally just as carefully as they manage the value
they create internally. And, this is not just outsourcing. Firms are relinquishing product
specifications to third parties that they do not even know. We provide a formal theory
of why this is happening and ask what levers platform firms have to encourage external
developers to innovate on their behalf. Our particular focus is on digital innovations that
developers create to extend platforms. Digital platforms di↵er from physical systems such
as automobiles because the subsystem boundaries can be more loosely defined, which makes
recombination of elements less costly, and because information is non-rival. Reusable code,
for example, facilitates knowledge spillovers. Interestingly, however, even firms that produce
product platforms such as automobiles, tractors, and turbines are adding a digital layer to
create an Internet of Things (Evans and Annunziata, 2012).
The role of developers has become so central in digital ecosystems that firms have de-
veloped strategies for “platform evangelism” to manage third party contributions that have
become central to platform success.2 Reasons that developers are so important in digital
platforms include well known features of digital technology such as malleability of the code,
1
http://finance.google.com. Accessed 9 Nov. 2015.
2
The Open Automotive Alliance?s mission statement provides an example of the trend towards digital
innovation on top of physical products: “The members of the Open Automotive Alliance share a vision for
the connected car, and are committed to collaborating around a common platform to make this vision a
reality.” http://www.openautoalliance.net/. Accessed 9 Nov. 2015.

Electronic copy available at: https://ssrn.com/abstract=2861574

the low cost of investing in tools to develop code, close to zero cost reproduction, and the
potential to profit from application successes while shedding the costs of failures. Put more
broadly, developers are key to a platform?s ability to scale rapidly because what the platform
firm does is not limited by the processes of hiring, training, project selection, and coordina-
tion. Instead, these processes are distributed outside of the platform, allowing much more
rapid growth (Parker et al., 2016).
Others have observed the importance of digital ecosystems and the strategies that firms
have to seek advantage. In an agenda setting paper for the field, Yoo et al. (2010) note that
firms need to ask what they should open and what they should close in a digital product
platform. Kallinikos et al. (2013) adopt the term “digital artifact” and characterize digital
objects as open, reprogrammable, and accessible by other digital objects.
We are far from alone in observing that an historic shift is underway, driven by rapid im-
provements in network connectivity and computing power (Brynjolfsson and McAfee, 2014).
Earlier improvements in transportation technology changed the locus of economic activity
from vertically integrated firms to firms organized around a nexus of supplier networks. We
believe that the current shift goes farther still. Using loosely affiliated ecosystems, firms are
able to harness a global network of partners they don’t even know beforehand who can con-
nect through digital networks to innovate on top of a core set of resources thereby creating
highly valuable products and services for their users.
In our context, a relevant innovation is any digital application, including a product or
service, that is produced by an ecosystem partner using core platform resources. The rate
of innovation is the rate at which developers produces digital applications for the platform.
A platform can be conceived as a “layered architecture of digital technology” (Yoo et al.,
2010). This includes a device layer, network layer, service layer and content layer. To this
definition, we also add the governance rules that organize the ecosystem. The platform owner
then influences developer innovation by exposing more platform resources e.g. by opening

Electronic copy available at: https://ssrn.com/abstract=2861574

the architecture (APIs, SDKs, code libraries, templates, etc.), and by o↵ering more favorable
standard licensing agreements (SLAs), e.g. by o↵ering exclusivity for new apps.

Average selling price (and gap) of Total Number of Apps by App Store
iOS and Android smartphones
iTunes Google Play
worldwide 1,5000 k

iOS Android Price Gap

1,250 k

$702 $261 $710 $330 $686 $366

$650 $374 1,000 k

750
$441 $380 k
$320 $276
500
k

250
k

0k
2010 2011 2012 2013 2010 2011 2012 2013 2014

(a) The more open system has the lower price. (b) The more open system grows faster.
Source: Statista using IDC data. Source: AppFigures.

Figure 1: Openness entails a tradeo↵. It nets less revenue on the core platform but fosters
more applications development.

The battle between Apple iOS and Google Android provides a useful example of dif-
ferent platform choices around layered architecture and openness. Although Apple courts
third party developers, it remains a relatively closed system along the dimensions catalogued
in Eisenmann et al. (2009). For example, developers who wish to publish iOS applications
must submit to Apple’s rigorous quality review and Apple controls the only distribution
channel. Eaton et al. (2015) describe the tensions between Apple, developers who build on
top of iOS, and its end-users many of whom have “bricked” their devices in order to gain
access to apps outside Apple’s control. In contrast, Google has released Android under an
open source license and has attracted more developer attention even though it launched after
iOS.3 Google’s Android system is open to hardware manufacturers to the point that Google
published reference designs to reduce their fixed costs in creating Android handsets (McAl-
lister, 2011). To retain modest control over Android, Google makes API access dependent
upon a Google Play subscription (Amadeo, 2013). Parker and Van Alstyne (2014) describe
the strategic challenges such firms face when competing as platform ecosystems.
3
http://readwrite.com/2014/01/08/app-store-sales-google-play-android

3
 The di↵erent decisions have led to di↵erent market outcomes for both the price each
platform charges and the number of applications provided by developers. As Figure 1a
shows, the closed Apple platform is able to charge higher prices for the core platform, giving
Apple much higher margins. By contrast, Figure 1b shows that the more open Android
platform has attracted more applications development. The number of Android apps passed
those of Apple iOS in 2014. Openness therefore entails a tradeo↵. It nets less revenue on
the core platform but fosters more applications development.
The need to open a platform to facilitate developer innovation is nicely stated by Laursen
and Salter (2014): “... in order to obtain knowledge, organizations have to reveal some parts
of their own knowledge to external actors” (page 868). In e↵ect, this argues for the benefit
of knowledge spillovers. Attaching to and building on others’ ideas is easier conditional
on access to these ideas. Importantly, in the case of software, Eilhard and Ménière (2009)
provide empirical evidence that open code produces significant knowledge spillovers. In a
study of 10,553 open source projects on SourceForge, they find evidence of non-decreasing
returns to scale. Productivity of developers rises substantially with access to code libraries,
especially if modules use the same language. The decision to fold new private code back into
an existing public code library thus represents a design parameter of a healthy ecosystem.
Unsurprisingly, given the complexity of platform markets, executives struggle and have
disagreed over how to manage their developer ecosystems (Libert et al., 2014). At the closed
end of the spectrum, Tivo used provisions of the Digital Millennium Copyright Act to lock out
industry players who sought to attach to its proprietary systems (Slater and Schoen, 2006).
This allowed Tivo to charge more for its innovations. However, closure kept its ecosystem
small. Ironically, the iPhone was also completely closed when Apple introduced it in 2007.
Not until hackers broke into it in order to add new features did Apple release an approved
System Development Kit (SDK) (Eaton et al., 2015). At the open end of the spectrum, UNIX
firms have lost control over Application Programming Interfaces (APIs) to committees and

