Dynamic Commercialization Strategies for Disruptive Technologies (PDF)

This article was downloaded by: [131.175.147.147] On: 24 January 2025, At: 01:32 Publisher: Institute for Operations Research and the Management Sciences (INFORMS) INFORMS is located in Maryland, USA Management Science Publication detail...

This article was downloaded by: [131.175.147.147] On: 24 January 2025, At: 01:32 Publisher: Institute for Operations Research and the Management Sciences (INFORMS) INFORMS is located in Maryland, USA Management Science Publication details, including instructions for authors and subscription information: http://pubsonline.informs.org Dynamic Commercialization Strategies for Disruptive Technologies: Evidence from the Speech Recognition Industry Matt Marx, Joshua S. Gans, David H. Hsu To cite this article: Matt Marx, Joshua S. Gans, David H. Hsu (2014) Dynamic Commercialization Strategies for Disruptive Technologies: Evidence from the Speech Recognition Industry. Management Science 60(12):3103-3123. https://doi.org/10.1287/ mnsc.2014.2035 Full terms and conditions of use: https://pubsonline.informs.org/Publications/Librarians-Portal/PubsOnLine- Terms-and-Conditions This article may be used only for the purposes of research, teaching, and/or private study. Commercial use or systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisher approval, unless otherwise noted. For more information, contact [email protected]. The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitness for a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, or inclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, or support of claims made of that product, publication, or service. Copyright © 2014, INFORMS Please scroll down for article—it is on subsequent pages With 12,500 members from nearly 90 countries, INFORMS is the largest international association of operations research (O.R.) and analytics professionals and students. INFORMS provides unique networking and learning opportunities for individual professionals, and organizations of all types and sizes, to better understand and use O.R. and analytics tools and methods to transform strategic visions and achieve better outcomes. For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org  MANAGEMENT SCIENCE Vol. 60, No. 12, December 2014, pp. 3103–3123 ISSN 0025-1909 (print) — ISSN 1526-5501 (online) http://dx.doi.org/10.1287/mnsc.2014.2035 © 2014 INFORMS Dynamic Commercialization Strategies for Disruptive Technologies: Evidence from the Downloaded from informs.org by [131.175.147.147] on 24 January 2025, at 01:32. For personal use only, all rights reserved. Speech Recognition Industry Matt Marx MIT Sloan School of Management, Massachusetts Institute of Technology, Cambridge, Massachusetts 02142, [email protected] Joshua S. Gans Rotman School of Management, University of Toronto, Toronto, Ontario M5S 3E6, Canada, [email protected] David H. Hsu The Wharton School, University of Pennsylvania, Philadelphia, Pennsylvania 19104, [email protected] W hen start-up innovation involves a potentially disruptive technology—initially lagging in the predomi- nant performance metric, but with a potentially favorable trajectory of improvement—incumbents may be wary of engaging in cooperative commercialization with the start-up. While the prevailing theory of dis- ruptive innovation suggests that this will lead to (exclusively) competitive commercialization and the eventual replacement of incumbents, we consider a dynamic strategy involving product market entry before switching to a cooperative commercialization strategy. Empirical evidence from the automated speech recognition industry from 1952 to 2010 confirms our main hypothesis. Keywords: technology commercialization strategy, disruptive innovation History: Received May 23, 2012; accepted July 3, 2014, by Bruno Cassiman, business strategy. Published online in Articles in Advance October 24, 2014. 1. Introduction of physics” (Brodsky 2008, p. 199) and accused Qual- Entrepreneurs seeking to commercialize their techni- comm of faking its first demonstration. Qualcomm cal innovations often rely on cooperative strategies, temporarily abandoned licensing and began manufac- such as technology licensing, with other organiza- turing both base stations and handsets to prove the tions. They do so both to access the skills or assets value of CDMA technology. It retained these comple- they may not possess and to minimize competitive mentary businesses for several years before selling the effects. Given that the decision to cooperate with former to Ericsson and the latter to Kyocera. In per- incumbents is not unilateral, the incumbent must see sonal communication, Qualcomm cofounder Andrew some advantage in accessing the technology from the Viterbi recounted the following: innovator. But if the incumbent is unsure about the value of the technology, cooperation may be initially [F]or this large and complex opportunity it was essen- tial to produce the infrastructure as well as the infeasible. Thus, the entrant may find it necessary handsets 0 0 0 it was necessary to convince the carriers to compete in