1. Introduction

Numerous high-tech products such as personal computers (PCs)

possess especially complex structures because they are a combination

of components built on platforms. When a product is functionally

interdependent with a majority of the other components of a system,

and the end-user demands the overall system, it may be termed a

“platform” [17]. The platform's market structure is determined by

an innovator's decision regarding the commercializing strategy. An

innovator may possess proprietary control over the entire system in

a vertically integrated production structure, or a monopoly over a

limited proprietary part, or only be a brand-name platform producer,

who integrates the components of the platform that are supplied by

third parties. This presents the intrinsic coordination question regarding

the commercialization of a platform [28]. Two historical examples

from the PC industry present rather different viewpoints regarding

this problem.

Apple Computers has been producing highly integrated PCs and

controlling the proprietary rights over its products since the 1970s.

Generally, the performance of highly integrated platform products is

expected to be superior. However, Apple decided not to establish a

large PC network and therefore did not establish interconnections

with others. Instead, it has been focusing on the development of

mania groups using fancy products. However, this niche strategy is

inherently dangerous in markets with strong network effects [34].

In contrast, by employing an open architecture strategy, IBM offers a

variety of IBM compatible PCs, thereby fulfilling the demand of numerous

customers. Since a majority of the personal computers sold

were IBM-compatible, IBM was recognized as a platform owner in

the market, and others identified their brands as IBM-compatible.

Apple Computers, who was the market leader in the 1970s, lost

their market share to IBM-compatibles; therefore, IBM became the

platform owner in the PC industry in the early 1980s. The competition

between these two extreme marketing strategies indicates that the

network effect is a critical factor that must be considered when a

firm determines a commercialization strategy for its platform product.

Clones were deliberately invited into the incumbent market in

order to maximize the benefit of the network effects [10]. The case

of IBM emphasizes that until the incumbent continues to produce

products of a quality that is superior than those produced by its

clones, the incumbent may be at an advantage when it acts as a monopolist

by protecting its technology. This is because the increased

user-base enables incumbents to enhance their profits by charging

high-value consumers a high price.

However, once IBM's open architecture began permitting numerous

manufacturers of IBM-compatible computers to produce PCs

whose quality was at par with those of IBM's, clones could no longer

be exploited by them to increase their profits and in fact became market

impediments that reduced IBM's profits, which resulted in the

creation of an almost perfectly competitive market. Compaq, the leader

among IBM clones aggressively threatened IBM's position. IBM's

market share declined from 30.7% in 1985 to 16.9% in 1989 [26].

This indicates that IBM's open-standard strategy failed to deliver

long-term success.

In this context, the existence of seemingly conflicting opinions regarding

the open-standard strategy resulted in the emergence of the leverage

theory. The leverage theory [5,9,32] encourages platform

owners to completely withhold proprietary technology in order to

avoid future competition with entrants. An open architecture and a

cloning strategy facilitate the reverse-engineering of proprietary components,

which enables newfirms to enter themarket. If a newcompetitor

succeeds in entering the market using reverse-engineering, the

negative impact of rent dissipation may exceed the benefit derived

from the network effects. Therefore, it is recommended that all components

are included in a vertically integratedmarket structure in order to

restrict the entry of new firms. This strategy is consistent with that of

Apple Computers; however, this strategy also failed [34]. Unlike IBM,

Apple Computers retained all the technological expertise for its Apple

series computers in-house, and therefore produced incompatible PCs.

Apple failed to establish a formidable standard in the PC market and

held only 20% of the market by 1983 [26].

The strategic failures of platform owners in the PC market indicate

the limitations of these two contrary perspectives. Now, we will identify

those aspects of network effects that were overlooked by Conner

[10] in the creation of an effective cloning strategy of a network platform.

Moreover, the reasons for the recent revision of the term “IBMcompatible”

to “Wintel-compatible” will be investigated. Currently,

Intel and Microsoft (MS) are essentially considered to be the platform

owners in the PC market. The term “Platform owner” represents a

firm that possesses the ability to control the evolution of the platform

architecture, and the likelihood of innovation in complementary markets.

