Compatibility, Standards, and Software Production - Semantic Scholar
Appeared in ACM StandardView, 6(4), 1998
Compatibility, Standards, and Software Production Giancarlo Succi‡, Andrea Valerio§, Tullio Vernazza§, Gianpiero Succiƒ ‡
§
Department of Electrical and Computer Engineering The University of Calgary
Dipartimento di Informatica, Sistemistica e Telematica Università di Genova ƒ
Dipartimento di Scienze Giuridiche Università di Genova
Abstract Compatibility is a key strategic decision in software production. Proposals exist for standards in several fields of software production, such as networking (ISO and IEEE), operating systems (Posix), object management (OMG). However a formal treatment of standards in software is still missing. This paper tries to overcome such lack, presenting a model of the effects of compatibility in software production. It overviews existing works on compatibility. It details a model on the effects of compatibility decisions in software development. It describes the application of this model to the cases of new products being introduced and of well-established incumbents. 1.
INTRODUCTION
Compatibility is a key strategic decision in software production, in particular when the software development process is based on domain analysis and reuse. For the last few years several software firms have been facing process improvement and product quality issues. However few are conscious of the role of market structure in their operations. Often "experts", following rules of thumb, perform the analysis of the application. They aim at identifying the feasibility and the potential return on investments, but usually do not consider in depth aspects related to market structure, such as compatibility and consumer network effects. Proposals exist for standards in several fields of software production, such as in networking (ISO and IEEE), in operating systems (Posix), in object management (OMG). However a formal treatment of standards in software is still missing. This paper describes a mathematical model of the effects of compatibility on software production, and discusses the application of this model in different scenarios. The works of Matutes and Regibeau [6,7] are the base. Strong references are also to the proposals of Farrel and Saloner [3,4] and of Katz and Shapiro [5]. This paper is organized as follows. A review of the existing approaches is presented in section 2. Section 3 proposes a framework to determine the effects of standards in software production. Section 4 analyses the
‡
Corresponding Address: Giancarlo Succi, Department of Electrical and Computer Engineering, The University of Calgary, 2500 University Drive, Calgary, Alberta, Canada T2N 1N4. E-mail: [email protected]
G. Succi, A. Valerio, T. Vernazza, G. Succi (1998), “Compatibility, Standards, and Software Production”, ACM StandardView, ACM Press, 6(4)
deductions coming from the application of the framework. Section 5 draws some conclusions and outlines future directions of this research. 2 DEFINITIONS AND MODELS OF COMPATIBILITY
Compatibility is the capability of performing in harmonious or congenial combination with different parts. Compatibility is a crucial aspect in many industrial sectors. It has strong repercussions in information technology, due to networking, data sharing and teams work.Compatibility may be achieved through standardization, or developing adapters between incompatible parts. Standardization means that there is an explicit or an implicit agreement to do certain things in a defined and uniform way. For instance in computers, two devices are compatible when they can be connected together to perform a desired functionality: they have been developed following the same standard. There are standard at the level of input/output, of inter-process communications (DCE), of messages (Corba), of system calls (Posix), of GUIs (Motif), and so on. Standardization is achievable through government actions, through the coordination activities of voluntary industry standards, or through independent actions of industries presents in the market. Adapters are devices that provide compatibility. They are put between two modules that have to be connected and translate from one to the other, processing the information exchanged between the two. The compatibility dilemma is: "When is it convenient to produce a good that is compatible with concurrent firms?" Compatibility decisions heavily influence the market behavior. The installed base is the set of customers which have chose to adopt a specific brand, i.e. it represents the number of people or organizations that have installed and use a particular product. To answer the compatibility dilemma, we need to consider network externalities. Network externalities are the effects of the installed base on the customers’ demand and enjoyment for a specific good or service. In other words, positive network externalities arise when customers value a product more if it is compatible with other consumers’ products. In markets with network externalities compatible designs may emerge as the preferred choice by producers. Network externalities are a source of scale economies that arise from the demand side of the market. A compatible technological design increases the value that consumers derive from a firm’s product. Compatibility gives the consumer the benefits of other firms’ networks. 2.1 A Model of Compatibility with Network Externalities In general, installed bases are important and consumption network externalities arise when customers wish or need to communicate directly or indirectly with each other. This is the case, for example, when files are shared among different computers, or when information is broadcast in the Internet. Consumption network externalities also arise when there is a large installed base for a good that allows producers of complementary goods to exploit economies of scale. This leads, for example, to a greater variety of software applications that runs on a specific computer, which has a broad installed base. Katz and Shapiro [5]developed a simple, static model of oligopoly to analyze markets which consumption externalities are present, and consumers value a product more when it is compatible with other consumers’ products. They examined different perspectives concerning network externalities and their influence on the market equilibrium. In particular, they studied the effects of consumption network externalities on competition, and demonstrated that consumption externalities give rise to demand-side economies of scale, which will vary with consumer expectations. Besides, they explored the effects of network externalities on compatibility decisions, especially regarding private and social incentives to produce compatible products,
or to switch from incompatible to compatible products. They found that firms with good reputation or large existing installed base tend to be against compatibility, aiming not to decline in market share, while firms with small networks or weak reputation tend to favor product compatibility. The work of Katz and Shapiro highlights and confirms the importance of consumers’ expectations concerning future installed base grown in markets where network externalities are present. 2.2 Innovation and Compatibility Decisions In situations characterized by strong network externalities and important compatibility issues, the introduction of innovative technology and products can be difficult. When the innovation is incompatible with the installed base, the likelihood and desirability of it has to be compared with private and social incentives, and its introduction can be inhibited by the effects on the market of network externalities and compatibility. Farrel and Saloner [3] studied the relationship between compatibility and innovation introduction in the presence of network externalities. In compatibility situation, where there is a dominant installed base composed by compatible brands, demand-side economies of scale arise, and there are real benefits in acting as other do. These benefits include interchangeability of complementary products, ease of communication among users in the same network, and cost savings, especially due to mass production and selling. In a situation of compatibility there are important network externalities effects; therefore, the customers that wants to switch to a new technology bears a disproportionate share of transient incompatibility costs. The same applies for products incompatible with the current network. This possibly creates a social excessive reluctance to adopt an innovative technology, which is called excess inertia by Farrel and Saloner. On the other hand, the new technology can offer to its first customer an important andvantage. Therefore, users may adopt a new technology as soon as it comes. If this is the case, an inefficient adoption of a new technology could arise, resulting in the stranding of the old network. This inefficient adoption of a new technology is called excess momentum. Different aspects influences the transition dynamic, whether there is a transition, toward the innovative technology: the size of the installed base, how quickly the network externality benefits of the new technology arise, the relative superiority of the new technology, and marketing strategy of the innovative product. A central role is also played by information and communication: in the model with incomplete information proposed by Farrel and Saloner [4], they showed that there is always excess inertia. One of the possible coordination mechanisms that can arise in a market where different standard have been proposed is named bandwagon. If an important organization makes a unilateral public commitment to one standard, other firms know that if they follow that leader, they will be compatible at least with the first mover, and probably with the later movers. In the bandwagon mechanism, one leader that adopt a specific standard is followed by other firms which want to benefit from the compatibility with the leader and to enter its network. Depending on the strength of the leader and on the market situation, the bandwagon mechanism can sometimes achieve rapid and effective coordination, resulting in a common standard adoption by the market. The earlier a firm joins the leader, the highest are the benefits that it can expect from its choice, mainly due to the experience and knowledge already acquired when new firms will join the network. But also, the earlier a firm joins the leader, the worse will the damage if the network does not survey in the market.