4
have reduced pricing power as a result (West, 2003). RedHat, for example, uses standard
GNU Public License terms that give anyone who receives their code the right to modify and
distribute copies of the enhanced code for free. This fosters ecosystem participation, but
competitors promptly absorb all valuable innovations.
Based upon our interviews with the global heads of platforms at Cisco, Haier, SAP, and
Thomson Reuters, we examine two key decisions that platform managers must make. These
are (1) how much of the core platform to open in order to spur developer innovation and (2)
how long to grant developers the right to benefit from sales on top of the platform before
the platform absorbs those innovations into the core. We explore how these decisions are
a↵ected by competition, level of vertical integration, number of developers, and the risk of
innovation failure.
Despite considerable research on prices, quantities, and network e↵ects, Yoo et al. (2010)
note that little formal analysis has investigated broader platform business models. In ad-
dition, Tiwana (2013) calls for research on how platform governance a↵ects ecosystem in-
novation. A growing literature has focused on leadership (Gawer and Cusumano, 2008),
economics (Bresnahan and Greenstein, 1999; Farrell et al., 1998), launch (Bhargava et al.,
2012), and strategies for managing platforms (Boudreau, 2010; Cusumano, 2010). Markovich
and Moenius (2009) analyze competitive platform dynamics and show that weak develop-
ers can benefit from value added by strong developers. Zhu and Iansiti (2012) show how a
platform entrant can overcome an incumbent based on the strength of developers’ indirect
network e↵ects. Huang et al. (2012) show that developers with stronger property rights can
more successfully resist expropriation by the platform. Scholten and Scholten (2011) identify
control points that allow the platform sponsor to charge for access. The two-sided literature
conceives of platforms as mediating markets with network externalities that cross distinct
user groups and shows how subsidies to one group become optimal (Parker and Van Alstyne,
2000a,b; Caillaud and Jullien, 2003; Rochet and Tirole, 2003; Parker and Van Alstyne, 2005;

5
Eisenmann et al., 2006; Rysman, 2009).
To build the ecosystem, platform sponsors often embrace modular technologies and en-
courage partners to supply downstream complements in competitive markets (Baldwin and
Clark, 2000; Fine, 1999; Boudreau, 2010). Loose integration promotes layered industries. In
the personal computer industry, for example, these layers consist of semiconductor manufac-
ture, PC assembly, operating system, and application software, among others (Baldwin and
Clark, 2000; Grove, 1996; Shapiro and Varian, 1999). The credit card and telecommunica-
tions industries are similarly layered (Evans et al., 2006).
Ondrus et al. (2015) study platform openness with a focus on how the openness decision at
the firm level (within and across industries), the technology level, and the user level a↵ects
the overall market potential. The key issue they study is the likelihood of a particular
platform achieving critical mass. In our study, we approach the openness decision from
the opposite direction and focus on how competition, technical success, size of developer
base, and network e↵ects change the decisions platform managers make around openness
and time-to-bundle.
To make progress toward understanding the trade-o↵s inherent in managing platforms,
we use tools from the economics of industrial organization. Specifically we begin with a
baseline analytic model introduced to the literature by Parker and Van Alstyne (2015). This
work provides a tractable model of the core elements of a platform and ecosystem economy
including production and recursive innovation. We extend this model by adding competition
among multiple developers to gain additional insight into the behavior of platform systems.
Our results also add to our understanding of the platform corporate form relative to vertical
integration. We extend the base model to add N developers who compete with one another.
From this baseline, we develop the following new results around optimal openness , optimal
developer innovation duration t, competition among developers, and network e↵ects.
Our analysis then yields the following three results that show how platforms invert the

6
firm by increasing the need to manage innovation and production that occurs externally.
(1) Platforms bundle innovation more quickly as the number of developers Nd increases
and as the likelihood of developer technical success increases so long as the second period
number of developers is equal to or below the number in the first period. Platforms also
increase their openness in both the number of developers and the likelihood of technical
success increases until a threshold level of N after which the platform reduces openness. (2)
Platform-to-platform competition strictly increases the level of openness as a function of the
number of developers. (3) Firms with small developer bases and minimal network e↵ects will
prefer vertical integration or closed sub-contracting to open innovation. We develop each of
these results and discuss their meaning below.
This paper addresses key topics of this special issue. In particular, we adopt the view
that IT platforms can reshape innovation ecosystems. When the number of developers is
sufficiently high, a platform with a default contract provides a new architecture for digital
innovation. Decisions about how long to protect developer innovation, and especially how
much to open the platform are at the intersection of IT, organizational design, and innovation
and, as such, respond to the questions raised in the call for papers.

2 Model with Developer Competition

We extend Parker and Van Alstyne (2015) by explicitly incorporating Nd , the number of
developers in the analytical framework. Such an extension enables analysis of how the key
platform decisions of exclusionary period and openness are a↵ected by technical risk, levels
of competition (among developers and among platforms) and network e↵ects. It also permits
analysis of how spillovers across developers can change organizational form.

7
2.1 Platform Formulation

Ecosystem participants include platform sponsors, developers, and consumers. Table 1 in the
Appendix summarizes variable definitions. We draw the following elements from Parker and
Van Alstyne (2015). V is the value of the core platform before developers add applications.
This is what a user would would pay for a standalone platform. For example, when the
Apple iPhone first shipped, it had network connectivity, an email reader, a web browser, a
calendar app and a handful of basic applications. In 2007, it was closed and did not o↵er
third party applications until 2008 (Eaton et al., 2015).
To capture sequential innovation, let time span two periods of equal length t. At time
zero, a platform sponsor makes fraction of its platform’s value publicly available to devel-
opers, representing free access to reference diagrams, code libraries, APIs and SDKs. The
other (1 ) portion remains private and is held by the platform sponsor. As a motivating
example, consider the “Price Gap” in Figure 1a. This value, represented by public code
V , is not captured by the platform owner. Instead, the platform owner captures only the
residual private code (1 )V . By contrast, a fully closed system would capture V and thus
have higher baseline profit.
Developers add value according to a standard Cobb Douglas production function with
technology parameter ↵. We denote the output of an individual developer in each period as y1
and y2 respectively. Period one output is y1 = k( V )↵ and period two output is y2 = k(y1 )↵ =
2
k 1+↵ ( V )↵ . Assume code reuse k > 0 and concave technology 0 < ↵ < 1. Developer
innovation is thus recursive. If developer code stays closed, other developers cannot build
upon it, there are no spillovers, and 3rd party production has no value. If developer code goes
open, a choice the platform determines in its contract, developers can build on it in the next
period, and there are code spillovers. Although the production function stays constant and
concave, the e↵ect of reuse rises from k to kk ↵ and the e↵ect of technology strengthens from