Hence platform owner leads the commercialization of a system

platform and receives the maximum benefit from a successful commercialization

[17]. By 1986, IBM realized that it had established a

standard and in doing so, they had spawned a number of imitators

by ceding the rights to their most valuable PC components to Intel

and MS [26].When IBM adopted the cloning strategy, it could not ensure

that the quality of its products would be superior than that of its

“clones” unless it maintained a veiled technology. However, IBM possessed

no such proprietary core technology that would enable it to

deliver a higher quality than its clones. Moreover, the term “clone”

implies that their product quality is comparable to that of the incumbent;

therefore, it is unlikely that users perceive IBM's products to be

of a higher quality than that of its “clones.”

These historical examples prompted us to investigate the characteristics

that a platform owner must possess in order to be successful.

An analytical model was developed in order to answer the following

research question: In a high-tech market, which is characterized by

strong network effects and entry threats, what enables a company

to become a sustainable platform owner? In a high-tech market, technological

innovation and consumer acceptance advance rapidly,

which makes it rather difficult for the incumbent to acquire a durable

first-mover advantage [31]. Our results indicate the strategic importance

of proprietary technology management and its synergistic resolution

with the network effects environment.

2. Theoretical background

In this section, we review the studies regarding the leverage theory

and the network effects. These are the two representative theories

regarding product commercialization that offer various insights on

platform strategy. The development of these two strategic schools of

thought is closely related to the production structure in the market.

Therefore, we investigate the meaning of each theory from this perspective.

Moreover, in order to understand our research question

more comprehensively, we further investigate the history and characteristics

of the PC industry in detail.

2.1. Leverage theory

The leverage theory focuses on leveraging the monopoly power of

the incumbent for protecting its position. In this section, we examine

the manner in which this theory is related to the platform strategy of

an innovator. Leverage theory encourages vertical foreclosure of entries

by tying components [5,32]. Basically, tying refers to a strategywherein

a seller ties and sells two or more goods together. However, for an incumbent,

this strategy is more significant than the concept of bundled

sales [4,33]. The incumbent may employ tying in order to protect its

monopolistic position, i.e., to create an entry barrier [5,9,32].

Previous studies indicate the impact of foreclosure of entry essentially

from two perspectives. First, tying reduces incentives of entrants'

investment [8,9]. For example, a monopolistic incumbent of a

PC platform may face competition from potential entrants for all its

components. However, when an incumbent adopts a tie-in sales

strategy for an entire platform, a potential entrant may enter the market

only if it succeeds in innovating all the components of the platform.

Alternatively, in order to complete the platform, an entrant

must depend on another entrants' provision of complementary components.

If an entrant only partially succeeds in innovating its components

and no other player produces the complementary parts, then

the entrant cannot enter the market when the incumbent employs a

tie-in sales strategy. Therefore, a comparison between tying and untying

may reduce the research and development (R&D) investment

incentives of entrants, thereby strengthening the incumbent's monopoly

position [5,9]. In particular, this concept is rather relevant in

a high-tech industry where the innovation of each component requires

substantial investments; however, the success of R&D is characterized

by a significant amount of uncertainty [9].

Second, if the incumbent adopts tying, it can protect its monopolistic

position more easily by employing a price-cost squeeze [2,12].

For example, assume that there exists a monopolistic incumbent

with a tied platform, which comprises only two components, A and

B. If there are two independent entrants for components A and B,

then the incumbent may establish the prices of the components in

such a manner that the price of one component is lower than that

of the entrant's in order to put competitive pressure on the independent

entrants. Although the price that the incumbent establishes for

component B is lower than that of marginal cost, the incumbent

may recover this loss by charging a high price for component A in

the presence of tying. Although such tactics require the incumbent

to charge prices that do not maximize the current profits of the two

components, the incumbent can compensate the lower short-term

profits with higher potential future profits once it has discouraged

the new firms from entering the market [2]. Owing to the practice

of such a prohibitive and predatory pricing, the entrant who considers

component B as a complementary product cannot enter the

market because it has no other components for recovering the loss accrued

on account the predatory pricing of B. Consequently, the entrant

for component B will not be able to enter the market. As a

result, the entrant of component A will also not be able to enter the

market owing to a lack of complementary components [12].