The bandwagon mechanism for coordinating is imperfect when there is not a clear leader and when there are different preferences among standards. This is, most of the times, the case in the software industry: generally there are a few standards, every one with its consolidated installed base, which compete in a application domain. 2.3 "Mix & Match" and Bundling Compatibly is directly related to modularity: when firms develop products composing different modules, or when consumers can assemble their own systems starting from parts commercially available, concurrent firms have to decide whether to make their components compatible with those of their rivals. In the information technology field this issue is particularly relevant, because of the ever-increasing communication and interoperability needs of customers. Moreover, the market proposes many different products, but computers and software environments create a web of users networks that exploit network externalities effects, and binds firms to value very carefully the need of compatibility with concurrent products and technologies. Matutes and Regibeau [6]showed that in some circumstances the producers of alternative technologies have strong incentives to design compatible products. When modules are composed in order to function as a whole, module compatibility reduces the incentive of competing firms to set low prices for component parts, because low prices increase the sale of a compatible rival, too. Despite higher prices, consumers can be better off because compatibility allows them to assemble systems that are closer to their ideal configurations, because of the variety increase of systems available. Matutes and Regibeau has demonstrated these results for fully integrated firms in a duopoly. Economides [2] verified them for n firms. In a successive works, Matutes and Regibeau [7] extended their model to firms pricing a bundle of their components separately from their individual components. This mixed bundling causes the firms, which produce compatible modules, to offer a discount on their full system, even if the best choice could be to agree not to bundle. Moreover, Matutes and Regibeau proved that firms’ ability to exploit the endogenous network externality associated with compatibility depends on the marketing strategies adopted by the firms. 3 FRAMEWORK FOR THE ANALYSIS The information technology market is highly dynamic and it is characterized by continuous technology innovation and product evolution. The typical software application is rather big and complex, generally it is a composite system where communicating modules cooperate to perform the desired functionality. This modules are the instances in a specific environment of reusable components. Figure 1 presents this scenario. The context considered in this work is a software organization that develops a standard component, to exploit the opportunity arose from the analysis of its installed base in a specific domain. The cost for the production of the component is compared to the return given by the reuse it inside the software applications already installed to the customers. The potential return on the investments has been evaluated considering a market share equal to the actual installed base of the firm in the specific domain under consideration. The framework developed is compatible with the current installed base, but the producer has to face the decision whether to make its framework compatible with concurrent firms’ applications. The information technology market, with its rapid growth and innovation, has put a new emphasis on the concept of compatibility and standardization.
MARKET SEGMENTS
APPLICATIVE DOMAIN
COMPATIBLE COMPONENTS application C
application B
ta i ns
nce
s
application A
Figure 1: Our scenario. One of the principal reasons for this can be found in the evolution that the role of computers has undergone: from task automation and speed-up tools to communication and information providers. Computer and software industry has driven most of the recent studies in formal standardization; however, these efforts have not been supported by an adequate theoretic work on the private and social effects of compatibility decisions. 3.1 The Reference Approach Most of the existing researches focus on identifying the equilibrium conditions that held in the market when the compatibility is the variable in the analysis. Private and social incentives and costs are considered as evaluation parameters in the compatibility game. This paper does not examine the social optimality of the firms’ compatibility decisions, but focus on the firms’ incentives to produce compatible products and the related conscious decision that they undertake. BB
B
B BB AA
A AA
A
Incompatibility AB
A AA
BB B
BA
B
B
A
AB
BB
AA
BA A
AB
BB
AA
BA
Compatibility
Figure 2: Equilibrium Market Configurations (from Matutes and Regibeau [6]).
