Business Strategy and Business Model: Extending the Strategy

Embed Size (px)

Citation preview

  • 1.Business Strategy and Business Model: Extending the Strategy-Structure-Performance ParadigmbyC. Zott and R. Amit 2004/84/ENT/SM/ACGRD 8Working Paper SeriesINSEAD-Wharton Alliance Center for Global Research & Development

2. Business Strategy and Business Model: Extending the Strategy-Structure-Performance Paradigm*Christoph Zott INSEAD Euro-Asia Center 006Boulevard de Constance 77305 Fontainebleau CedexFRANCE Telephone: 33 1 6072 4364 Fax: 33 1 60 72 42 23E-mail: [email protected] AmitThe Wharton School University of Pennsylvania3620 Locust walkPhiladelphia, PA 19104-6370 Telephone: (215) 898-7731Fax: (215) 573-7189E-Mail: [email protected] 6 August 2004 * Both authors contributed equally to this article. We gratefully acknowledge the financial support of the Wharton-INSEAD Alliance Center for Global Research & Development. Christoph Zott acknowledges support from the Rudolf and Valeria Maag Fellowship in Entrepreneurship at INSEAD. Raffi Amit acknowledges financial support from the Wharton e-business Research Center (a unit of WeBI) and the Robert B. Goergen Chair in Entrepreneurship at the Wharton School. We thank Iwona Bancerek, Amee Kamdar, Jenny Koelle and Gueram Sargsyan for valuable research assistance. 3. Business Strategy and Business Model:Extending the Strategy-Structure-Performance ParadigmABSTRACTWe extend the strategy-structure-performance paradigm to highlight the interdependencies among afirms product market strategy, the structure of transactions it enables with external stakeholders (itsbusiness model), and its performance. By developing a formal model, we are able to examinetheoretically the contingent effects of product market strategy and business model design on firmperformance. Using a unique manually collected data set, we find that novelty-centered businessmodel design, coupled with product market strategies that emphasize differentiation, cost leadership,or early market entry, enhances firm performance. KEYWORDS: Product market strategy, business model, performance, strategy-structure- performance paradigm, competitive strategy. 2 4. Business Strategy and Business Model: Extending the Strategy-Structure-Performance ParadigmThe Strategy-Structure-Performance (SSP) paradigm addresses linkages between corporatestrategy (e.g., the extent of diversification) and administrative structure (e.g., divisional versusfunctional forms) of firms (Chandler 1962). In this paper, we extend this paradigm by examining theinteraction between a firms product-market positioning choices and its business model design, whichcenters on the architecture of the firms transactions with external stakeholders. While research on theSSP paradigm has focused largely on the administrative structure of a firm within its boundaries, weare concerned with structures as architectural designs of transactions that span firm and industryboundaries. Specifically, we ask if business model design is distinct from a firms product-marketpositioning and how product market strategies and business model designs interact to affect firmperformance.Consider, for example, the case of Priceline.com Inc., a provider of an electronic pricingsystem, known as a demand collection system. Using the internet, it enables consumers to save moneyon a wide range of products and services by trading flexibility regarding brands, product features,timing, convenience and/or sellers in return for being able to buy products and services at prices thatare lower than those charged through traditional retail distribution channels. Further, Priceline enablessellers to generate incremental revenue by disposing of excess inventory or capacity at prices that arelower than the ones they offer through other channels while protecting their brand. Overall then, theproduct-market strategy of Priceline is cost leadership. However, Priceline's business model is novel,and thus innovative. At its core is a reverse market auction pricing mechanism for which the companyhas secured a business method patent. It allows the customer to name the price at which they wish totransact and the company will attempt to find a provider of the product or service within a specified3 5. range. So how does the cost leadership product market strategy of Priceline interact with its novelty-centered business model design to affect its performance? Is there fit among these constructs?By developing a formal model, we examine theoretically the interdependencies among salientdimensions of product market strategy, different business model design themes, and firmperformance, thereby answering these questions. Using our unique data set on business strategy andbusiness models, we establish empirically that a firms product market strategy and the design of afirms business model are distinct constructs that individually and jointly affect firm performance.Specifically, we find that novelty-centered business model design, coupled either with adifferentiation or cost leadership strategy, enhances firm performance. In addition, we ascertain that anovelty-centered business model design joined with early entry into a market positively affectsperformance.This study attempts to make the following contributions: First, it extends the perspective ofthe SSP paradigm on structure, from being concerned with the administrative structure of the firm to afocus on the architectural structure of transactions the focal firm enables with external stakeholders.As well, the study shifts the focus of the SSP paradigm from corporate to business strategy. Second,in this paper we argue theoretically, and validate empirically that the business model construct is validand distinct from received notions of a firms product market strategy. Third, we offer a formal modelthat allows us to articulate how interactions among our main constructs are expected to affect firmperformance. Fourth, we test these theoretical developments empirically, and show that novelbusiness model design can augment the competitive advantage realized through superior productmarket strategy. In other words, both product market strategy and structure as embodied by the designof the business model can enhance the firms competitive advantage, independently as well as jointly.The remainder of the paper is organized as follows: We proceed in the next section to presentour theory and model, after which we explain the data and methods we used to test it. We then presentour results, and we conclude with a discussion of our findings and implications for future research.4 6. THEORYStrategy and StructureIn his study of large American corporations and their approaches toward product-marketdiversification, Alfred Chandler (1962) observed that major increases in volume, geographicdispersion, and vertical and horizontal integration of firms were followed by changes in theiradministrative activity, which eventually led to the emergence of the M-form of organization. Thatline of reasoning, however, provoked the counterargument that strategy follows structure" (e.g.,Bower 1970), which was predicated on the logic that managerial cognition and skills mediate betweenstructure and strategy. The ensuing debate in the strategy literature on the relationship betweenstrategy, structure, and firm performance known as the SSP debate flourished in the 1970s and1980s, but (with few exceptions, e.g., Amburgey and Dacin 1994 and Yin and Zajac 