Therefore, from these two perspectives – reduction of the entrant's

R&D incentives and price-cost squeeze – tying enables incumbents to

maintain their monopolistic position. The logic behind Apple's tying

strategy for its Apple series computers may be understood by focusing

on the leveraging effect of tying. However, Apple's closed architecture

and tying strategy was unsuccessful for their PC products wherein a

number of strongly complementary components had been collectively

employed. In the 1980s, IBM declared an “open-standard,” following

which several IBM clones entered the PC market with PCs that were

rather similar to and compatible with the IBM PCs. However, Appleestablished an individual standard and attempted to differentiate its

products by following a strict non-licensing, patent-regulated policy.

Consequently, although IBM's market share increased to over 50% of

the market by 1984, Apple computers had a rather limited customer

base.

As indicated by the leverage theory, tying may be an effective

strategy for restricting the entry of competitors; however, it is detrimental

in terms of increasing the network of the platform. Apple's

tying strategy indicated that a closed architecture may severely jeopardize

a firm's business in a network effects environment. Now, we

will examine the concept of network effects.

2.2. Network effects and modular production

When a larger network yields a greater economic value for a product,

network effects are considered to be determinants of product

success. Moreover, numerous platform products in high-tech markets

operate through networks [28]. Even though these networks are virtual

and invisible, they may be rather critical because they make the

previously introduced or standardized technology more viable and

competitive as compared to the subsequent superior technology.

Therefore, innovators in the high-tech industry must thoroughly understand

the significance of establishing a large network [13,19].

There exist numerous IS studies that provide theoretical as well as

empirical evidence for indicating the strategic importance of network

effects. In the software market, an innovator may charge a higher

price for a standard program; this trend increases as the network of

the product expands [3,14,15]. In this context, a few studies suggested

that the software manufacturers intentionally permitted piracy

in order to increase their user base. Conner and Rumelt [11]

evidenced that providing software without any piracy protection

may be advantageous for both firms and consumers. Furthermore,

as the network of a software expands, an increased number of aftermarket

services are supplied to the network, which enhances the significance

of the network and increases profits [29].

A strategy for establishing a dominant network for a platform is to

establish the interface standard and adopt modular production with

various suppliers for sub-systems [27], that is, the “open-standard”

architecture. Standard interface components interact effectively

since they are not assigned particular configurations; this reduces

the specificity of a platform and increases its flexibility [1]. For example,

IBM developed the Industry Standard Architecture (ISA) and

combined various components, which were produced by various suppliers

in the industry.

When several firms adopt an industry standard for IBM compatible

PCs, the availability of a large network plays an important role

in enhancing the utility of the platform consumers. These consumers

can share information without converting the data from one format

to another, which further expands the available network effects.

Moreover, the recombination of various components in multiple configurations

enables a platform to fulfill heterogeneous demands with

a lower investment, thereby further enhancing the utility of this network.

All these reasons are responsible for increasing the demand for

this standard product [16,22,30].

2.3. PC industry

Conner [10] maintained that encouraging competitors to enter the

market is beneficial for a monopolist because the benefit derived from

an increased user base exceeds the rent dissipation by entrants under

strong network effects. However, IBM's strategy failed to deliver longterm

success. Originally, the key components of IBM PCs were not IBM

technologies. IBM – the dominant leader in the computer industry at

that time – intentionally adopted other suppliers' components such as

Intel'smicroprocessor and Microsoft's operating systemfor rapidmarket

penetration and success. The only proprietary component that was

maintained by IBM was the “basic input/output system (BIOS),” which

was not sufficient for ensuring that the reproduction of PCs comparable

to that of IBM was impossible. IBM clones like Compaq easily reverseengineered

IBM's proprietary component and promptly reproduced PCs

that were comparable to that of IBM's. By reproducing a product that

was already established and accepted in the market, the IBM clones circumvented

the significant investment and risk that is usually associated

with entering a market. They seized IBM's market share by shifting consumers'

preferences from IBMPCs to IBM-compatible clones, and subsequently

displaced IBM from its leading position in the PC market [23].