8
( V )↵ to (( V )↵ )↵ . This formulation has the attractive property of capturing knowledge
spillovers such as those found in Eilhard and Ménière (2009). Their use of the translog
production function for econometric analysis is a direct generalization of the Cobb Douglas
form used here.
If code is open, it is public and easily copied. No one can charge for public code. If
code is closed, it is private and developers can charge for what they produce. The platform
cannot tax developers who have no revenues. Openness thus has a benefit of increasing code
spillovers and boosting innovation but a cost of decreasing ability to charge. To balance these
conflicting interests, the platform sets an exclusionary or non-compete period t during which
developers can charge for their code. Like a patent, however, the platform will make the code
public after the non-compete period expires. We have direct evidence that the non-compete
period is a strategic variable whose duration companies choose. For example, SAP publishes
an 18–24 month roadmap that articulates “white space,” alerting developers that they will
not copy their their innovations until after this time. After that, any developer innovation
may be absorbed into the SAP core. Apple reserves the right to appropriate developer
innovations and reuse them in its ecosystem.4 Similarly, Cisco bundles features that have
appeared among multiple developer products into its core network operating system where
they can be accessed via API calls by ecosystem partners. “Developers don’t like it but
realize it’s good for the ecosystem.”5
At the time developer innovations become open, the competitive price falls to zero.
Knowing apps will be free in the future, a strategic customer optimizes between consum-
ing the innovation at p = v or waiting until time t, which at a conventional interest rate
4
“Apple works with many application ... developers and some of their products may be similar to
or compete with your applications. Apple may also be developing its own ... competing applications
... or may decide to do so in the future. ... Apple will be free to use and disclose any licensee
[code] on an unrestricted basis without notifying or compensating you.” Section 10.3 of the standard
iOS license. https://developer.apple.com/programs/terms/ios/standard/ios_program_standard_
agreement_20140909.pdf. Accessed Nov. 9, 2015.
5
Authors’ interview with Guido Jouret CTO Emerging Markets Group, Cisco Systems Inc. 9-8-2006.

9
 rt
implies a discount of = e . The option to wait thus means that a customer will only
pay for the incremental value of immediate consumption v relative to the discounted value
of future consumption v. The indi↵erent consumer thus accepts a price no higher than
p = v v = v(1 ). For simplicity, we assume market size is 1, and that a developer
charges the highest price possible to the indi↵erent consumer, who then enjoys surplus V .
Deviating from the original setup in Parker and Van Alstyne (2015), we then introduce
a new parameter, Nd 1, to represent the number of developers who compete with one
another. This will interact with the production technology and the choice of openness to
drive spillovers. Platform sponsor profit can then be written as:

1 1 2
⇡p = V (1 ) + v(1 )k( V )↵ + v(1 )k 1+↵ Nd↵ ( V )↵ . (1)
2 2

Increasing the number of developers Nd raises output in period one such that y˜1 =
Nd y1 . Recursive production then yields y˜2 = Nd1+↵ y2 , where the additional Nd↵ follows from
production spillovers y2 (Nd y1 ) of period one developers. Additional developers, however,
reduce the pricing power of any given developer in the manner of Cournot competition. The
1
exact formula for price under Cournot competition is p̃ = N +1
p.6 To make analysis more
tractable, simply interpret Nd as N 1, representing the number of other developers beyond
1
the first one. This allows us to use the simpler form p̃ = Nd
p.
Finally, splitting profit between platform sponsor and developers, we have

⇡p = V V + 12 p̃ỹ1 + 1
2
p̃ỹ2
= V (1 ) + 12 1/Nd pNd y1 + 1
2
1/Nd pNd↵+1 y2
(2)
= V (1 ) + 12 v(1 )y1 + 1
2
v(1 )Nd↵ y2
2
= V (1 ) + 12 v(1 )k( V )↵ + 1
2
v(1 )k 1+↵ Nd↵ ( V )↵ ,
6
See, e.g., Tirole (1988), p. 220.

10
which confirms Eq. (1).

3 Model Analysis

To facilitate the analysis below, we introduce a generic parameter and re-write platform
sponsor profit as

1 1 2
⇡p = V (1 ) + v(1 )k( V )↵ + v(1 )k 1+↵ ( V )↵ . (3)
2 2

Eqn (1) is recovered when is substituted back as D = Nd↵ . Conceptually, we can interpret
as a measure of the spillover e↵ect. We will later see in this and the next section that
platforms exposed to developer technical risk and platforms subject to network e↵ects would
experience di↵erent levels of spillover, say T R, or NE. Nonetheless, the dependency of the
0
platform’s choice of and on the s is directionally the same.

3.1 Optimization under Competition and Developer Risk

Consider the situation where individual developers face the risk of technical failure (with
probability ⇢). Given failure in earlier rounds, the number of developers in the ecosystem
may decline over time. Assume the number of developers on Stage 1 is Nd and denote the
number of developers who survive to Stage 2 as n. n follows a Binomial distribution with
parameter Nd and ⇢. The mean of the total value generated in Stage 1 remains y1 = ( V )↵ ,
and in Stage 2, it becomes E(n↵ )y2 .
Consequently, the corresponding platform profit can be expressed in terms of the generic
form Eqn (3) with T R (Nd , ⇢) = E(n↵ ). For the special case ⇢ = 0, n = Nd , and T R (Nd , 0) =

D = (Nd )↵ . For the general case 0 < ⇢ < 1, it is clear that 0 < T R (Nd , ⇢) < (Nd )↵ , and
it monotonically increases in Nd and in ! = 1 ⇢. Denoting Nr := [E(n↵ )]1/↵ , we have

11
 T R (Nd , ⇢) = Nr↵ , and the mean value of platform profit under technical risk, still denote by
⇡p , can be written as

1 1 2
⇡p = V (1 ) + v(1 )k( V )↵ + T R (Nd , ⇢) v(1 )k 1+↵ ( V )↵ , (4)
2 2
1 1 2
= V (1 ) + v(1 )k( V )↵ + (Nr )↵ v(1 )k 1+↵ ( V )↵ . (5)
2 2

In other words, the mean profit of a risk-prone platform system is identical to that of a
no-risk system (Eqn. (1)) with the number of developers at a reduced level of Nr instead
⇣ ⌘ p ↵
1 ↵ 1 2 ↵
of Nd . Further, define R := (↵v)/(2Nr ), U := Nr k /V , and := 2
< 1/2. We
have the following characterization of the optimal interior solution ( ⇤ , ⇤
). Denote Nd as
the unique Nr that induces optimal at .

Proposition 1 Given k, v, V, Nd > 0, 0 < ↵ < 1, 0  ⇢ < 1. Under condition R, U < 1,

the optimum ( ⇤ , ⇤
) of the risk-prone platform system corresponds to the optimum in the
no-risk platform system with an equivalent number of developer Nr < Nd . Consequently, the
following results hold true.
⇤
(i) An interior optimum 2 (0, ] uniquely exists.
⇤
(ii) An interior optimum 0 < < 1 uniquely exists.
(iii) Optimum increases monotonically in Nr . Consequently, it also increases in Nd
and in ! = 1 ⇢.
(iv) monotonically increases in when  and in Nr (Nd , !) when Nr  Nd ; and
monotonically decreases in when and in Nr (Nd , !) when Nd Nd .