2004) seems tohave received less attention recently.Some management scholars have pointed to the relatively narrow focus of the traditional SSPdebate on corporate strategy and administrative firm structures, which might limit the applicability ofSSP theory to large, established corporations, and which might also explain why some researchersseem to consider the subject as settled. However, there might be much more to the concepts of bothstrategy and structure (Miller 1986: 233). In this paper, we propose to reopen -- and enrich -- the SSPdebate by examining business strategy instead of corporate strategy, and by focusing on the firmsbusiness model instead of its administrative structures.In terms of business strategy, we focus on some salient aspects of a firms product marketstrategy. We view product market strategy as the way in which a firm chooses to build, exploit, andsafeguard advantages in its addressable market spaces by making the following decisions: (1) Whattype of product market positioning approach to adopt (i.e., cost leadership and/or product/servicedifferentiation; see Porter 1985); and (2) When to enter the market (Lieberman and Montgomery1988). The answers to these questions are central to our understanding of how firms that operate incompetitive product markets create and appropriate value.5 7. Towards a New Perspective on StructureThe received literature on the strategy-structure debate has focused on administrativestructures within firm boundaries, and paid little attention to extending the question ofstrategy/structure fit issues for other structural forms of organization (Yin and Zajac 2004: 365). Yet,rapid advances in information and communication technologies have enabled new types oftechnology-mediated interactions between economic agents (Geoffrion and Krishnan 2003). Inparticular, such new information technologies as the internet and broadband wire- and wirelesscommunication technologies have facilitated the design of novel architectures of economicexchanges, thus enabling firms to change fundamentally the way they organize and transact, bothwithin and across firm and industry boundaries (Mendelson 2000). Consequently, new opportunitieshave emerged for the creation of organizational arrangements among firms, partners, and customers.On-line communities for open source software development are a case in point (Lee and Cole 2003).These developments have facilitated a shift in the locus of value creation beyond theboundaries of the firm (Gulati, Nohira and Zaheer 2000; Normann 2001). For example, a new type ofrelational rent may accrue to alliance partners (Dyer and Singh 1998). Consequently, to betterunderstand value creation and capture by firms, a broader conceptualization of organizationalboundaries beyond the legally relevant demarcation of the firm from its environment is needed toencompass, for example, the exchange relationships that exist between the firm and multiple externalparties (Santos and Eisenhardt Forthcoming).The holistic design of an organizations boundary-spanning economic exchanges can bedefined as the firms business model. The business model is the template of how a focal firm interactsand transacts with customers, partners, and vendors; it refers to the overall gestalt of these possiblyinterlinked transactions. The design elements of the business model can then be defined as thestructure, content, and governance of transactions between the focal firm and its exchange partners6 8. (Amit and Zott 2001: 511). Thus, the business model serves as a conceptualization of the architecturallinks between the firm and its ecosystem.1In describing business models, Christensen, Verlinden and Westerman (2002) distinguishbetween two design principles for business model architectures vertical integration and horizontalstratification. A business model can also be characterized by design themes, which capture commonthreads that orchestrate and connect its design elements.In particular, we distinguish betweennovelty-centered and efficiency-centered business model designs. Novelty-centered businessmodel design refers to new ways of conducting economic exchanges among various participants (seethe example of Priceline). Efficiency-centered business model design refers to the measures firmsmay take with the intention to achieve transaction efficiency, i.e., reduce transaction costs for allparticipants; it does not refer to the outcome (i.e., efficiency) itself. An example of an efficiency-centered design element would be the order-tracking feature in Amazons business model, which isaimed at enhancing transaction transparency, and thus at increasing efficiency. These design themesare neither orthogonal (for instance, novel design elements may engender lower transaction costs), norare they mutually exclusive: Both may be present in the design of any given business model.Moreover, the design themes are not exhaustive as there may be other themes present in the design ofa business model. The design themes describe the holistic gestalt of a firms business model, and theyfacilitate its conceptualization and measurement (Zott and Amit 2003).The business model can be a source of competitive advantage that is distinct from the firmsmarket position (Christensen 2001). In Table 1 we contrast business model and product marketstrategy. 1 We note that the business model construct is distinct from the value net strategic analysis frameworkdeveloped by Brandenburger and Nalebuff (1996). In the value net the unit of analysis is the firmwhereas the business model considers the eco system of the focal firm as the unit of analysis. Theplayers in the value net such as competitors and certain complementors may or may not be part of thebusiness model because some of these players may not transact with the focal firm. 7 9. [INSERT TABLE 1 HERE]Examining Table 1 leads us to conclude that product market strategy differs from the businessmodel mainly through its focus on the positioning of the firm vis--vis rivals, whereas the businessmodel centers more on the architectural design (i.e., structure) of transactions with customers,suppliers and partners so as to enable economic exchanges among parties in the firms addressableproduct markets. This leads us to establish the following corollary.Corollary: Business model design (as, for example, measured by design themes) isdistinct from product market strategy (as, for example, measured by generic strategies).Recent research has emphasized the contingent nature of the relationship between strategyand structure, suggesting a moderating, rather than a mediating, effect of these constructs on firmperformance (Mintzberg 1990; Siggelkow and Levinthal 2003). This research has highlighted theusefulness of examining interactions between salient dimensions of strategy and structure on firmperformance. Hence, we explore the contingency relationship between a firms product marketstrategy and the structural design of its business model. A core tenet of contingency theory is thatalignment between contingency factors, such as business model and product market strategy, mightresult in higher performance. This is also reflected in SSP research, where alignment betweenadministrative structure and firm strategy is argued to have positive implications on firm performance(e.g., Chandler 1962; Siggelkow and Levinthal 2003).Fit Between Product Market Strategy and Business Model DesignTo evaluate the interactions between business model design