In the framework first explored by Katz and Shapiro [5] and Farrell and Saloner [3], a system composed by compatible components is treated as a single good characterized by positive consumption externalities. These authors model firms as selling a single good, assuming implicitly that every firms sells every component of a system, identified as the single good useable by customers, or that manufacturers sell a component to customers who already own the rest of the system. The last is, for example, the case of software which is usually sold without the computer that the consumer already own. A different approach is proposed by Matutes and Regibeau [6,7]: they show that additional considerations concerning compatibility issues can be achieved by explicitly modeling a system as a set of related components, instead of treating it as a single good. Firms sell complete systems, but it may be possible to sell single components, allowing the use of systems which combines components produced by different companies. Following the approach proposed by Matutes and Regibeau, the model can be restricted to the consideration of two independent firms which sell a functionally equivalent product composed by two cooperating modules. Both firms produce the two modules of a product, and the modules produced by one firm are differentiated from the modules produced by the other firm. The marginal costs for the production are set equal to zero, and there are no economies of scope. Figure 2 (from Matutes and Regibeau [6]) represents the situation: the two firms can be placed in the corners of a square (with side length normalized to one). If the two modules sold by the two firms are not compatible, then consumers can chose only between the two pure system proposed by the two firms. Considering the modules compatible, it is possible to combine a component sold by the first firm with the complementary component from the second firm. As a result, consumers can purchase one of four possible systems, the two pure systems and the two hybrid system obtained combining one module of the first firm with one module of the second firm. Distributing the consumers on the unit square in a uniform way, the first firm, named A, is placed in the lower-left corner, while the second firm, named B, is placed in the upper-right corner, with coordinates (1,1) for the normalization. Normalizing the user preference for a module of a firm in the range [0,1], a user placed at a point of coordinates (x, y) in the square has a preferred first module that is x away from the first module of firm A, and a preferred second module that is y away from the second module sold by firm A. A complementary situations, with distances 1-x and 1-y, holds concerning firm B. The reservation price, R, is the price a customer is available to pay for purchasing its ideal system. The customer surplus is the hypothetical saving a customer has buying an existing technology. The customer surplus, S, is difference of the reservation price and the hypothetical cost of the technology. The hypothetical cost of the technology is the sum of the real price, P, and the hypothetical cost of the distance of the purchased technology and the ideal technology, D. That is: S=R-P-D A customer purchases a technology if s/he feels that the surplus is positive. If both the technologies provide the user a positive surplus, s/he will purchase the cheaper. The question is: when is it convenient to provide compatibility for a firm, (1) to maximize S for its products but also (2) to avoid the competitor from taking hostile advantages from its compatibility? The worst situation for a firm would be to spend money to make a product compatible and to have its competitor taking the full advantage from the compatibility.
We need to study the structure of S, that is, of R, P, and D. 3.2 Our Hypotheses We assume the following. 1. For software firms there are always economies of scales. Once the first instance of a product is produced, the other copies are almost for free. 2. P depends on the absolute time, t. P depends on the cost to produce the component. As from Katz and Shapiro [5], the cost is strictly decreasing in t. Therefore, P is strictly decreasing, too. 3. The reservation price depends both on the value of the intrinsic good and the value of the network. 4. The intrinsic value of a good is independent of time. 5. The network evolves with the absolute time. 6. Network externalities are important in the software industry, since communication, data sharing and coordination are widely present. 7. The bandwagon effect is strong. From 1. and 2., we can model the behavior of the price of a component as: P(t) = p e-x(t-t’). p depends on the cost of the resources to produce the product; x depends on technological factors of the application domain; t’ is the time of introduction of the product in the market. From 3., the reservation price has the shape of: R(t,n) = Ri(t) + Rnet(n) where Ri(t) is the intrinsic value of the good, the value that the customer associates directly to the good as a stand-alone entity. Rnet(n) is the value given to the product by the presence of a network made up by n consumers. From 4., we assume that the intrinsic value of the good Ri(t) is independent from the absolute time: Ri(t) is Ri. We assume also that Rnet(n) is linearly related to the size of the network. Ri(t) = Ri Rnet(n) = rn From 5., the network evolves with time. As from the literature, we assume two possible evolutions, a linear evolution, and an exponential evolution: n(t) = h (t – t”) [linear evolution] n(t) = b eg(t-t”) [exponential evolution] t” is the time when the good is purchased for the first time. The constants b, h, and g may be positive or negatives, depending of the lifecycle phase of the object. This implies that: Rnet(n) = r n(t); therefore, Rnet(t) = h (t-t”) or Rnet(t) = b eg(t–t”)
This evolution occurs while the network is not saturated. Once saturated, a constant period may occur or a negative trend may start. For simplicity, from now on, we assume t’=0 and we define a new t" as t’-t". Figure 3 details the behavior of Rnet; for space constraint the time of constant value is omitted. Shape 1 represents the initial behavior of the reservation price as a function of the time. Point M represents the maximum asymptotic reservation price achievable considering the network maximum dimension. The coordinates (F,L) indicates the change point from the ideal behavior and the step to the shape 2. At the growing of the network size, a new effect appears (shape 3): positive network externalities effects begin to grow less and then to decrease due to possible congestion of the network (this is, for example, the case of Internet when there are too many users connected at the same time). The point at coordinates (N,S) marks the step to this situation, and at the point (Ri, E) there is a perfect equilibrium between positive network externalities effects and network saturation effects. From 6., and 7. we focus mainly on the exponential evolution. Moreover we focus in this paper on the growing pattern of the curve. Rnet
M N 2 3
F 1 Rn(d) Ri
t-t” L
S
E
Rnet
M N 1
2 3
F
R n(d)
Ri
t-t” L
S
E
Figure 3: Evolution of Rnet. 3.3 The Model From our assumptions the customer surplus is expressed by: S(t) = Ri + b eg(t-t”) - p e-xt - D All the constants are positive. A customer will choose the technology with the highest value of S, provided that it is positive. Otherwise it will not buy any product. -xt
S(t) = Ri + b eg(t-t”) - p e –D =
-xt = (Ri) + (- D - p e ) + (b eg(t-t”))
There are three components in the surplus. A value in the ideal product, (Ri) , a value in the real product, (D - p e-xt), and a value coming from the network (b eg(t-t”)). The value of the ideal product is fixed in our analysis. The value of the real product depends on the compatibility decisions. From Figure 2 it is evident that compatibility decisions may help both competitors, since they are able to sell products to customers that would not buy their own product individually. The value coming from the network is also strongly influenced by compatibility decisions. t" plays an important role: it is the time of the first introduction of the product in the network. Firms want to lower t" as much as they can. We can express the profit of a firm, Q, as: +¥
Q = å diP(i) - C i = t"
Where di is the number of firms buying the product P at instant i, C is the cost for producing the product. That is the profit is the sum of the purchase price of the firm and the cost to produce the product, that is constant, since we assume marginal costs are zero. A firms increases its profits if it increases the number of customers, that is, if it maximizes: +¥
åd
i
i = t"
Therefore the profits of a firm Q is strictly and monotonically related to the customer surplus. A firm could also increase its profits increasing the price, P(i) in the equation; however, if a firm increases its prices, a lower number of customers will buy its products, therefore lowering the global value of Q. This can be accomplished through different ways: 1. A firm need to enter the market, i.e., to have t" lower than +¥. 2. A firm t has to reduce the distance with the ideal product D. In these two cases compatibility is a key feature, since it allows a higher surplus due to network externalities of incumbent products. 3. A firm has to enlarge its market share. There are two different cases, according to our hypotheses on R and P. Initially compatibility supports synergy of incumbents. However the fast evolution of the market makes compatibility decision hard. The entire problem in this situation is the understanding of the value of g and x. Unfortunately the writers have not been able to run suitable simulations, so the whole discussion is only qualitative. g and x are two different constants: g refers to the growing pace of the network, while x depends on the ability of the firm to discount the initial cost of its product. However they are related. Static situations are characterized by low values of them. Dynamic situations are characterized by high values of them. Static situations occur when few large customers are present, the market is captive, personal relationships play a relevant role. Dynamic situations occur when a lot of small customers. Unfortunately we are not able to provide any quantitative definition of what is static and what is dynamic.