Proof. Please see Appendix

To interpret the result, we first note that, Nr < Nd suggests that risk damps the spillover
e↵ect, consequently lowering the level of incentive for bundling. In the meantime, an in-
creased success rate encourages earlier bundling. In the case that the bundling durations are

12
too long for reasons not modeled (such as fundamental change in the industry), then enlisting
additional developers could compensate for technical risk and then maintain (or speed up)
the bundling process. The result that having more developers helps mitigate risk is consistent
with both logic and empirical research that finds handheld device platforms opened to more
developers precisely to reduce the risk of technological innovation (Boudreau, 2010). For the
same reason, social network platforms encourage developers to experiment because “much
remains unknown concerning preferences and technical approaches to social applications”
(Boudreau and Hagiu, 2009, p. 11).
The finding that first increases in N and then decreases after a threshold level has
an important antecedent in the literature. Laursen and Salter (2006, 2014) find a concave
relationship between appropriability and openness. To join their context and ours, additional
developers can be interpreted as increasing collaboration. An interesting case arises when
there is friction in adjusting . In this situation, the firm might anticipate growth in the
developer base at launch and set a non-optimal and then hope that platform adoption
would make the choice better over time. This parallels platform launch as described in
Ondrus et al. (2015).

3.2 Platform Competition

We now analyze competition among platforms. Continuing with the Green and Scotchmer
(1995) approach, we let competition moderate platform pricing power in the same way that
V
it moderates developer pricing power, reducing platform price from V to Np
where Np 1 is
the number of platform competitors. Note from the proof of Proposition 1 that the optimum
⇤ ⇤
is independent to V , and that depends on V . We conclude that increasing Np implies
⇤ ⇤ ⇤
a linear dependence Np := p until p hits 1, complete openness, meaning the platform
sponsor would give away 100% of its original value. We summarize this discussion formally.

13
Corollary 1 Increasing the intensity of platform competition has no e↵ect on t⇤ , but pro-
⇤ ⇤
portionally increases to min{Np , 1}.
V
Proof. The claims follow from Proposition 1 by substituting Np
for V .

Holding all else constant, greater platform competition reduces the sponsor’s direct plat-
form profits. The sponsor’s incentive is therefore to open the platform in order to increase
indirect profits from developer innovation. In terms of competition policy, the social planner
should promote platform competition, which motivates sponsors to open up platforms and
seek growth. This result directly parallels empirical findings. Based on case studies of IBM,
Sun Microsystems, and Apple, West (2003) concluded that sponsors prefer the higher rents
from closing their systems unless their platforms face pressure from rival platforms.

4 Permissionless versus Negotiated Platform Access

To this point, our analysis has assumed a platform with default contracts that allow develop-
ers to innovate on top. In this setting, it is always strategically optimal to open it up ( > 0)
at least some degree. However, a more fundamental question remains to be answered: is
opening the platform to all developers the best way to organize for innovation? Might not
vertical integration, which implies closing o↵ developer access and only working with nego-
tiated parternships, be better? The two-stage platform model suggests that one mechanism,
the network spillover e↵ect, could help open innovation to gain competitive advantage. In
fact, the spillover e↵ect induced by multiple developers Nd scales up profit in the second
stage. Below, we will examine how a network e↵ect scales up the ecosystem by mobilizing
both developers and users. The generic form Eqn (3) facilitates the analysis through analysis
of the spillover factor NE that is implied by a network e↵ect.

14
4.1 Negotiated Platform Access

Numerous mergers and acquisitions are predicated on the theory that a rational firm could
improve profits by buying developers to acquire their technology. By only allowing negotiated
platform access in our model, the sponsor might gain three advantages over permissionless
innovation. First, closing the platform saves the open innovation subsidy V , which increases
profits from direct platform sales. Second, the developer can build on the entire platform,
not just the portion opened; so output rises from y( V ) to y(V ). Third, application prices
rise to monopoly levels p = v because users cannot acquire apps by waiting for them to
becomes a public good. Thus app profits rise. Allowing developers to keep half the value of
their technology, based on Nash bargaining, the platform’s profit under vertical integration
becomes ⇡vi = V (1 )| =0 +y1 | =1 +y2 | =1 which simplifies to

1 1 2
⇡vi = V + vkV ↵ + vk 1+↵ V ↵ . (6)
2 2

The corresponding platform system with Nd = 1 is listed below for comparison.

1 1 2
⇡p = V (1 ) + v(1 )k( V )↵ + v(1 )k 1+↵ ( V )↵ . (7)
2 2

Apparently, vertical integration (Eqn (6)) yields higher profit than open platform with one
developer (Eqn (7)). It has higher output; it has no subsidy cost; and it has higher prices.7
We then ask how might profits from open innovation ever dominate those from vertical
integration? Following Eqn (3) and viewing the platform profit ⇡p as an increasing function
of , denoted as ⇡p ( ), we conclude
7
Model analysis can easily extend to subcontracting, an organizational form between vertical integration
and open innovation, by choosing di↵erent levels of .

15
Proposition 2 Vertical integration outperforms an open platform when there is no spillover
which reduces the spillover multiplier to = 1. However, there is a unique breakeven spillover
parameter such that with all other parameters remaining constant, the open platform system
outperforms vertical integration if and only if .

Proof. Notice that the profit in the second stage is the only item depending on ; more
specifically, proportionally increases in , we claim that increasing the spillover e↵ect
would improve the profit of open system ⇡p ( ), which ultimately outperforms the vertically-
integrated system ⇡vi .
We posit two distinct answers to the question of whether to negotiate access or allow
“permissionless innovation” (Cerf, 2012). One is that there exist developers the sponsor
does not know and therefore cannot acquire. The other is that network e↵ects can increase
disproportionately under openness. The former might arise if there are numerous small
developers who participate if they see an opportunity. Permissionless innovation matters to
developers who risk disclosing their novel ideas by identifying themselves or their applications
to the platform sponsor. Owning the indispensable asset, the sponsor has bargaining power
and needs only the ideas to steal them (Bessen and Maskin, 2009; Parker and Van Alstyne,
2000a, 2012). Commitment to stay out of the developer’s market during the exclusionary
period provides the incentive such developers need to step forward. The law literature
(Eisenberg, 1976) notes that making such a commitment will a↵ect the downstream conduct
of other parties whenever the mere act of negotiating reveals sensitive information. This
result is clearly in evidence in SAP’s platform, for example, where, as noted above, the
platform sponsor commits to stay out of “white spaces,” functionality that anyone is free to
develop, for minimum periods of 18–24 months.
The second answer arises because, relative to closed systems, open systems invite more
third party participation. Mechanisms by which openness might increase participation in-

16
clude transparency, bug reporting and feedback that can reduce R&D costs and increase
platform quality, and user ability to modify open systems (Chesbrough, 2003; West, 2003).
Openness can reduce negotiation costs, facilitate free redistribution (Raymond, 1999), and
serve as a low price commitment analogous to second sourcing (Farrell and Gallini, 1988).
It can aid horizontal integration (Farrell et al., 1998). The “two-sided” network literature
(Parker and Van Alstyne, 2000a, 2005; Rochet and Tirole, 2003) specifically demonstrates
how openly subsidizing one community (i.e., developers), can increase value to and participa-
tion of another community (i.e., end-users). For a variety of reasons, openness can increase
both value and participation.
As both answers rely on growing the platform microeconomy, we now modify the earlier
open platform model to include classic two-sided network e↵ects across consumers and devel-
opers who value one another’s participation on the platform (e.g., Parker and Van Alstyne
(2005)). For tractability, we develop a novel yet simplified version of two-sided network e↵ects
to understand how their strength a↵ects a sponsor’s choice to provide access to all developers
versus working with a select few. Thus we introduce market multiplier Mi , i 2 (u, d) derived
from two-sided market feedback in order to represent the sizes of spillover externalities from
content creation and content consumption.