and product market strategy interms of their implications on firm performance, we consider two main business model design themes novelty-centered and efficiency-centered business model design (Zott and Amit 2003) along withtwo generic product market positioning strategies cost leadership and differentiation (Porter 1985) as well as the timing of entry into a market (Lieberman and Montgomery 1988).Which design of a firms business model fits best with its choice of product market strategy?In other words, what constitutes a good fit between these constructs? The literature on fit generally8 10. considers coherent configurations of design elements as good fit that manifest themselves as peaks inthe performance landscape (Siggelkow 2001). Concretely, two design elements (A and B) fit well ifcomplementarities exist between them, that is, if the marginal benefit of A increases with the level ofB, and if the levels of A and B are adjusted optimally to achieve a local performance optimum(Milgrom and Roberts 1995).We next introduce a formal model that allows us to investigate which combinations ofbusiness model design themes and product market strategies fit well. Building on the conceptualframework developed by Brandenburger and Stuart (1996) for value creation in a simple static settingwith one firm, one customer, and one supplier, the total value created by a business model can beexpressed as the sum of the values created for all the participants in a business model, over alltransactions that the business model enables in a given time period. More formally, drawing onBesanko, Dranove and Shanley (1996), let m be an index ranging from 1 to M, with M denoting thetotal number of market segments served by a focal firm through its business model, and let Pm(t) bethe price that a homogeneous customer from segment m pays for a good acquired in transaction t, orfor the right to participate in the transaction. Furthermore, let Bm(t) denote the customers perceivednet benefit from participating in t. Bm(t) is net of the transaction, purchasing, and user costs thataccrue to the customer (Besanko et al. 1996: 443). It can be thought of as the customers willingness-to-pay. Consequently, the value created for a customer in transaction t can be written asVm(t) = Bm(t) - Pm(t)(1)The focal firm has adopted a business model of design type d, where d is a vector describing theextent to which the business model emphasizes the design themes novelty and efficiency. As well, it hasadopted a product market strategy s, where s is a vector describing the extent to which the firmemphasizes differentiation, cost leadership, and entry timing. For simplicity, denote that firm as Fds F.Denote the focal firms suppliers and partners (other than customers) as i, where i is an index rangingfrom 1 to I, the total number of suppliers and partners in the business model. Let Ri(t,m) be the revenuesthat focal firm F gets from partner i in a particular transaction, t, involving a customer from segment m,let Ci(t,m) denote the flow of revenues from F to i, and let OCF(t,m) be Fs opportunity costs for 9 11. providing its own resources. Then the value created for firm F in transaction t involving a customer fromsegment m can be expressed as VF(t,m) = Pm(t) + i Ri(t,m) - i Ci(t,m) - OCF(t,m)(2) Let us furthermore denote the opportunity costs of supplier or partner i of supplying resources(including, for example, complementary products or services) to F as OCi(t,m). Then the value createdfor partner i in transaction t with a customer from segment m can be written as Vi(t,m) = Ci(t,m) - Ri(t,m) - OCi(t,m)(3) Assuming that the set of stakeholders in a business model comprises the focal firm, itscustomers, suppliers, and partners, and given that the total value created by the business model equalsthe sum of values created for all business model stakeholders, it follows that the total value created intransaction t is TV(t,m) = Vm(t) + VF(t,m) + i Vi(t,m)(4) Inserting (1), (2) and (3) into (4) yieldsTV(t,m) = Bm(t) - OCF(t,m) - i OCi(t,m) (5) which is a generalized version of Brandenburger and Stuarts (1996) formula for total valuecreated. However, equation (5) focuses on a particular transaction, t, with a particular market segment,m, rather than on a particular product or service. Finally, the total value created through a businessmodel, TVC, is the value created over all market segments m, and over all the types of transactions t thatthe business model enables, where t is an index ranging from 1 to T, and T denotes the number oftransaction types. n(t,m) is the average number of transactions of type t conducted with customers fromsegment m: TVC = t m [TV(t,m)*n(t,m)](6) Inserting (5) into (6) yieldsTVC = t m {[Bm(t) - OCF(t,m) - i OCi(t,m)]*n(t,m)}(7) Similarly, the total value appropriated by the focal firm, TVA, can be expressed as TVA = t m [VF(t,m)*n(t,m)](8) Inserting (2) into (8) yields10 12. TVA = t m {[Pm(t) + i Ri(t,m) - i Ci(t,m) - OCF(t,m)]*n(t,m)} (9) TVA as a proxy for firm Fs performance is contingent on Fs business model design, d, andits product market strategy, s. If d and s are choice variables of the firm, their impact on each term ofthe right hand side of equation (9) must be considered to understand their collective impact on TVA.Following Siggelkow (2002), a useful thought experiment for evaluating the fit between a particularbusiness model design theme and a particular product market strategy is to consider whether themarginal value of a the business model design theme would be affected (in particular, whether itwould increase) if a firm were to put more emphasis on the product market strategy. This thoughtexperiment is consistent with the definition of fit as indicative of complementarity (Milgrom andRoberts 1995; Siggelkow 2001). Hence, we proceed to explore the marginal effects of business modeldesign themes and product market strategies on TVA. Novelty-centered design and TVA. Novelty-centered business model design refers to theconceptualization and adoption of new ways of conducting economic exchanges among transactionparticipants. Novelty primarily aims at creating new types of transactions, i.e., increasing T, but alsoat addressing new market segments, i.e., raising M. It also strengthens the focal firms bargainingpower vis--vis other business model stakeholders (Zott and Amit 2003). Consequently, strongeremphasis on novelty-centered design will have a positive effect on Pm(t) and will exert downwardpressure on Ci(t,m) due to the increased bargaining power of the focal firm. Hence, we observe that amarginal increase in a firms emphasis on novelty-centered design may affect TVA in equation (9)through T (+), M (+), Pm (+), and Ci (-).2 We next examine the marginal effect on TVA of changing a particular product marketstrategy, followed by an analysis of the impact of such a change on the marginal value of novelty-centered business model design on TVA. 2 The sign in brackets gives the expected direction of change in the respective variable from amarginal increase in novelty-centered business model design. E.g., T(+) means that T increases innovelty-centered design. 