4 DISCUSSION
From our model, it seems that compatibility is not a good or a bad property of a software product. Its value depends mainly on the new product and on the market structure. We support our view with a brief analysis of few possible scenarios. Scenario 1. A firm entering a completely virgin, narrow market. The firm is the first one in the market. It can take advantage of its position. Any other firm entering the market will have to face an already installed network. Compatibility is not advisable. Scenario 2. A firm entering a complete virgin, wide but dynamic market. The firm cannot satisfy all the expectations from the habitants. However it may expect that the benefits coming from network externalities will overcome the gap soon, thus making compatibility not desirable. Scenario 3. A firm entering a complete virgin, wide but static market. The firm cannot satisfy all the expectations from the habitants. If another firm enters the market, a profitable situation for both firms may occur. Compatibility enables such situation. The two products from the two firms can work together to satisfy the more distant customers. However, due to the exponential nature of the curve, the firm should prepare for a competition with the other firm, on the long run. Strategy for making a compatible product not compatible should be devised. Scenario 4. A firm entering an existing and populated market. Compatibility is the way to go. Firms which have not or have a little installed base (or suffers of little consideration from customers), are usually oriented toward the production of modules compatible with bigger firms’ ones. The firm should try as hard as possible to take advantage of the existing network. It is worthwhile to notice that the European Union legislation permits a firm to disassemble somebody else’s product to make its own product compatible with it. A comprehensive discussion of this topic can be found in [8]. Scenario 5. A firm keeping a leading role in a populated market. Compatibility does not carry any advantage, because of the firm already has a large installed base and a good consideration from customers. In the software and hardware market, technology play a fundamental role: due to the rapid evolution of hardware and software products, consumer network are highly dynamic and difficult to estimate. This situation is thicken by the strong correlation existing among hardware, software and applications evolution. New technologies introduced in the market, having great intrinsic potential but incompatible with the actual technologies, could strand the competing networks. 5 CONCLUSIONS AND FURTHER RESEARCH
Compatibility decisions influence the success of a product. The answer to the compatibility dilemma depends on multiple factors: the size of the network, the number and the kinds of incumbents, the technology employed and so on. In this paper we have presented a simple model to represent the role of compatibility in software production. The specificity of software firms has been captured through a sequence of working hypothesis. We have conducted the analysis from the firm’s profit standpoint; we have not considered social welfare. It appears that the high dynamics of the software industry makes compatibility advantageous mainly to new firms entering a wide virgin market or a populated market. In the other cases incompatibility seems the way to go. Several questions arise from this research. It would be relevant to perform simulations of the behavior of incumbents when a compatible product enter an existing dynamic market, and of the long-term compatibility behavior of incumbents in a static market. Tools and methods to build a product that is initially compatible and then can be transformed to incompatible would also be extremely significant.
REFERENCES
1. Boehm, Barry B. (1984): Software Engineering Economics, IEEE Transactions on Software Engineering, Vol. SE-10, No. 1, January 1984, pages 4-21 2. Economides, Nicholas (1989): Desirability of Compatibility in the Absence of Network Externalities, American Economic Review, Vol. 79, No. 5, pages 1165-1181, 1989 3. Farrell, Joseph, and Garth Saloner (1985): Standardization, Compatibility, and Innovation, Rand Journal of Economics, Vol. 16, No. 1, Spring 1985 4. Farrell, Joseph, and Garth Saloner (1986): Installed Base and Compatibility: Innovation, Product Preannouncements, and Predation, American Economic Review, Vol. 76, No. 5, December 1986 5. Katz, Michael L., and Carl Shapiro (1985): Network Externalities, Competition, and Compatibility, American Economic Review, Vol. 75, No. 3, June 1985, pages 425-440 6. Matutes, Carmen, and Pierre Regibeau (1988): "Mix and Match": Product Compatibility without Network Externalities, Rand Journal of Economics, Vol. 19, No. 2, Summer 1988 7. Matutes, Carmen, and Pierre Regibeau (1992): Compatibility and Bundling of Complementary Goods in a Duopoly, Symposium on Compatibility, Journal of Industrial Economics, Vol. XL, No. 1, March 1992 8. Succi, Giancarlo, Gianpiero Succi, and Marco Ronchetti (1997): Legal Issues regarding Software Use and Reuse within the European Union Legislation. Journal of Computing and Information Technology.(In publication in 1997)
Compatibility, Standards, and Software Production Giancarlo Succi‡, Andrea Valerio§, Tullio Vernazza§, Gianpiero Succiƒ ‡
§
Department of Electrical and Computer Engineering The University of Calgary
Dipartimento di Informatica, Sistemistica e Telematica Università di Genova ƒ
Dipartimento di Scienze Giuridiche Università di Genova
Abstract Compatibility is a key strategic decision in software production. Proposals exist for standards in several fields of software production, such as networking (ISO and IEEE), operating systems (Posix), object management (OMG). However a formal treatment of standards in software is still missing. This paper tries to overcome such lack, presenting a model of the effects of compatibility in software production. It overviews existing works on compatibility. It details a model on the effects of compatibility decisions in software development. It describes the application of this model to the cases of new products being introduced and of well-established incumbents. 1.