4.2 Network E↵ect

While advantages of vertical integration include eliminating the subsidy, increasing prices,
and increasing output, the advantage of open innovation is growing the market. Higher
adoption and network e↵ects can then justify open innovation relative to vertical integration.
To gain some insight about network e↵ects which we label Mi , consider the following
mechanism that allows more users to attract more developers and more developers to at-
tract more users. Based on an externality spillover eud , augment baseline developers Nd

17
proportional to the number of users Nu , increasing developers by eud Nu . Likewise, based on
externality spillover edu , augment baseline users Nu proportional to the number of developers
Nd , increasing users by edu Nd . New users attract additional new developers, and vice versa,
in amounts edu eud Nu and eud edu Nd , a recursion process that defines Cauchy sequences for
both groups. Developer size increases according to Nd (1 + eud edu + (eud edu )2 + (eud edu )3 + ...)
and similarly for users. To keep market size finite, impose the constraint edu eud < 1. These
1 edu
sequences converge to Nd Md = Nd 1 eud edu
and Nu Mu = Nu 1 eud edu
respectively.
We can now present the platform profit function increased by network e↵ects as follows.

⇣ 1 1 ⌘
↵ 1+↵ ↵ ↵2
⇡open = Mu V (1 ) + v(1 )k( V ) + v(1 )k (Nd Md ) ( V ) . (8)
2 2

E↵ectively, the network e↵ect has two impacts: (1) the spillover e↵ect at the level of

NE := (Nd Md )↵ , (2) re-scaling of the total platform profit proportional to the scale of the
user population Mu . These two impacts jointly work to make open permissionless innovation
outperform negotiated access-vertical integration. The following results parallel those of
Proposition 2.

Proposition 3 For the platform system with network e↵ects, there exists a monotonically
decreasing threshold curve N Md (Mu ) such that for any given user-side multiplier Mu , the
open platform system outperforms vertical integration if and only if Nd Md N Md .

Proof. By observing that ⇡open /Mu in Eqn (8) is the generic ⇡p in Eqn (3) with =
(Nd Md )↵ , the proof follows Proposition 2 for all Mu > 0.
In the absence of network e↵ects, the platform sponsor should own all means of produc-
tion. Opening the platform to outside developers, however, becomes more attractive (i) as
network e↵ects rise (or the sizes of user or developer pools grow), (ii) as developer output
rises, and (iii) content becomes more reusable. Negotiated access/vertical integration be-
comes more attractive as platform value itself grows. Note that the decentralized innovation

18
is achieved without bargaining costs. A default contract with h > 0, t > 0i gives developers
an option to enter the market without disclosure to the platform sponsor. Open innova-
tion, with a guarantee of lead time, preserves the information asymmetry that protects the
innovator and prevents a powerful monopoly platform from stealing the full value of the inno-
vation. The importance of third party contributions also becomes clearer as we observe that
price e↵ects, which are one-time gains, yield lower returns than production e↵ects, which are
recursive gains.

5 Discussion & Conclusions

Using a formal model of sequential innovation with code spillovers, our contribution is to
show why developers might cause a shift in organizational form and to provide a theory of how
platform firms optimize their own intellectual property regimes in order to maximize growth.
This extends the sequential innovation literature (Green and Scotchmer, 1995; Chang, 1995;
Parker and Van Alstyne, 2015) to add developer competition, spillovers, and network e↵ects.
From this baseline, we add three new results on optimal openness ( ), optimal IP duration
(t), and e↵ects of competition among developers (Nd ).

1. We show how a rising number of developers can invert the firm. That is, firms will
choose to innovate using open external contracts in preference to closed vertical in-
tegration or subcontracts. The locus of value creation moves from inside the firm to
outside. Distinct from physical goods, digital goods a↵ord firms the chance to optimize
spillovers. If the number of developers is small or network e↵ects are modest, firms
prefer vertical integration or closed sub-contracts. However, once a threshold developer
base Nd is reached, firms prefer to o↵er an open default contract to any developer who
wishes to build upon the platform. Distinct from traditional buyer-supplier networks,
this attracts resources from third parties that the platform firm does not even know.

19
 2. Competition has di↵erent implications depending upon whether it occurs between plat-
forms or among developers. Platform-to-platform competition strictly increases the
level of openness. Because the platform makes less direct profit, openness costs less and
the firm prefers to subsidize developer spillovers. By contrast, developer-to-developer
competition has a non-monotonic e↵ect. Openness first rises in the number of devel-
opers to promote R&D spillovers but then falls due to developer price competition.
The result that openness has an invert-U relationship to innovation is consistent with
empirical literature (Boudreau, 2010; Laursen and Salter, 2014).

3. Firms that pursue high risk innovations with more developers can be more profitable
than firms that pursue low risk innovations with fewer developers. More developers
give platform firms more chances at success. Platforms will increase their openness
( ) as the likelihood of technical success increases until a threshold level of Nd after
which the platform reduces openness. Further, the more developers, the shorter is the
proprietary period (t) o↵ered under an optimal IP policy. The reason is that more
developers increase the value of spillovers. Thus it makes sense to subsidize N-1 other
developers by making the code of a given developer public in addition to the open code
of the platform itself.