11 13. First, consider product market differentiation. A stronger emphasis on differentiation willpositively influence customers willingness-to-pay, Bm(t), and therefore make it easier for the focalfirm to charge higher prices to customers, Pm(t), and possibly to lower the costs of suppliers, Ci(t,m)(Porter 1985). Hence, a marginal increase in a firms emphasis on differentiation may affect TVAthrough Pm (+), and Ci (-). In addition, a focus on innovation in multiple domains (business model,product market strategy) may help firms create a consistent image, thus increase n, and allow foreconomies of scope, thus decrease OCF. Moreover, a firm that focuses all its activities andtransactions on innovation may become an even more skillful innovator over time, thus decreasingOCF even further (Zott 2003). Hence, we expect a positive joint effect on TVA.Second, consider cost leadership. A stronger emphasis on cost leadership implies lower pricescharged to customers, Pm(t), as well as lower input and production costs, Ci(t,m) and OCF(t,m) (Porter1985). Furthermore, new low-cost customer segments can be addressed, thus raising M, andcustomers within given segments will be motivated to increase their number of repeat transactions,thus raising n. In other words, a marginal increase in a firms emphasis on cost leadership may affectTVA through Pm (-), Ci (-), OCF(-), M (+) and n (+). Moreover, a greater emphasis on cost leadershipwill also enhance the marginal effect of novelty-centered business model design on TVA. First, amore pronounced cost leadership approach interacts positively with the firms strengthenedbargaining power over its suppliers through increased novelty-centered design as it puts additionaldownward pressure on Ci. Second, customers in the new market segments will now have two motivesto be drawn to the firm - novelty and low cost - which enhances the positive impact of novelty-centered design on M. Therefore, we expect a positive joint effect of cost leadership and novelty-centered business model design on TVA.Third, consider timing of market entry. Firms that enter markets earlier may enjoyconsiderable advantages. These stem from the creation of customer switching costs, brand awareness,and reputation, thus allowing these firms to charge higher prices, Pm(t) (Lieberman and Montgomery1988). Early market entrants can also gain by learning and accumulating proprietary knowledge andby pre-empting scarce resources, thus lowering their opportunity costs, OCF(t,m) (Lieberman and12 14. Montgomery 1988). As the early entrant attempts to address and perhaps even create a newmarket, the number of transactions, n, is likely to be limited. In other words, a marginal increase in afirms emphasis on early market entry timing may affect TVA through Pm (+), OCF(-), and n (-). Inaddition, moving into a market earlier allows the firm to capture the rents from business modelinnovation, which can be considered entrepreneurial rents, i.e., rents that accrue between theintroduction of an innovation and its diffusion (Rumelt, 1987). In particular, the positive effect ofnovelty-centered design on Pm may be more pronounced, and more sustainable if the firm enters amarket early. Hence, we expect a positive joint effect on TVA.In summary, the above analysis of novelty-centered business model design suggests thatcoupling a novelty-centered design with a product market strategy of either differentiation, costleadership, or early market entry represents good fit.Efficiency-centered design and TVA. Efficiency-centered business model design aims atreducing transaction costs for all transaction participants. This explains the likely negative effects of amarginal emphasis in such design on OCF(t,m). By reducing transaction costs, efficiency-centeredbusiness model design may also lead to higher transaction volume, n(t,m); more new customers willbe drawn to transact with the focal firm, and existing customers may transact more frequently as aresult of the lowered transaction costs. Hence, a marginal increase in a firms emphasis on efficiency-centered business model design may affect TVA through OCF (-), and n (+).3Next, to evaluate the possible fit between efficiency-centered business model design andparticular choices of business strategy, we examine whether the marginal value of efficiency-centereddesign would increase if a firm were to put more emphasis on a particular product market strategy.First, consider the strategy of differentiation. As shown above, a marginal increase in a firmsemphasis on product market differentiation may affect TVA through Pm (+), and Ci (-). It is not clearper se, however, whether and how product differentiation would affect the marginal benefit of 3 Again, the sign in brackets gives the expected direction of change in that variable from a marginalincrease in the degree of efficiency-centered business model design.13 15. efficiency-centered business model design through OCF, and n. Hence, the joint effect ofdifferentiation and efficiency-centered business model design on TVA is expected to beindeterminate. Second, consider cost leadership. We have seen that a marginal increase in a firms emphasison cost leadership may affect TVA through Pm (-), Ci (-), OCF(-), n (+) and M (+). In addition, a focuson low costs in multiple domains (business model, product market strategy) may help firms create aconsistent image, thus increase n, and allow for economies of scope, thus decrease OCF. Moreover, afirm that focuses in all its activities and transactions on cost reductions may become even moreskillful at reducing costs over time, thus decreasing OCF even further. Hence, we expect a positivejoint effect of cost leadership and efficiency-centered business model design on TVA. Third, consider timing of market entry. As shown above, a marginal increase in a firmsemphasis on early market entry timing may affect TVA through Pm (+), OCF(-), and n (-). However, itis not clear per se whether and how early market entry timing would affect the marginal benefit ofefficiency-centered business model design. We expect, therefore, that the joint effect of early marketentry and efficiency-centered business model design on TVA is indeterminate. In summary, the above analysis of efficiency-centered business model design suggests thatcoupling an efficiency-centered design with a product market strategy of cost leadership representsgood fit, whereas the fit with either a product market strategy of differentiation or with early marketentry cannot be clearly predicted.DATA AND METHODSSample We collected data on a sample of Internet-related firms that had gone public in Europe or inthe U.S. between April 1996 and May 2000. This sample selection strategy enabled us to create a dataset of about 300 firms that conducted part of their business over the Internet, and hence served asfertile ground to investigate business model designs. We randomly sampled 170 firms on theirbusiness model design characteristics and product market strategies. We considered public companiesin order to ensure the availability and consistency of the data. Data collection from initial public14 16. offering documents is an acknowledged method for studying firms strategies (e.g., Dowling andMcGee 1994).Data CollectionThe data collection proceeded in two stages. We first built composite scales for businessmodel design themes, and we identified and measured relevant items on the basis of a content analysisof IPO prospectuses. We then followed a similar procedure to ascertain and measure relevantdimensions of product market strategies.To determine the business model design themes, we built measurement scales as none werereadily available in the literature. This process proceeded in five stages: (1) development of surveyinstrument, (2) development of