INTRODUCTION
Compatibility is a key strategic decision in software production, in particular when the software development process is based on domain analysis and reuse. For the last few years several software firms have been facing process improvement and product quality issues. However few are conscious of the role of market structure in their operations. Often "experts", following rules of thumb, perform the analysis of the application. They aim at identifying the feasibility and the potential return on investments, but usually do not consider in depth aspects related to market structure, such as compatibility and consumer network effects. Proposals exist for standards in several fields of software production, such as in networking (ISO and IEEE), in operating systems (Posix), in object management (OMG). However a formal treatment of standards in software is still missing. This paper describes a mathematical model of the effects of compatibility on software production, and discusses the application of this model in different scenarios. The works of Matutes and Regibeau [6,7] are the base. Strong references are also to the proposals of Farrel and Saloner [3,4] and of Katz and Shapiro [5]. This paper is organized as follows. A review of the existing approaches is presented in section 2. Section 3 proposes a framework to determine the effects of standards in software production. Section 4 analyses the
‡
Corresponding Address: Giancarlo Succi, Department of Electrical and Computer Engineering, The University of Calgary, 2500 University Drive, Calgary, Alberta, Canada T2N 1N4. E-mail: [email protected]
G. Succi, A. Valerio, T. Vernazza, G. Succi (1998), “Compatibility, Standards, and Software Production”, ACM StandardView, ACM Press, 6(4)
deductions coming from the application of the framework. Section 5 draws some conclusions and outlines future directions of this research. 2 DEFINITIONS AND MODELS OF COMPATIBILITY
Compatibility is the capability of performing in harmonious or congenial combination with different parts. Compatibility is a crucial aspect in many industrial sectors. It has strong repercussions in information technology, due to networking, data sharing and teams work.Compatibility may be achieved through standardization, or developing adapters between incompatible parts. Standardization means that there is an explicit or an implicit agreement to do certain things in a defined and uniform way. For instance in computers, two devices are compatible when they can be connected together to perform a desired functionality: they have been developed following the same standard. There are standard at the level of input/output, of inter-process communications (DCE), of messages (Corba), of system calls (Posix), of GUIs (Motif), and so on. Standardization is achievable through government actions, through the coordination activities of voluntary industry standards, or through independent actions of industries presents in the market. Adapters are devices that provide compatibility. They are put between two modules that have to be connected and translate from one to the other, processing the information exchanged between the two. The compatibility dilemma is: "When is it convenient to produce a good that is compatible with concurrent firms?" Compatibility decisions heavily influence the market behavior. The installed base is the set of customers which have chose to adopt a specific brand, i.e. it represents the number of people or organizations that have installed and use a particular product. To answer the compatibility dilemma, we need to consider network externalities. Network externalities are the effects of the installed base on the customers’ demand and enjoyment for a specific good or service. In other words, positive network externalities arise when customers value a product more if it is compatible with other consumers’ products. In markets with network externalities compatible designs may emerge as the preferred choice by producers. Network externalities are a source of scale economies that arise from the demand side of the market. A compatible technological design increases the value that consumers derive from a firm’s product. Compatibility gives the consumer the benefits of other firms’ networks. 2.1 A Model of Compatibility with Network Externalities In general, installed bases are important and consumption network externalities arise when customers wish or need to communicate directly or indirectly with each other. This is the case, for example, when files are shared among different computers, or when information is broadcast in the Internet. Consumption network externalities also arise when there is a large installed base for a good that allows producers of complementary goods to exploit economies of scale. This leads, for example, to a greater variety of software applications that runs on a specific computer, which has a broad installed base. Katz and Shapiro [5]developed a simple, static model of oligopoly to analyze markets which consumption externalities are present, and consumers value a product more when it is compatible with other consumers’ products. They examined different perspectives concerning network externalities and their influence on the market equilibrium. In particular, they studied the effects of consumption network externalities on competition, and demonstrated that consumption externalities give rise to demand-side economies of scale, which will vary with consumer expectations. Besides, they explored the effects of network externalities on compatibility decisions, especially regarding private and social incentives to produce compatible products,
or to switch from incompatible to compatible products. They found that firms with good reputation or large existing installed base tend to be against compatibility, aiming not to decline in market share, while firms with small networks or weak reputation tend to favor product compatibility. The work of Katz and Shapiro highlights and confirms the importance of consumers’ expectations concerning future installed base grown in markets where network externalities are present. 2.2 Innovation and Compatibility Decisions In situations characterized by strong network externalities and important compatibility issues, the introduction of innovative technology and products can be difficult. When the innovation is incompatible with the installed base, the likelihood and desirability of it has to be compared with private and social incentives, and its introduction can be inhibited by the effects on the market of network externalities and compatibility. Farrel and Saloner [3] studied the relationship between compatibility and innovation introduction in the presence of network externalities. In compatibility situation, where there is a dominant installed base composed by compatible brands, demand-side economies of scale arise, and there are real benefits in acting as other do. These benefits include interchangeability of complementary products, ease of communication among users in the same network, and cost savings, especially due to mass production and selling. In a situation of compatibility there are important network externalities effects; therefore, the customers that wants to switch to a new technology bears a disproportionate share of transient incompatibility costs. The same applies for products incompatible with the current network. This possibly creates a social excessive reluctance to adopt an innovative technology, which is called excess inertia by Farrel and Saloner. On the other hand, the new technology can offer to its first customer an important andvantage. Therefore, users may adopt a new technology as soon as it comes. If this is the case, an inefficient adoption of a new technology could arise, resulting in the stranding of the old network. This inefficient adoption of a new technology is called excess momentum. Different aspects influences the transition dynamic, whether there is a transition, toward the innovative technology: the size of the installed base, how quickly the network externality benefits of the new technology arise, the relative superiority of the new technology, and marketing strategy of the innovative product. A central role is also played by information and communication: in the model with incomplete information proposed by Farrel and Saloner [4], they showed that there is always excess inertia. One of the possible coordination mechanisms that can arise in a market where different standard have been proposed is named bandwagon. If an important organization makes a unilateral public commitment to one standard, other firms know that if they follow that leader, they will be compatible at least with the first mover, and probably with the later movers. In the bandwagon mechanism, one leader that adopt a specific standard is followed by other firms which want to benefit from the compatibility with the leader and to enter its network. Depending on the strength of the leader and on the market situation, the bandwagon mechanism can sometimes achieve rapid and effective coordination, resulting in a common standard adoption by the market. The earlier a firm joins the leader, the highest are the benefits that it can expect from its choice, mainly due to the experience and knowledge already acquired when new firms will join the network. But also, the earlier a firm joins the leader, the worse will the damage if the network does not survey in the market.