5.1 Limitations and Research Implications

The model analyzed above requires a number of assumptions for tractability. Key among
these are (1) a shared consumer value for the platform V and developer additions, v, (2) use
of a Cobb-Douglas formulation for developer output, and (3) no developer entry between
periods one and two. A shared value is clearly a simplification since consumer valuations are
more likely to follow an exponential distribution. However, Bakos and Brynjolfsson (1998)
observe that, based on the central limit theorem, the average value converges rapidly for

20
any bounded distribution as the population of consumers grows or the number of items in a
bundle grows.
The practical impact of the point-mass valuation is likely to be limited because the focus
of the analysis is on the behavior of the platform with respect to technical risk, the number
of developers, and network e↵ects. If the markets were made up of heterogeneous consumers,
then some consumers would be priced out and the served market would be smaller. It would
be interesting to examine the impact of the distribution of consumer values on the choices
that platforms make. We leave that analysis to future researchers who might need to use
simulation tools given the complexity of the analysis.
The Cobb-Douglas output assumption was also made in order that the model be solvable
using formal analytic techniques. However, the formulation is widely used and, importantly,
enjoys empirical support in work by Eilhard and Ménière (2009), who found the translog
generalization fits the data for open source projects in real applications.
Our model of technical risk assumes that developers who fail are assumed to exit. If
developers were to enter between period one and two, the impact on the platform’s choices
of and t should be the same as if the number of developers were simply larger in period
one. However, the issue of when and why developers join platforms merits further study.
More generally, we note that results in the sequential innovation literature have not, until
now, begun to account for recursive R&D spillovers that are feasible with digital goods. By
not accounting for digital reuse, prior literature has recommended a patent period that is
too long (Gilbert and Shapiro, 1990; Green and Scotchmer, 1995; Chang, 1995). Further, the
prior works do not account for the variety of possible subsequent uses, which conditions and
shortens these baseline predictions. Each of the results derived from the analysis presented
above can also be formulated as a testable hypothesis, which we hope to explore in follow-on
empirical research.

21
5.2 Managerial and policy implications

There are several managerial and policy conclusions. First, consider a classic M&A policy
that would recommend acquiring complementary assets. We show that this can be subop-
timal in a platform context. More precisely, we show that “permissionless innovation” can
dominate vertical integration in cases where the number of developers becomes large because
openness promotes R&D spillovers, which do not occur when the firm internalizes all pro-
duction. Moreover, the platform owner does not always know which developers will succeed
in the market and therefore which assets to acquire. This result implies that a platform
strategy has a longer term likelihood of success than a purchasing/subcontracting strategy
so long as the developer base reaches a sufficient size. This inverts the firm; the platform can
wait longer to observe and profit from external developers before (if ever) acquiring them.
Second, as noted after Proposition 1, suppose that changing the level of openness is not
frictionless. This might be the case due to technological or cultural constraints. Then firms
prefer to set openness higher than is initially optimal whenever they anticipate growth of
the developer base. Firms can also restrict network growth beyond a given size in order to
control for poor developers.
Third, the core platform value can become so great that the owner no longer needs to
give it away to stimulate growth. In that case the owner prefers to monetize the platform by
reasserting control and further limiting openness. We can observe this in practice based on
two examples. MakerBot, a 3D printing company, that gave away designs as well as design
specs, was more open at first but has since moved to a more closed strategy by pursuing
patents and creating applications that are not governed by open source licenses (Brown,
2012). Similarly, Android was more open at first in order to foster growth but has since
moved to close the platform by exerting control over application programming interfaces
(APIs) and critical applications such as Google maps (Amadeo, 2013).

22
 Fourth, the manager’s role must change in a platform context. Rather than optimizing
the profits of the firm’s own product in isolation, the manager needs to optimize the value
created by an ecosystem, even if it cannibalizes a core product. In the context of large
numbers of developers, this “platform” model is more profitable in the long run. In a
real world context, considering developer profit is consistent with the policies of SalesForce
and SAP as these firms specifically measure themselves by how much value they create for
ecosystem partners. Indeed, addressing the cannibalization problem via a platform strategy
is a significant topic addressed in the book Platform Leadership (Gawer and Cusumano,
2002). Here we show this result formally and prove that it can be more profitable.
Fifth, managing the ecosystem can also be interpreted as setting the optimal growth
policy for an intellectual property (IP) regime where code may be reused. This is a unique
property of digital and information goods. The period (t) during which a developer can sell
an app parallels the duration of patent protection. The end of this period parallels when
developer code becomes “public” and useful for R&D spillovers that help other developers.
How the reuse of digital code a↵ects optimal IP policy, and in turn is a↵ected by competition,
is not analyzed in other literature.

6 Appendix

Proof of Proposition 1
Recall from the model setup (Eqn. (5)) that
1 1 2
⇡p = V (1 ) + v(1 )k( V )↵ + v(1 )k 1+↵ Nr↵ ( V )↵ . (9)
2 2
The corresponding first-order condition w.r.t. and becomes
@⇡p 1 ↵ 1 ↵2 1 2
0= = V + v(1 )[k↵ V ↵ + ↵2 k 1+↵ Nr↵ V↵ ] (10)
@ 2
@⇡p 1 1 2 1 2
0= = vk( V )↵ + v(1 )k 1+↵ ( V )↵ vk 1+↵ Nr↵ ( V )↵ . (11)
@ 2 2 2
23
 Var Parameter Definition
– Share of platform (%) opened to developers
t, – Time until exclusionary period expires (discount = e rt )
↵ – Technology in Cobb Douglas production
k – Coefficient of reuse
Md , Mu – Market spillovers from developers & users, index sizes of network e↵ects
Nd , Nu – Numbers of developers and users respectively
p – Price of individual developer applications p = v(1 )
⇢ – Technological uncertainty; equal to 1 !
v – Value, per unit, of developer output
V – Standalone value of sponsor’s platform
yi – Output of a single developer in period i and input to developers
in period i + 1 with yi = kyi↵ 1 and y0 = S
! – Probability of success for a given innovation; equal to 1 ⇢

Table 1: Parameter Definitions

Multiply Eqn. (10) by to get

1 2
0= V + k↵v(1 )[( V )↵ + ↵k ↵ Nr↵ ( V )↵ ]. (12)
2
Denote
S := V. (13)
↵2
Then, Eqn. (12) becomes S = 12 k↵v(1 )[S ↵ + ↵k ↵ Nr↵ S ], or
v
1 = (k↵S ↵ 1 ) (1 )[1 + ↵(kNr S ↵ 1 )↵ ]. (14)
2
1 2 2
Similarly, Eqn. (11) becomes 0 = 2
vkS ↵ + 12 v(1 )k 1+↵ S ↵ 1
2
vk 1+↵ Nr↵ S ↵ . Equivalently,
1
= [1 (kNr S ↵ 1 ) ↵
]. (15)
2
Let

M := kS ↵ 1 . (16)
Then
1 1
= (1 ) (17)
2 (Nr M )↵
Then Eqns (14,15) become
↵v
1 = (1 )M [1 + ↵(Nr M )↵ ] (18)
2
1 1
= [1 ]. (19)
2 (Nr M )↵

24
 Substituting (19) into (18), we obtain
↵v ↵
⇥ 1 ↵
⇤
1= 1 + (Nr M ) M 1 + (1 (Nr M ) )↵(Nr M )↵ . (20)
4 2
Eqn. (20) serves as the basis for our analysis of and .
First, about as claimed in (i). Denote

X := Nr M, (21)

and view the right-hand side of (20) as a function of X and Nr , f (X, Nr ), i.e.,
↵v ↵
⇥ 1 ↵ ↵
⇤
1 = 4N 1 + X X ⇥ 1 + 2
(1 X )↵X⇤
↵v
r
1 ↵ 1 1 ↵ (22)
= 4N r
X + X (1 2
↵) + 2
↵X := f (X; Nr ).