measurement scales, (3) pre-testing of survey, (4) development ofonline web interface and of central database, and (5) data collection. In collecting the data, we built onthe use of expert panelists in management research (see, for example, MacCormack, et al. 2001). Wehired eleven part- or full-time research assistants (primarily MBA students), and we trained them asexpert raters to analyze assigned sample companies. We carefully selected our raters and trained themin data collection and data analysis. On average, it took a rater about two days to collect data on agiven business model, to understand the model, and to assess it. Data sources included IPOprospectuses primarily (Dowling and McGee 1994), but also, annual reports, investment analystsreports, and web sites. The data were collected from May 2000 to June 2001. During that time period,we were able to take one measurement of the design themes for each of the business models in oursample.We validated inter-rater reliability by assigning a randomly chosen business model to twodifferent expert raters (each of whom was assigned to a different project manager), and by conductinga pair-wise comparison of responses, yielding a Cronbach alpha of 0.81, and a Pearson correlationcoefficient of 0.72. Raters were in broad agreement with each other for 82% of the individual items.We repeated the test periodically, and we found that all indicators of reliability further improved.Regarding the product market strategy scales, we first consulted with the strategy andmanagement literatures which aspects of product market strategy should in theory affect the 15 17. performance of a firm (i.e., product market positioning through differentiation and cost leadership,timing of market entry), and which measurement scales had been used in previous research for thesevariables. We found that most of the empirical work on Porters (1985) generic strategies, forexample, had been conducted on the basis of surveys administered to managers (e.g., Miller 1988). Afew researchers (e.g., Dowling and McGee 1994) have used IPO prospectuses to measure these items.We then adapted these survey-based instruments so as to analyze the content of our primary datasource.We iteratively selected a set of items to measure product market strategy dimensions. Afterpilot-testing these items on sample firms, we refined some items and dropped others, mainly on thebasis of data availability. Starting with 51 items derived from the literature that measured variousaspects of generic firm-level strategy, we retained five items referring to differentiation and four itemsreferring to cost leadership. We also retained a single-item measure for market-entry timing. Tworaters then used these measures to analyze independently all 170 firms in our sample. Inter-raterreliability was established by conducting a pair-wise comparison of responses for five randomlychosen firms, yielding a Cronbach alpha of 0.92, and a Pearson correlation coefficient of 0.91. Raterswere in exact agreement with each other on 77% of the individual items (on a five point scale). Allinitial differences were resolved through discussions, so the final agreement percentage was 100%.Econometric Modelling and Estimation ApproachWe conducted a confirmatory factor analysis and a partial least squares regression analysis inorder to establish the discriminant validity of our business model and product market strategyconstructs. We then proceeded to analyze the data using multivariate regression techniques. Weconfirmed that conventional assumptions underlying OLS regression analysis held in our data set.First, after performing a logarithmic transformation of our dependent variable, we found that the nullhypothesis of normality could not be rejected at the 5% level of significance using a Shapiro-Wilktest. Second, we used Whites general test for homoskedasticity to detect evidence ofheteroskedasticity. We corrected the p-values and t-statistics of estimates using Whites variance-covariance matrix for those models in which heteroskedasticity appeared to be present (White 1980). 16 18. As a third measure to verify the validity of our model, we tested for multicollinearity amongindependent variables by calculating Variance Inflation Factors (VIF) (Kleinbaum et al. 1998) inregression models that contained only first-order terms before mean-centering our measures. The VIFlevels that we observed were smaller than 2, hence much smaller than the critical threshold of 10, thuseliminating the concern about multicollinearity among first-order terms in the regression analysis.Multicollinearity may, however, arise due to the introduction of the interaction term, in which casemean-centering can be applied to all first- and second-order variables as a standard and validprocedure to attenuate multicollineariry (Aiken and West 1991). Interaction terms are entered asorthogonalized effects, and this approach yields interaction variables that are uncorrelated with theircomponent variables. The VIF levels that we observed in regression models containing first- andsecond-order terms after mean-centering our first-order measures were all smaller than 2. Our modelspecification, therefore, proved robust to multicollinearity.Independent VariablesIn this study, two latent variables characterize the design of a business model (novelty andefficiency), and another two latent variables characterize the product market positioning of the firm(differentiation and cost leadership). After scale purification, we retained 13 items as measures ofnovelty, and 11 items as measures of efficiency. Similarly, we retained 3 items for differentiation, and4 items for cost leadership (see the online supplement for details on the scales). Given the difficulty ofobtaining objective measures of business model design and product market strategy, we deemed theuse of perceptual measures obtained from expert raters appropriate (Dess and Robinson 1984). Thestrength of each of these items was measured using five point Likert-type scales, which we coded intoa standardized score. After coding, we aggregated the item scores for each design theme into anoverall score for the composite scale using equal weights (see Mendelson 2000). This process yieldeddistinct quantitative measures of business model design and product market strategy.We validated the internal consistency and reliability of our measures using standardizedCronbach alpha coefficients, which were 0.71 for the business model novelty measure, 0.70 for thebusiness model efficiency measure, 0.66 for the differentiation strategy measure, and 0.76 for the cost17 19. leadership strategy measure. Hence, our measures sufficiently satisfy Nunnallys (1978) guidelines,which suggest 0.7 as a benchmark for internal consistency.Dependent VariablesA firms stock-market value reflects the markets expectations of future cash flows toshareholders, and hence can be viewed as a measure of perceived firm performance, as opposed torealized performance, which is typically embodied in historical measures of firm profitability (e.g.,ROI, ROA). Given the level of uncertainty often associated with the true prospects of firms that had arecent Initial Public Offering, perceived performance operationalized as stock market value is ameasure that is particularly germane in such a setting (Stuart, Hoang and Hybels 1999). Measures ofrealized performance, such as ROI, ROA, or Tobins q, are less appropriate for these firms, whichoften have negative