The bandwagon mechanism for coordinating is imperfect when there is not a clear leader and when there are different preferences among standards. This is, most of the times, the case in the software industry: generally there are a few standards, every one with its consolidated installed base, which compete in a application domain. 2.3 "Mix & Match" and Bundling Compatibly is directly related to modularity: when firms develop products composing different modules, or when consumers can assemble their own systems starting from parts commercially available, concurrent firms have to decide whether to make their components compatible with those of their rivals. In the information technology field this issue is particularly relevant, because of the ever-increasing communication and interoperability needs of customers. Moreover, the market proposes many different products, but computers and software environments create a web of users networks that exploit network externalities effects, and binds firms to value very carefully the need of compatibility with concurrent products and technologies. Matutes and Regibeau [6]showed that in some circumstances the producers of alternative technologies have strong incentives to design compatible products. When modules are composed in order to function as a whole, module compatibility reduces the incentive of competing firms to set low prices for component parts, because low prices increase the sale of a compatible rival, too. Despite higher prices, consumers can be better off because compatibility allows them to assemble systems that are closer to their ideal configurations, because of the variety increase of systems available. Matutes and Regibeau has demonstrated these results for fully integrated firms in a duopoly. Economides [2] verified them for n firms. In a successive works, Matutes and Regibeau [7] extended their model to firms pricing a bundle of their components separately from their individual components. This mixed bundling causes the firms, which produce compatible modules, to offer a discount on their full system, even if the best choice could be to agree not to bundle. Moreover, Matutes and Regibeau proved that firms’ ability to exploit the endogenous network externality associated with compatibility depends on the marketing strategies adopted by the firms. 3 FRAMEWORK FOR THE ANALYSIS The information technology market is highly dynamic and it is characterized by continuous technology innovation and product evolution. The typical software application is rather big and complex, generally it is a composite system where communicating modules cooperate to perform the desired functionality. This modules are the instances in a specific environment of reusable components. Figure 1 presents this scenario. The context considered in this work is a software organization that develops a standard component, to exploit the opportunity arose from the analysis of its installed base in a specific domain. The cost for the production of the component is compared to the return given by the reuse it inside the software applications already installed to the customers. The potential return on the investments has been evaluated considering a market share equal to the actual installed base of the firm in the specific domain under consideration. The framework developed is compatible with the current installed base, but the producer has to face the decision whether to make its framework compatible with concurrent firms’ applications. The information technology market, with its rapid growth and innovation, has put a new emphasis on the concept of compatibility and standardization.
MARKET SEGMENTS
APPLICATIVE DOMAIN
COMPATIBLE COMPONENTS application C
application B
ta i ns
nce
s
application A
Figure 1: Our scenario. One of the principal reasons for this can be found in the evolution that the role of computers has undergone: from task automation and speed-up tools to communication and information providers. Computer and software industry has driven most of the recent studies in formal standardization; however, these efforts have not been supported by an adequate theoretic work on the private and social effects of compatibility decisions. 3.1 The Reference Approach Most of the existing researches focus on identifying the equilibrium conditions that held in the market when the compatibility is the variable in the analysis. Private and social incentives and costs are considered as evaluation parameters in the compatibility game. This paper does not examine the social optimality of the firms’ compatibility decisions, but focus on the firms’ incentives to produce compatible products and the related conscious decision that they undertake. BB
B
B BB AA
A AA
A
Incompatibility AB
A AA
BB B
BA
B
B
A
AB
BB
AA
BA A
AB
BB
AA
BA
Compatibility
Figure 2: Equilibrium Market Configurations (from Matutes and Regibeau [6]).