Recall 0 < ↵ < 1, which implies 1 ↵ > 0 and 1 ↵/2 > 0. Therefore, all the terms in
the expression of f (X; Nr ) are both positive and monotonically nondecreasing. We have the
follow properties of f (X; Nr ):
(1) f (0; Nr ) = 0; f (1; Nr ) = 1 for all v, Nr > 0.
(2) f (X; Nr ) increases strictly in X and decreases strictly in Nr .
Consequently, there exists a unique X(Nr ) > 0 such that f (X(Nr ); Nr ) = 1. Clearly
X(Nr ) monotonically increases in Nr due to the monotonicity of f (X; Nr ) w.r.t X and Nr .
By further expressing in terms of X via (19) and (21), = 12 [1 X ↵ ], we see increases
in X, thus in Nr . Moreover, the natural bound for an interior > 0 requires X > 1,
which is equivalent to f (1; Nr ) < 1 due to the monotonicity of f in X. By straightforward
rearrangement, f (1; Nr ) < 1 becomes Condition R = (↵v/2)/Nr < 1. This completes the
proof of Part (i).
Now, consider . The uniqueness of X > 1 satisfying f (X; Nr ) = 1 implies the uniqueness
of . Indeed, by definitions of X, M, and S, we have X = Nr M = Nr kS ↵ 1 = Nr k( V )↵ 1 .
Under condition Nr , k > 0 and X > 0 according to argument above, we have
⇣ N k ⌘ 1 1↵
r
= /V > 0. (23)
X
Therefore, it is never optimal for the platform to be completely close, = 0 as long as
v, Nr , k > 0. We now demonstrate the monotonicity property of with respect to Nr , or
equivalently, to , to complete the proof of Part (ii).
Noticing (23) can be re-written as
⇣ k ⌘ 1 1↵
= /V, (24)
M
we now convert f (X; Nr ) into a function of M and Nd . To be more precise, for

Q := Nr↵ , (25)

25
define function
↵v ⇥ 1 1 ⇤
g(M ; Q) := f (X; Nr ) = M + M1 ↵
/Q (1 ↵) + ↵QM ↵ . (26)
4 2 2
Parallel to previous arguments, we have g(0; Q) = 0, g(1; Q) = 1; thus, for all Q > 0,
there exists a unique M (Q) > 0 such that g(M (Q); Q) = 1. The monotonicity property of
M (Q) w.r.t. Q is thus implied in the monotonicity of g(M ; Q) w.r.t. both M and Q.
As for the monotonicity of g, it is clear g(M ; Q) increases strictly in M . With respect to
Q, consider the first-order partial derivative
@g
⇥ ⇤
@Q
= ↵v 4
M 1 ↵ /Q2 (1 12 ↵) + 12 ↵QM ↵ + ↵v 4
M + M 1 ↵ /Q 12 ↵M ↵
= ↵v 4
1 ↵ 2
h M 1/Q (1 2 ↵)
1
+ ↵v 1
4 2
↵M 1+↵ (27)
↵vM (1 2 ↵) 1 ↵
⇤
= 4 M ↵ Q2
+ 2 ↵M .

Clearly, n o
@g
@Q
>0 () (QM ↵ )2 > 1 ↵/2
↵/2
q
() (QM ↵ ) > 1 ↵/2 ↵/2
q (28)
() (1 2 ) 1 > 1 ↵/2 ↵/2
[by (19)]
p ↵
1 2 ↵
() < 2
=
Combining equations f (X, Nr ) = 1 and = 12 (1 X1↵ ), uniquely determines an Nd . The
monotonicity of w.r.t. Nr in Part (i) further yields
n @g o
> 0 () Nr < Nd . (29)
@Q

Therefore, we conclude on {Nr < Nd } or { < },

g(M ; Q) = 1 =) M & Q[g increases in M and in Q]

() % Q [by (24)]
(30)
() % Nr [by (25)]
() % [monotonicity of w.r.t. Nr in Part (i)].

In parallel, on {Nr Nd } or { }, & , Nr . Consequently, achieves its maximum at

= , Nr = Nd . This completes the proof of Part (iii).
By combining Eqns (19, 21, 23) under condition R < 1, we can further express as
function of . ⇣ ⌘ 1
1/↵ 1 ↵
= Nr k 1 2 /V (31)
⇣ 1
Clearly, < 1 is guaranteed by Nr k) 1 ↵ /V < 1, or equivalently, (Nr k)/V 1 ↵ = U < 1.
This confirms Part (ii) of the proposition. Finally, it is easy to see Nr monotonically increases
in Nd and ! = 1 ⇢, and the proof is complete.

26
References
Amadeo, R. 2013. “Google’s iron grip on Android: Controlling open source by any means
necessary,” .
URL http://arstechnica.com/gadgets/2013/10/googles-iron-grip-on-android-controlling-o

Bakos, Y., and Brynjolfsson, E. 1998. “Bundling information goods: Pricing, profits and
efficiency,” Management Science (45:12), pp. 1613–1630.

Baldwin, C., and Clark, K. 2000. Design rules: The power of modularity, vol. 1, The MIT
Press.

Bessen, J., and Maskin, E. 2009. “Sequential innovation, patents, and imitation,” The RAND
Journal of Economics (40:4), pp. 611–635.

Bhargava, H., Kim, B., and Sun, D. 2012. “Commercialization of Platform Technologies:
Launch Timing and Versioning Strategy,” Production and Operations Research pp. 1–15.

Boudreau, K. 2010. “Open Platform Strategies and Innovation: Granting Access versus
Devolving Control,” Management Science (56:10), pp. 1849–1872.

Boudreau, K., and Hagiu, A. 2009. Platforms, Markets and Innovation, Edward Elgar Pub-
lishing Limited, Cheltenham, UK, chap. Platform rules: Multi-sided platforms as regula-
tors, pp. 163–189.

Bresnahan, T., and Greenstein, S. 1999. “Technological competition and the structure of the
computer industry,” The Journal of Industrial Economics (47:1), pp. 1–40.

Brown, R. 2012. “Pulling back from open source hardware, MakerBot angers some adher-
ents,” .
URL http://www.cnet.com/news/pulling-back-from-open-source-hardware-makerbot-angers-s

Brynjolfsson, E., and McAfee, A. 2014. The second machine age: Work, progress, and pros-
perity in a time of brilliant technologies, WW Norton &amp; Company.

Caillaud, B., and Jullien, B. 2003. “Chicken & egg: Competition among intermediation
service providers,” RAND Journal of Economics (34:2), pp. 309–328.

Cerf, V. 2012. “Remarks at the Digital Broadband Migration: The Dynamics of Disruptive
Innovation: Internet Speculations,” J on Telecomm & High Tech L (10), p. 21.