earnings, few tangible assets, and low (or even negative) book values.Since most firms in our sample have relatively low levels of debt, the market value of a firmsequity is a good approximation of the market value of the whole firm. We measured the market valueof equity at a given date as the number of shares outstanding multiplied by the firms stock price,taken from the combined CRSP and Datastream databases. We then took the logarithm of the marketvalue of the equity in order to comply with the normality assumption of OLS.Control VariablesWe included further factors that might influence the market value of a firms equity as controlvariables in the analysis because their omission might confound the analysis. On the firm level, weincluded variables that controlled for the age and size (i.e., the number of employees) of the firm. Wealso controlled for additional dimensions of a firms product market strategy, such as the mode ofmarket entry, and its product and market scope (see the online supplement for details on thesevariables). On the industry level of analysis we controlled for the degree of competition4 andestimated market size5. 4 Our raters measured the degree of competition on a four-point Likert scale based on informationfound in annual reports, prospectuses, competitors SEC documents and web sites, benchmark studies,18 20. RESULTSDescriptive Statistics Table 2 provides an overview of the data we use in this study. We note that our sample firmsare relatively young and small, having an average age of seven years (median of 4.3 years) in 2000,and a median of 270 employees. We also note the large variance among sample firms as evidencedby the median, minimum, and maximum values of these variables. Furthermore, our sample firmsdraw from relatively broad and highly competitive market segments and focus on a narrow array ofproducts. There are few early entrants into the market among our sample firms. Our sample, thus,consists of emerging growth companies that address relatively established markets.[INSERT TABLE 2 ABOUT HERE] Table 2 also lists the Pearson correlations among the variables used in the regression analysis.The correlations between novelty-centered business model design and differentiation strategy (0.148),and between efficiency-centered business model design and cost leadership strategy (-0.064) are low,which supports the argument that business model design themes and product market strategies aredistinct. We also note that while some correlations among explanatory variables are significant andrelatively high (e.g., between age and entry mode: 0.488), they do not appear to pose amulticollinearity problem as the Variance Inflation Factors (VIF) are low for all these variables.Confirmatory Factor Analysis and Partial Least Squares Regression To establish the discriminant validity of our main constructs, we first conducted theconfirmatory factor analytic method proposed by Gatignon, Tushman, Smith, and Anderson (2002). Themethod consists in selecting pairs of constructs and then conducting confirmatory factor analysis (CFA)for each pair. In applying this method, we first ran a CFA for each pair of factors in an unconstrainedHoovers Database (which lists each focal firms main competitors), as well as investment analystsreports. 5 This information was obtained from Forrester research reports and from the U.S. Department ofCommerce.19 21. measurement model with the two factors. In this first model, the correlation between the factors wasestimated. For example, take novelty and differentiation as the chosen pair of factors. Novelty traitsloaded onto the novelty factor, and the differentiation traits loaded onto the differentiation factor. Table3 depicts the results from this analysis in the rows where the correlation between the factors is reportedas freely estimated (i.e., not set equal to 0 or 1). For example, the estimated correlation between noveltyand differentiation was 0.19. [INSERT TABLE 3 ABOUT HERE]We then ran a CFA on a measurement model with only one factor, where the correlationbetween the constructs of interest was constrained to be 1. If the unconstrained model where thecorrelation is freely estimated improves the fit significantly compared to the constrained model, the twoconstructs can be said to be distinct from each other, although they still can be significantly correlated(Gatignon et al. 2002; Gatignon 2003). To illustrate this, consider novelty and differentiation. Theresults from the CFA demonstrate that novelty-centered business model design and differentiation inproduct markets are distinct constructs, although they are positively correlated. This is confirmed by asignificantly (at the 0.01 level) improved confirmatory factor analytic model when the correlation isestimated, compared to a measurement model where the correlation is constrained to 1 (chi-squared =260 186 = 74, degrees of freedom = 104 103 = 1). As Table 3 shows, we obtain similar results for allother pairs involving generic product market strategies and business model design themes, whichprovides support for our corollary [Business model design (as, for example, measured by designthemes) is distinct from product market strategy (as, for example, measured by generic strategies)].In addition to CFA, the literature suggests partial least squares (PLS) as another method forassessing discriminant validity. Using PLS, one can determine whether a construct shares more variancewith its measures than it shares with other constructs in the model (Hulland 1999). This is achieved by(1) calculating the square roots of the Average Variance Extracted (AVE) values, which measure theaverage variance shared between a construct and its measures, and by (2) calculating the correlationsbetween different constructs. A matrix can then be constructed where the square root of AVE is in thediagonal, and the correlations between the constructs are in the off-diagonal. This matrix is shown in 20 22. Table 4. For adequate discriminant validity, the diagonal elements should be greater than the off-diagonal elements in the corresponding rows and columns (Fornell and Larcker 1981). This is the casehere, which is further evidence in support of the discriminant validity of our main constructs.[INSERT TABLE 4 ABOUT HERE] We note that the CFA can also be used to assess the convergent validity of the constructs(Gatignon et al. 2002; Gatignon 2003). For this, a measurement model where the correlation betweenthe two constructs is estimated and a model where the correlation is constrained to be 0 are compared. Asignificant improvement in fit (moving from zero to estimated correlation) would indicate that the twoconstructs are indeed related, which would confirm convergent validity. Using as an illustration againthe example of novelty and differentiation in Table 3, the results from the CFA demonstrate thatnovelty-centered design and product market differentiation are independent constructs. Theconfirmatory factor analytic model when the correlation is estimated, compared to a measurement modelwhere the correlation is constrained to 0, is not significantly improved (chi-squared = 189 186 = 3,degrees of freedom = 104 103 = 1). This same qualitative result holds for all pairs of generic strategiesand business model design themes.Hierarchical OLS Regressions Table 5 depicts the results from selected hierarchical OLS regression runs. Panel A reports thefull results for the models that included the interaction