In the framework first explored by Katz and Shapiro [5] and Farrell and Saloner [3], a system composed by compatible components is treated as a single good characterized by positive consumption externalities. These authors model firms as selling a single good, assuming implicitly that every firms sells every component of a system, identified as the single good useable by customers, or that manufacturers sell a component to customers who already own the rest of the system. The last is, for example, the case of software which is usually sold without the computer that the consumer already own. A different approach is proposed by Matutes and Regibeau [6,7]: they show that additional considerations concerning compatibility issues can be achieved by explicitly modeling a system as a set of related components, instead of treating it as a single good. Firms sell complete systems, but it may be possible to sell single components, allowing the use of systems which combines components produced by different companies. Following the approach proposed by Matutes and Regibeau, the model can be restricted to the consideration of two independent firms which sell a functionally equivalent product composed by two cooperating modules. Both firms produce the two modules of a product, and the modules produced by one firm are differentiated from the modules produced by the other firm. The marginal costs for the production are set equal to zero, and there are no economies of scope. Figure 2 (from Matutes and Regibeau [6]) represents the situation: the two firms can be placed in the corners of a square (with side length normalized to one). If the two modules sold by the two firms are not compatible, then consumers can chose only between the two pure system proposed by the two firms. Considering the modules compatible, it is possible to combine a component sold by the first firm with the complementary component from the second firm. As a result, consumers can purchase one of four possible systems, the two pure systems and the two hybrid system obtained combining one module of the first firm with one module of the second firm. Distributing the consumers on the unit square in a uniform way, the first firm, named A, is placed in the lower-left corner, while the second firm, named B, is placed in the upper-right corner, with coordinates (1,1) for the normalization. Normalizing the user preference for a module of a firm in the range [0,1], a user placed at a point of coordinates (x, y) in the square has a preferred first module that is x away from the first module of firm A, and a preferred second module that is y away from the second module sold by firm A. A complementary situations, with distances 1-x and 1-y, holds concerning firm B. The reservation price, R, is the price a customer is available to pay for purchasing its ideal system. The customer surplus is the hypothetical saving a customer has buying an existing technology. The customer surplus, S, is difference of the reservation price and the hypothetical cost of the technology. The hypothetical cost of the technology is the sum of the real price, P, and the hypothetical cost of the distance of the purchased technology and the ideal technology, D. That is: S=R-P-D A customer purchases a technology if s/he feels that the surplus is positive. If both the technologies provide the user a positive surplus, s/he will purchase the cheaper. The question is: when is it convenient to provide compatibility for a firm, (1) to maximize S for its products but also (2) to avoid the competitor from taking hostile advantages from its compatibility? The worst situation for a firm would be to spend money to make a product compatible and to have its competitor taking the full advantage from the compatibility.
We need to study the structure of S, that is, of R, P, and D. 3.2 Our Hypotheses We assume the following. 1. For software firms there are always economies of scales. Once the first instance of a product is produced, the other copies are almost for free. 2. P depends on the absolute time, t. P depends on the cost to produce the component. As from Katz and Shapiro [5], the cost is strictly decreasing in t. Therefore, P is strictly decreasing, too. 3. The reservation price depends both on the value of the intrinsic good and the value of the network. 4. The intrinsic value of a good is independent of time. 5. The network evolves with the absolute time. 6. Network externalities are important in the software industry, since communication, data sharing and coordination are widely present. 7. The bandwagon effect is strong. From 1. and 2., we can model the behavior of the price of a component as: P(t) = p e-x(t-t’). p depends on the cost of the resources to produce the product; x depends on technological factors of the application domain; t’ is the time of introduction of the product in the market. From 3., the reservation price has the shape of: R(t,n) = Ri(t) + Rnet(n) where Ri(t) is the intrinsic value of the good, the value that the customer associates directly to the good as a stand-alone entity. Rnet(n) is the value given to the product by the presence of a network made up by n consumers. From 4., we assume that the intrinsic value of the good Ri(t) is independent from the absolute time: Ri(t) is Ri. We assume also that Rnet(n) is linearly related to the size of the network. Ri(t) = Ri Rnet(n) = rn From 5., the network evolves with time. As from the literature, we assume two possible evolutions, a linear evolution, and an exponential evolution: n(t) = h (t – t”) [linear evolution] n(t) = b eg(t-t”) [exponential evolution] t” is the time when the good is purchased for the first time. The constants b, h, and g may be positive or negatives, depending of the lifecycle phase of the object. This implies that: Rnet(n) = r n(t); therefore, Rnet(t) = h (t-t”) or Rnet(t) = b eg(t–t”)
This evolution occurs while the network is not saturated. Once saturated, a constant period may occur or a negative trend may start. For simplicity, from now on, we assume t’=0 and we define a new t" as t’-t". Figure 3 details the behavior of Rnet; for space constraint the time of constant value is omitted. Shape 1 represents the initial behavior of the reservation price as a function of the time. Point M represents the maximum asymptotic reservation price achievable considering the network maximum dimension. The coordinates (F,L) indicates the change point from the ideal behavior and the step to the shape 2. At the growing of the network size, a new effect appears (shape 3): positive network externalities effects begin to grow less and then to decrease due to possible congestion of the network (this is, for example, the case of Internet when there are too many users connected at the same time). The point at coordinates (N,S) marks the step to this situation, and at the point (Ri, E) there is a perfect equilibrium between positive network externalities effects and network saturation effects. From 6., and 7. we focus mainly on the exponential evolution. Moreover we focus in this paper on the growing pattern of the curve. Rnet
M N 2 3
F 1 Rn(d) Ri
t-t” L
S
E
Rnet
M N 1
2 3
F
R n(d)
Ri
t-t” L
S
E
Figure 3: Evolution of Rnet. 3.3 The Model From our assumptions the customer surplus is expressed by: S(t) = Ri + b eg(t-t”) - p e-xt - D All the constants are positive. A customer will choose the technology with the highest value of S, provided that it is positive. Otherwise it will not buy any product. -xt
S(t) = Ri + b eg(t-t”) - p e –D =
-xt = (Ri) + (- D - p e ) + (b eg(t-t”))
There are three components in the surplus. A value in the ideal product, (Ri) , a value in the real product, (D - p e-xt), and a value coming from the network (b eg(t-t”)). The value of the ideal product is fixed in our analysis. The value of the real product depends on the compatibility decisions. From Figure 2 it is evident that compatibility decisions may help both competitors, since they are able to sell products to customers that would not buy their own product individually. The value coming from the network is also strongly influenced by compatibility decisions. t" plays an important role: it is the time of the first introduction of the product in the network. Firms want to lower t" as much as they can. We can express the profit of a firm, Q, as: +¥
Q = å diP(i) - C i = t"
Where di is the number of firms buying the product P at instant i, C is the cost for producing the product. That is the profit is the sum of the purchase price of the firm and the cost to produce the product, that is constant, since we assume marginal costs are zero. A firms increases its profits if it increases the number of customers, that is, if it maximizes: +¥
åd
i
i = t"
Therefore the profits of a firm Q is strictly and monotonically related to the customer surplus. A firm could also increase its profits increasing the price, P(i) in the equation; however, if a firm increases its prices, a lower number of customers will buy its products, therefore lowering the global value of Q. This can be accomplished through different ways: 1. A firm need to enter the market, i.e., to have t" lower than +¥. 2. A firm t has to reduce the distance with the ideal product D. In these two cases compatibility is a key feature, since it allows a higher surplus due to network externalities of incumbent products. 3. A firm has to enlarge its market share. There are two different cases, according to our hypotheses on R and P. Initially compatibility supports synergy of incumbents. However the fast evolution of the market makes compatibility decision hard. The entire problem in this situation is the understanding of the value of g and x. Unfortunately the writers have not been able to run suitable simulations, so the whole discussion is only qualitative. g and x are two different constants: g refers to the growing pace of the network, while x depends on the ability of the firm to discount the initial cost of its product. However they are related. Static situations are characterized by low values of them. Dynamic situations are characterized by high values of them. Static situations occur when few large customers are present, the market is captive, personal relationships play a relevant role. Dynamic situations occur when a lot of small customers. Unfortunately we are not able to provide any quantitative definition of what is static and what is dynamic.