Chang, H. 1995. “Patent scope, antitrust policy, and cumulative innovation,” The RAND
Journal of Economics (26:1), pp. 34–57.

Chesbrough, H. 2003. Open innovation: The new imperative for creating and profiting from
technology, Harvard Business Press.

27
Cusumano, M. 2010. “Technology strategy and management The evolution of platform think-
ing,” Communications of the ACM (53:1), pp. 32–34.

Eaton, B., Elaluf-Calderwood, S., Sørensen, C., and Yoo, Y. 2015. “Distributed Tuning of
Boundary Resources: The Case of Apple s iOS Service System,” MIS Quarterly (39:1),
pp. 217–244.

Eilhard, J., and Ménière, Y. 2009. “A Look Inside the Forge: Developer Productivity and
Spillovers in Open Source Projects,” Available at SSRN 1316772 .

Eisenberg, M. 1976. “Private Ordering Through Negotiation: Dispute-Settlement and Rule-

making,” Harvard Law Review pp. 637–681.

Eisenmann, T., Parker, G., and Van Alstyne, M. 2006. “Strategies for two-sided markets,”
Harvard Business Review (84:10), pp. 92–101.

Eisenmann, T., Parker, G., and Van Alstyne, M. 2009. Opening platforms: How, when and
why? , Edward Elgar Publishing Limited, Cheltenham, UK, chap. 6, pp. 131–162.

Evans, D., Hagiu, A., and Schmalensee, R. 2006. Invisible engines: how software platforms
drive innovation and transform industries, The MIT Press.

Evans, P. C., and Annunziata, M. 2012. “Industrial Internet: Pushing the Boundaries of
Minds and Machine,” General Electric .

Farrell, J., and Gallini, N. 1988. “Second-sourcing as a commitment: Monopoly incentives

to attract competition,” The Quarterly Journal of Economics (103:4), p. 673.

Farrell, J., Monroe, H., and Saloner, G. 1998. “The vertical organization of industry: Sys-
tems competition versus component competition,” Journal of Economics & Management
Strategy (7:2), pp. 143–182.

Fine, C. 1999. Clockspeed: Winning industry control in the age of temporary advantage,
Basic Books.

Gawer, A., and Cusumano, M. 2002. Platform leadership: How Intel, Microsoft, and Cisco
drive industry innovation, Harvard Business Press.

Gawer, A., and Cusumano, M. 2008. “How companies become platform leaders,” MIT Sloan
management review (49:2), pp. 28–35.

Gilbert, R., and Shapiro, C. 1990. “Optimal patent length and breadth,” The RAND Journal
of Economics (21:1), pp. 106–112.

Green, J., and Scotchmer, S. 1995. “On the division of profit in sequential innovation,” The
RAND Journal of Economics (26:1), pp. 20–33.

28
Grove, A. 1996. Only the paranoid survive, Currency Doubleday.

Huang, P., Ceccagnoli, M., Forman, C., and Wu, D. 2012. “Appropriability Mechanisms and
the Platform Partnership Decision: Evidence from Enterprise Software,” Management
Science ((Forthcoming)).

Kallinikos, J., Aaltonen, A., and Marton, A. 2013. “The ambivalent ontology of digital
artifacts,” MIS Quarterly (37:2), pp. 357–370.

Laursen, K., and Salter, A. 2006. “Open for innovation: the role of openness in explaining
innovation performance among UK manufacturing firms,” Strategic management journal
(27:2), pp. 131–150.

Laursen, K., and Salter, A. J. 2014. “The paradox of openness: Appropriability, external
search and collaboration,” Research Policy (43:5), pp. 867–878.

Libert, B., Wind, Y. J., and Fenley, M. B. 2014. “What Airbnb, Uber, and Alibaba Have in
Common,” Harvard Business Review .

Markovich, S., and Moenius, J. 2009. “Winning while losing: Competition dynamics in
the presence of indirect network e↵ects,” International Journal of Industrial Organization
(27:3), pp. 346–357.

McAllister, N. 2011. “Google empowers hardware hackers with the Android ADK,” .

Ondrus, J., Gannamaneni, A., and Lyytinen, K. 2015. “The impact of openness on the
market potential of multi-sided platforms: a case study of mobile payment platforms,”
Journal of Information Technology pp. 1–16.

Parker, G., and Van Alstyne, M. 2000a. “Information complements, substitutes, and strategic
product design,” in Proceedings of the twenty first international conference on Information
systems, , Association for Information Systems.

Parker, G., and Van Alstyne, M. 2000b. “Internetwork Externalities and Free Information
Goods,” in Proceedings of the Second ACM conference on Electronic Commerce, , Associ-
ation for Computing Machinery.

Parker, G., and Van Alstyne, M. 2005. “Two-sided network e↵ects: A theory of information
product design,” Management Science (51:10), pp. 1494–1504.

Parker, G., and Van Alstyne, M. 2012. “A Digital Postal Platform: Definitions and a
Roadmap,” Tech. rep., MIT Center for Digital Business.

Parker, G., and Van Alstyne, M. 2014. “Platform Strategy,” Palgrave Encyclopedia of Strate-
gic Management .

29
Parker, G., Van Alstyne, M., and Choudary, S. 2016. Platform Revolution, WW Norton and
Company.

Parker, G., and Van Alstyne, M. W. 2015. “Innovation, Openness, and Platform Control,”
SSRN Working Paper .

Raymond, E. 1999. “The cathedral and the bazaar,” Knowledge, Technology & Policy (12:3),
pp. 23–49.

Rochet, J., and Tirole, J. 2003. “Platform competition in two-sided markets,” Journal of the
European Economic Association (1:4), pp. 990–1029.

Rysman, M. 2009. “The economics of two-sided markets,” The Journal of Economic Per-
spectives (23:3), pp. 125–143.

Scholten, S., and Scholten, U. 2011. “Platform-based Innovation Management: Directing

External Innovational E↵orts in Platform Ecosystems,” Tech. rep., SAP Research and
Karlsruhe Institute of Technology.

Shapiro, C., and Varian, H. 1999. Information rules: a strategic guide to the network econ-
omy, Harvard Business Press.

Slater, D., and Schoen, S. 2006. “Who Killed TiVo to Go?” Tech. rep., Electronic Frontier
Foundation, w2.e↵.org/IP/pnp/cablewp.php.

Tirole, J. 1988. The Theory of Industrial Organization, MIT press.

Tiwana, A. 2013. Platform Ecosystems: Aligning Architecture, Governance, and Strategy,

Newnes.

West, J. 2003. “How open is open enough? Melding proprietary and open source platform
strategies,” Research Policy (32:7), pp. 1259–1285.

Yoo, Y., Henfridsson, O., and Lyytinen, K. 2010. “The New Organizing Logic of Digital
Innovation: An Agenda for Information Systems Research,” Information Systems Research
(21:4), pp. 724–735.

Zhu, F., and Iansiti, M. 2012. “Entry into platform-based markets,” Strategic Management
Journal (33:1), pp. 88–106.

30