In contrast, Gawer and Henderson's [17] analysis of Intel's market

entrance strategy emphasizes the importance of strategic handling of

in-house capability and external complementary markets in creating

a successful platform owner position. By entering the complementary

market, Intel managed to establish a position that was strong enough

to control the entire platform. Intel's core competency was the production

of the microprocessors. However, in order to accelerate its

primary business and increase its network, Intel continuously encouraged

the innovation of its complementary products by indicating that

it would not charge excessive rent in these markets and would subsidize

entry into the market [17]. Currently, Intel plays a major role in

developing and distributing the “plug-and-play” interface standards

among component suppliers. Intel focuses on interconnecting the

components and enhancing their performance because faster,

cheaper, and easier use of counterpart components increases the demand

of their own microprocessors. Although Intel controlled the

platform network by focusing on a single component, it encouraged

complementors to invest in additional state-of-the-art platforms.

Moreover, although Intel was an “ingredient brand” of the PC platform,

it made efforts to shift consumer perception from being recognized

as “IBM-compatible” to “Intel Inside,” which essentially made

Intel a platform owner in the PC market.

We developed a mathematical model for explaining the fundamental

reason for Intel's success and providing managerial implications

for firms exploring successful platform owner strategies by

examining this strategic choice of Intel. Furthermore, we explain

why Wintel—the dual platform owner strategy—is sustainable.

3. Model

3.1. Player categories

There are three categories of firms in the market: the innovator,

intermediate goods producers, and potential entrants in the innovator's

monopolistic components market. The last platform is commercialized

as the innovator's brand name. Therefore, an “innovator”

represents a firm that wants to be a platform owner by developing

a few or all the component parts of the platform.

The underlying technology of a product is for a system platform

that consists of numerous components; the innovator must identify

which components it should produce proprietarily and which it

should outsource under an open strategy. In order to formalize this

idea, we denote the monopolistic products as β, which assumes any

value between 0 and 1. In the market, the innovator is the monopolist

for the β portion of the system platform. When the complementary

markets are sufficiently attractive, there are numerous intermediate

goods producers that produce the components for the remainder 1

−β portion of the system. This modular production structure of a

platform is created by the innovator's open technology and licensing

strategies when they develop the technology for the entire line of

components. As indicated in the IBM case, the innovator may occasionally

choose to maintain limited technology for the system platform,

adopt the other suppliers' components, and fuse them with

their proprietary components for producing the final system.

The innovator plays the role of a final platform producer by assembling

its β portion of components with the 1−β portion of thecomponents supplied by the intermediate goods producers. If the innovator

decides to maintain all the technology for the system, there

would be no intermediate goods producers for its final system. Even

if the final platform is produced by third parties, the results would

be identical with those obtained under the current setting of a competitive

final goods market.

Other firms may enter the market by independently conducting

R&D for the innovator's proprietary components. Firms observe the

innovator's market and decide to enter the market only if they expect

their profits to exceed their R&D cost. Henceforth, the subscript i denotes

the innovator, m denotes the intermediate goods producer, and

e denotes the new entrant.

3.2. Demand

In order to derive the demand function, the network effects were

considered for modifying the Conner [10] model. For the final platform,

every consumer possesses a constant valuation or reservation

price,τ, which is uniformly distributed between 1−a and 1: τ ~u

[1−a,1]. Here, aN1; therefore, if their post-purchase cost of learning

to use the product exceeds its benefit, then the reservation price of

the final platform is negative for a few customers. Let the number

of consumers with positive use values (τ≥0) be N. However, the underlying

platform is subject to positive network effects.1 Therefore,

considering the network effects, the following assumptions have

been derived:

Assumption 1. (consumer valuation):

τ~u[1−a+γ(1−β)Q, 1+γ(1−β)Q]

Network effects are represented by the parameters Q, γ, and β. Q

is the expected size of the user base in rational expectations equilibrium,

which is unique and equal to the actual equilibrium demand.2

γ is the marginal value that is offered to consumers when the user

base increases by one person, that is, the magnitude of network effects.

β is the innovator's monopolistic production level. The consumers

can obtain a larger utility with a larger network than with

a stand-alone use of the platform; this is because consumers can

connect with more users (e.g., through email) or collaborate more

easily (e.g., using a word processor) with a larger user base [22].