between novelty-centered business model designand differentiation strategy. In the Panel, the top display refers to regressions that used the logarithm ofmarket value averaged over the fourth quarter of 2000, and the bottom display refers to regressions thatused the logarithm of market value averaged over the entire year 2000. Panel B shows the main resultsfor the other interactions of interest. [INSERT TABLE 5, Panels A & B HERE] Table 5 Panel A supports the prediction that coupling a novelty-centered business model designwith a differentiation product market strategy represents good fit; these variables jointly produce asignificant positive effect on perceived performance -- for both dependent variables used (see the topand bottom display of the Panel) -- in most models that we ran. Furthermore, Table 5 Panel B supports 21 23. the hypothesized good fit between novelty-centered business model design and cost leadership strategy,and between novelty-centered business model designs and early market entry timing. Our data produce apositive coefficient on the relevant interaction terms in all of our regressions. That coefficient isstatistically significant at the 5% level in a majority of the models that exhibit an adequate F-value.Regarding the fit between efficiency-centered business model design and product marketstrategies, we note that our empirical analysis (as shown in Table 5 Panel B) did not support thepredicted good fit between efficiency-centered design and cost leadership strategy; it producedinsignificant results. However, the finding of statistically insignificant interaction terms involvingefficiency-centered design and product market differentiation or early market timing is consistent withour model, which suggested neither good nor bad fit between these variables.Lastly, we note that even when the interaction terms reported in Table 5 were statisticallysignificant, the coefficients on the corresponding main variables were sometimes insignificant. Thispoints to the importance of interactions between product market strategy and business model design.DISCUSSION AND CONCLUSIONOur empirical analysis reveals that product market strategy and business model design aredistinct constructs that affect firm performance in important ways. Using hierarchical OLS regressiontechniques, we find significant effects of their interaction on the perceived performance of firms, asmeasured by market capitalization. More specifically, we find empirical support for the theoreticalpredictions about the positive and significant interactions between novelty-centered business modeldesign and product market strategies. With respect to efficiency-centered business model design, ourempirical findings are, in general, consistent with the theoretical analysis: No clear predictions can bemade with respect to complementarities with a differentiation strategy or with respect to the timing ofentry, and indeed our empirical analysis did not reveal any such relationship.We believe that our study makes two important contributions to the management andorganization literatures. First, it extends the concept of structure beyond administrative elements toinclude boundary-crossing transactions between a focal firm and its ecosystem of partners, customers,and suppliers. By examining the business model, which captures the structure of a firms ecosystem22 24. rather than its administrative structure, and by considering business-level competitive strategy insteadof corporate strategy we extend the strategy-structure-performance paradigm to highlight theinterdependence between the competitive strategy of a firm, the architectural structure of thetransactions it enables with external stakeholders, and its performance. We show empirically thatadopting a broader view of organizations, one that transcends traditional firm boundaries, can bevaluable for understanding wealth creation and performance. By doing so, our study may inspire newresearch on the relationship between strategy and structure, and on the boundaries of firms. Second, our study explores the fit between the design of a business model and a firmsproduct market strategy. In doing so, we elaborate on the notion of good fit between theseconstructs by offering a formal model and performing a marginal effects analysis. Our study points to the need to investigate competition among various business models withinan industry. Such rivalry may have implications both for the wealth-creation potential of a givenbusiness model and for value capture by the focal firm relative to its rivals. In order to betterunderstand these phenomena, we need to know more about the strategic effects of business modeldesign and how business model design influences the positioning of firms in their competitiveenvironment. The empirical results presented in this paper show that both product market strategy andstructure, as embodied by the design of the business model, can enhance the firms competitiveadvantage, independently as well as jointly, thus supporting previously held conjectures (Christensen2001). Finally, our study raises the issue of timing of business model and product market strategydesign. Business model and product market strategy may be simultaneously determined. For example,when entrepreneurs define and refine their business models, they may concurrently identify customerneeds and map them against the products and services offered by competitors (McGrath and MacMillan2000). However, it is also conceivable that product market strategy follows business model design, orvice versa. Little research has been conducted so far on how business models evolve and in particularhow they co-evolve with the product market strategy of the firm. In this study, we hope to have laidsome of the foundations that are necessary to fruitfully explore these new avenues for research.23 25. REFERENCESAiken, L. S., S. G. West. 1991. (Paperback 1996.) Multiple Regression: Testing And InterpretingInteractions. Sage, Thousand Oaks, CA.Amburgey, T. L., T. Dacin. 1994. As the left foot follows the right? The dynamics of strategic andstructural change. Acad. Management J. 37 1427-1452.Amit, R., C. Zott., 2001. Value creation in e-business. Strategic Management J. 22 493-520.Besanko, D., D. Dranove, M. Shanley. 1996. Economics of Strategy. John Wiley & Sons. New York.Bower, J. L. 1970. Managing the Resource Allocation Process: A Study of Corporate Planning andInvestment. Harvard Business School Press. Boston.Brandenburger, A. M., H. Stuart. 1996. Value-based business strategy. J. of Econom. andManagement Strategy 5 5-25.Brandenburger, A. M., B. J. Nalebuff (1996). Co-opetition. Doubleday, New York.Chandler, A. 1962. Strategy and Structure: Chapters in the History of American Industrial Enterprise.MIT Press. Cambridge.Christensen, C. M. 2001. The past and future of competitive advantage. MIT Sloan Management Rev.Winter 105-109.Christensen, C. M., M. Verlinden, G. Westerman. 2002. Disruption, disintegration and the dissipationof differentiability. Indust. and Corporate Change 11 955-993.Dess, G. G., R. Robinson. 1984. Measuring organizational performance in the absence of objectivemeasures. Strategic Management J. 5 265-73.Dowling, M. J., J. E. McGee. 1994. Business and technology strategies and new venture performance:A study of the telecommunications equipment industry. Management Sci. 40 1663-1677.Dyer, J. H., H. Singh. 