4 DISCUSSION
From our model, it seems that compatibility is not a good or a bad property of a software product. Its value depends mainly on the new product and on the market structure. We support our view with a brief analysis of few possible scenarios. Scenario 1. A firm entering a completely virgin, narrow market. The firm is the first one in the market. It can take advantage of its position. Any other firm entering the market will have to face an already installed network. Compatibility is not advisable. Scenario 2. A firm entering a complete virgin, wide but dynamic market. The firm cannot satisfy all the expectations from the habitants. However it may expect that the benefits coming from network externalities will overcome the gap soon, thus making compatibility not desirable. Scenario 3. A firm entering a complete virgin, wide but static market. The firm cannot satisfy all the expectations from the habitants. If another firm enters the market, a profitable situation for both firms may occur. Compatibility enables such situation. The two products from the two firms can work together to satisfy the more distant customers. However, due to the exponential nature of the curve, the firm should prepare for a competition with the other firm, on the long run. Strategy for making a compatible product not compatible should be devised. Scenario 4. A firm entering an existing and populated market. Compatibility is the way to go. Firms which have not or have a little installed base (or suffers of little consideration from customers), are usually oriented toward the production of modules compatible with bigger firms’ ones. The firm should try as hard as possible to take advantage of the existing network. It is worthwhile to notice that the European Union legislation permits a firm to disassemble somebody else’s product to make its own product compatible with it. A comprehensive discussion of this topic can be found in [8]. Scenario 5. A firm keeping a leading role in a populated market. Compatibility does not carry any advantage, because of the firm already has a large installed base and a good consideration from customers. In the software and hardware market, technology play a fundamental role: due to the rapid evolution of hardware and software products, consumer network are highly dynamic and difficult to estimate. This situation is thicken by the strong correlation existing among hardware, software and applications evolution. New technologies introduced in the market, having great intrinsic potential but incompatible with the actual technologies, could strand the competing networks. 5 CONCLUSIONS AND FURTHER RESEARCH
Compatibility decisions influence the success of a product. The answer to the compatibility dilemma depends on multiple factors: the size of the network, the number and the kinds of incumbents, the technology employed and so on. In this paper we have presented a simple model to represent the role of compatibility in software production. The specificity of software firms has been captured through a sequence of working hypothesis. We have conducted the analysis from the firm’s profit standpoint; we have not considered social welfare. It appears that the high dynamics of the software industry makes compatibility advantageous mainly to new firms entering a wide virgin market or a populated market. In the other cases incompatibility seems the way to go. Several questions arise from this research. It would be relevant to perform simulations of the behavior of incumbents when a compatible product enter an existing dynamic market, and of the long-term compatibility behavior of incumbents in a static market. Tools and methods to build a product that is initially compatible and then can be transformed to incompatible would also be extremely significant.
REFERENCES
1. Boehm, Barry B. (1984): Software Engineering Economics, IEEE Transactions on Software Engineering, Vol. SE-10, No. 1, January 1984, pages 4-21 2. Economides, Nicholas (1989): Desirability of Compatibility in the Absence of Network Externalities, American Economic Review, Vol. 79, No. 5, pages 1165-1181, 1989 3. Farrell, Joseph, and Garth Saloner (1985): Standardization, Compatibility, and Innovation, Rand Journal of Economics, Vol. 16, No. 1, Spring 1985 4. Farrell, Joseph, and Garth Saloner (1986): Installed Base and Compatibility: Innovation, Product Preannouncements, and Predation, American Economic Review, Vol. 76, No. 5, December 1986 5. Katz, Michael L., and Carl Shapiro (1985): Network Externalities, Competition, and Compatibility, American Economic Review, Vol. 75, No. 3, June 1985, pages 425-440 6. Matutes, Carmen, and Pierre Regibeau (1988): "Mix and Match": Product Compatibility without Network Externalities, Rand Journal of Economics, Vol. 19, No. 2, Summer 1988 7. Matutes, Carmen, and Pierre Regibeau (1992): Compatibility and Bundling of Complementary Goods in a Duopoly, Symposium on Compatibility, Journal of Industrial Economics, Vol. XL, No. 1, March 1992 8. Succi, Giancarlo, Gianpiero Succi, and Marco Ronchetti (1997): Legal Issues regarding Software Use and Reuse within the European Union Legislation. Journal of Computing and Information Technology.(In publication in 1997)