Therefore, the demand for a large network platform increases with

an increase in the number of users. We capture this network effect

by γQ. Furthermore, as β decreases, the platform becomes more

open and standardized since the production of various compatible

components enables a more flexible re-configuration and assembly

of the platform. This technological openness enables the platforms

to meet the demand of heterogeneous consumers more effectively,

which in turn increases the demand for this platform and expands

the network effects. This network effect is reflected in the openness

of the platform, (1−β). Considering this effect, we assume that the

consumers' utility is enhanced byγ(1−β)Q,which changes the original

distribution of τ as assumed in Assumption 1.

Assumption 2. (potential buyers):

1+γ(1−β)Q

Once again, it is assumed that despite the network effects consideration,

there exist consumers with negative utility, 1−a+γ(1−β)

Qb0. Under such conditions, based on Assumption 1, the number of

potential buyers is N(1+γ(1−β)Q. Here, we normalize N to be 1,

which results in 1+γ(1−β)Q.

Assumption 3. (utility function):

U(τk,pf)=τk−pf.

If sf and pf represent the quality and price of the final platform, respectively,

then consumer k's utility from the final platform consumption

is defined as U(τk, sf,pf)=sfτk−pf. Since the valuation on the

platform quality, τk, varies for each consumer, let sf be 1 assuming

that the quality of the original product is consistent. Therefore, the

utility function is as provided in Assumption 3.

Assumption 4. (demand for final platform):

qf ¼

1−pf

1−γً1−βق

:

Assumption 3 presents the condition for product purchasing such that

τk≥pf. Therefore, consumers whose τ is between pf and 1+γ(1−β)Q

will purchase the product (see Assumption 1). When F(⋅) is a cumulative

distribution function of τ conditional on τ≥0 among potential buyers, the

demand for the final platform is qf={1+γ(1−β)Q}{1−F(pf)}. Since qf

indicates the market demand for the product, it is equal to the user base

Q in rational expectations equilibrium, that is, Q=qf. Moreover, since F

(⋅) is conditional on τ≥0, F(pf)equals pf

1‏γً1−βقqf

. Therefore, the resulting

demand is given as 1−pf

1−γً1−βق

.

Assumption 5. (magnitude of network effects):

0bγb1.

In order to guarantee a downward sloping demand function (see

Assumption 4), we assume that 0bγ(1−β)b1. Here, the magnitude

of network effects, γ, is an exogenous variable of the innovator,

which is given in the market and cannot be limited by the size of

the endogenous decision variable β. In other words, 0bγ(1−β)b1

must always be satisfied for any value of β between 0 and 1, and Assumption

5 has been employed in order to ensure this. This assumption

is identical to that used in Conner [10].

3.3. Cost

Assumption 6. (intermediate good):

ci(x)=ce(x)=cm(x)=cx for x∈[0, 1].

The system platform is a set of infinitely small intermediate goods,

which are uniformly distributed between [0, 1]. In order to eliminate

the impact of economies of scale on production, which may induce

the innovator to curtail proprietary control, we assume that every

producer of intermediate goods has a constant unit cost c for each intermediate

good x∈[0, 1]. In addition, for the sake of simplifying

1 Instead of network externalities, this study employs the concept of network effects.

Originally, network effects referred to those circumstanceswherein the net value of an action

is influenced by the number of agents undertaking the same action. In this context, if

the consumption utility of a good is based on the number of other agents consuming the

good, then such a good is believed to demonstrate ‘network externality’ [19]. Here, the difference

between network externality and network effect is characterized by the existence

of market failure. Therefore, Liebowitz and Margolis [21] indicated that the application of

the concept of network externality is limited since network externalities as market failures

are theoretically fragile and empirically undocumented, whereas the concept of network

effects is commonly accepted and important. Liebowitz and Margolis [21]

emphasized that numerous external effects of network size are merely pecuniary; therefore,

network externality was considered to be a particular type of network effect where

equilibrium exhibits unexploited gains from trade regarding network participation.

According to the distinction suggested by Liebowitz and Margolis [21], ‘network effects’

is the more appropriate term for this paper.

2 Rational expectations equilibrium is typically based on the fundamental economic

assumption that people behave in ways that maximize their utility; therefore, all relevant

information has been used in formulating expectations for economic variables.

Therefore, it is assumed that expectations must not be systematically biased and must

essentially constitute informed predictions of future events [24]. Although, it is a

strong assumption that the expected Q equals the real Q, this assumption has been

adopted in several previous studies [7,10,18,20].