1998. The relational view: Cooperative strategy and sources ofinterorganizational competitive advantage. Acad. Management Rev. 23 660-679.Fornell, C., D. F. Larcker. 1981. Evaluating structural equation models with unobservable variablesand management error. J. of Marketing Res. 18 39-50.Gatignon, H. 2003. Statistical Analysis of Management Data. Kluwer Academic Publishers. Boston. 24 26. Gatignon, H., M. Tushman, W. Smith, P. Anderson. 2002. A structural approach to assessing innovation: Construct development of innovation locus, type and characteristics. Management Sci. 48 1103-1122.Geoffrion, A. M., R. Krishnan. 2003. E-business and management science: Mutual impacts (Part 1 of 2). Management Sci. 49 1275-1286.Gulati, R., N. Nohira, A. Zaheer. 2000. Strategic networks. Strategic Management J. 21 203215.Hulland, J. 1999. Use of partial least squares (PLS) in strategic management research: A review of four recent studies. Strategic Management J. 20 195-204.Kleinbaum, D. G., L. L. Kupper, K. E. Muller, A. Nizam. 1998. Applied Regression Analysis and Other Multivariable Methods (3rd ed.). Duxberry Press. Pacific Grove, CA.Lee, G. K., R. E. Cole. 2003. From a firm-based to a community-based model of knowledge creation: The case of Linux kernel development. Organization Sci. 14 633-649.Lieberman, M., D. Montgomery. 1988. First-mover advantages. Strategic Management J. 9 41-58.MacCormack, A., R. Verganti, M. Iansiti. 2001. Developing products on Internet time: The anatomy of a flexible development process. Management Sci. 47 133-150.McGrath, R., I. MacMillan. 2000. The Entrepreneurial Mindset. Harvard Business School Press. Boston.Mendelson, H. 2000. Organizational architecture and success in the information technology industry. Management Sci. 46 513-529.Milgrom, P., J. Roberts. 1995. Complementarities and fit: strategy, structure, and organizational change in manufacturing. J. Account. and Econom. 19 179-208.Miller, D. 1986. Configurations of strategy and structure: Towards a synthesis. Strategic Management J. 7 233-249.Miller, D. 1988. Relating Porters business strategies to environment and structure. Acad. Management J. 31 280-308.Mintzberg, H. 1990. The design school: Reconsidering the basic premises of strategic management. Strategic Management J. 11 171-195.25 27. Normann, R. 2001. Reframing Business: When The Map Changes The Landscape. Wiley. Chichester.Nunnally, J. C. 1978. Psychometric Theory. McGraw-Hill. New York.Porter, M.E. 1985. Competitive Advantage: Creating and Sustaining Superior Performance. Free Press/MacMillan. New York.Rumelt, R. 1987. Theory, strategy, and entrepreneurship. In D. J. Teece (Ed.), The competitive challenge: 137-158. Ballinger Publishing Company. Cambridge.Santos, F. M., K. M. Eisenhardt. Forthcoming. Perspectives on organizational boundaries. Organization Sci. Special Issue on New Directions for Organizational Theory.Siggelkow, N. 2001. Change in the presence of fit: The rise, the fall, and the renaissance of Liz Clairborne. Acad. Management J. 44 838-857.Siggelkow, N. 2002. Evolution toward fit. Admin. Sci. Quart. 47 125-159.Siggelkow, N., D. A. Levinthal. 2003. Temporarily divide to conquer: Centralized, decentralized, and reintegrated organizational approaches to exploration and adaptation. Organ. Sci. 14 650-669.Stuart, T. E., H. Hoang, R. C. Hybels. 1999. Interorganizational endorsements and the performance of entrepreneurial ventures. Admin. Sci. Quart. 44 315-349.White, H. 1980. A heteroskedasicity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 48 817-838.Yin, X., E. Zajac. 2004. The strategy/governance structure fit relationship: Theory and evidence in franchising arrangements. Strategic Management J. 25 365-383.Zott, C. 2003. Dynamic capabilities and the emergence of intra-industry differential firm performance: Insights from a simulation study. Strategic Management J. 24 97-125.Zott, C., R. Amit. 2003. Business model design, and the performance of entrepreneurial firms. Working Paper 2003/94/ENT/SM/ACGRD 4, INSEAD, Fontainebleau, France. 26 28. TABLE 1: Business Model, And Product Market StrategyBusiness Model Product Market StrategyDefinition Blueprint of the firms transaction with its Pattern of managerial actions that explainexternal stakeholdershow firm achieves and maintains competitive advantage through positioning in product marketsMain How to do business?How to position firm against rivals? QuestionsWho are the parties that can beWhat kind of generic strategy to Addressedbrought together to exploit aadopt (i.e., cost leadership and/orbusiness opportunity, and how candifferentiation)?they be linked to the focal firm to When to enter the market, andenable transactions? (i.e., what are how to enter it?the exchange mechanisms?)What information or goods areexchanged among the parties, andwhat resources and capabilities areneeded to enable the exchanges?How are the transactions betweenthe parties controlled, and what arethe incentives for the parties?Unit ofEcosystem of the focal firmFirm AnalysisFocusExternally oriented: focus on firms Internally/externally oriented: focus onexchanges with othersfirms activities and actions in light of competition27 29. TABLE 2: Pearson Correlations And Descriptive Statistics Ln (market valueLn (market valueLn [market size] Cost Leadership Ln [employees]Differentiation Product ScopeMarket Scopeavg. Q4 2000)Entry TimingCompetition Entry ModeAge of firm avg. 2000) EfficiencyNovelty Variable Name (Acronym)Indep. Var. Novelty 1.000Efficiency 0.193*1.000Differentiation0.148 0.0531.000Cost Leadership-0.013-0.064 -0.061 1.000Entry Timing 0.238** 0.004 0.197*0.164* 1.000Dependent Var. Ln (market value 0.176*0.79 0.1150.0080.1251.000 average Q4 2000) Ln (market value 0.241** 0.120 0.279** -0.037 0.170*0.929** 1.000 average 2000) Control Var. Competition -0.476** -0.198** -0.151*0.025-0.148-0.148 -0.189*1.000Ln [market size] -0.065-0.004-0.260**0.094-0.0520.217** 0.105 0.179*1.000Age of firm-0.135-0.101-0.295**0.072-0.0260.219** 0.044 0.0710.191* 1.000Ln [employees]0.067-0.027-0.164*0.11-0.0160.632** 0.547** 0.0120.339** 0.459** 1.000Entry Mode0.069-0.0370.443** -0.068 0.075 -0.163* 0.014 0.007-0.222** -0.488** -0.301**1.000Product Scope-0.060-0.016 0.0090.0540.092 -0.093 -0.144 0.073 -0.0760.106-0.140-0.1341.000Market Scope 0.155*0.107 -0.131-0.153*-0.060-0.012 -0.026 -0.1360.045 0.031-0.035-0.1450.1001.000Descriptive Stat. Mean0.3820.7423.5982.6572.147517883 0.624 224107.01145 3.971 3.7651.871Median0.3720.7503.6672.500177 183 0.639 5400 4.32704 41Std. Deviation0.1380.1240.7961.0281.590 1491 2262 0.175 691117.93749 1.275 1.0111.047Min 0.0770.3861.667 1 1 250120 0.4 171 11Max 0.8141 55 5 12304166510.972744000 46 31000 5 55N 170 170170170170 161169170 170170 170 170170170 Note on descriptive statistics: (1) The independent variables are indices that have been coded such that low values represent a low emphasis, and high values represent a high emphasis on the respective business model design theme, or product market strategy. High values of Entry Timing indicate early market entry timing. (2) Market value and market size are given in $ millions, without taking the logarithm. (3) Firm size is given as number of employees, without taking the logarithm. (3) High values of Entry Mode indicate high reliance on strategic partnerships and/or joint ventures in developing, producing, or marketing products. (4) High values of Product Scope indicate a highly focused product offering. (5) High values of Market Scope indicate a very focused market approach. ** p