the product market, at least until the that CDMA was indeed a workable technology which incumbent becomes convinced regarding the value of had a major advantage over alternates: GSM, U.S. the technology. and Japanese TDMA standards. All of this took a Consider the case of Qualcomm’s code-division lot of effort, several successful demonstrations, some multiple access (CDMA) technology for handling cel- luck and about three or four years; there were many lular communications. CDMA took the controversial skeptics. (Viterbi 2012) approach of handling multiple calls on the same fre- quency simultaneously and managing the interfer- Qualcomm’s strategy of temporarily entering the ence as opposed to sequentially as in the prevail- product market and subsequently switching to its pre- ing protocol, time-division multiple access (TDMA). ferred licensing model serves as an example of how Although CDMA promised to be more efficient than firms can demonstrate the value of their technology TDMA, there were many skeptics, including a Stan- to would-be partners. ford University professor who declared that the One category of innovations that may be particu- frequency-sharing approach would “violate the laws larly difficult to commercialize in a cooperative setup 3103  Marx, Gans, and Hsu: Dynamic Commercialization Strategies for Disruptive Technologies 3104 Management Science 60(12), pp. 3103–3123, © 2014 INFORMS are “disruptive” technologies. Disruptive technolo- notion that disruptive technologies necessarily result gies exhibit an initially worse performance profile on in the demise of incumbents, such as in the disk- the dimension valued by mainstream consumers (e.g., drive industry (Christensen 1997). Although the ini- OECD 1967, Foster 1986, Christensen 1997), so the tially unattractive nature of disruptive technologies gains to trade with incumbents required for cooper- does entail first stage entrant/incumbent competition, ative commercialization may not exist. If deployed, cooperation may ultimately ensue. however, they may exhibit a favorable trajectory of Downloaded from informs.org by [131.175.147.147] on 24 January 2025, at 01:32. For personal use only, all rights reserved. improvement. In such a circumstance, the commer- 2. Theory and Main Hypothesis cialization partner may have little financial incentive The literature on commercialization strategy has early on to develop the innovation in-house or access focused on the entrant choice between competing it via contractual means, as combining it with their or cooperating with incumbents (Teece 1986, Gans existing activities is costly. However, should a poten- and Stern 2003). The empirical investigation of those tially disruptive technology prove to be valuable, choices has correlated them with characteristics of the these incentives may change. Thus, in contrast to the market environment, including competition (Arora main predictions of existing analyses that find incum- et al. 2001), access to complementary assets (Gans bent firm market leadership routinely replaced in the et al. 2002), frictions (Hsu 2006, Chatterji and Fabrizio face of disruptive innovation by entrepreneurs, coop- 2013) and the strength of intellectual property protec- erative commercialization—which preserves incum- tion. Here, we instead consider how different technol- bent market leadership—may still be a long-term ogy types within an industry correlate with commer- outcome. cialization choices. In addition, we examine changes We explore a two-stage commercialization strategy in commercialization strategy throughout the life of in which a start-up entrant temporarily enters the an entrepreneurial firm, thereby moving away from product market to establish the value of its technol- the static, one-time choice that has been the hallmark ogy. Ultimately, the entrant may switch to a strat- of the TCS literature to date. egy of cooperating with incumbents once uncertainty over the disruptive technology is resolved and/or the 2.1. How Does Technology Innovation Type incumbent’s costs of integrating the new technology Impact Commercialization Choice? declines. This dynamic technology commercialization There have been many classifications of technology strategy (TCS) extends extant frameworks linking the that have been used to inform strategic management. environmental, organizational, and competitive fac- Here we focus on those that have been argued to tors to an entrant’s initial choice of TCS (Teece 1986, impact the nature of the commercialization choice Gans and Stern 2003). Such work characterizes TCS as for entrants between competing and cooperating. To a one-time, static decision to cooperate with incum- date, the literature on commercialization strategy has bents via licensing or to compete against them in the emphasized entrant costs in competitive entry. In the product market. predominant static TCS framework (Teece 1986, Gans Perhaps one reason commercialization strategy has and Stern 2003), the lower the cost of product market not been explored dynamically is the difficulty of entry, including the costs of assembling the requi- obtaining longitudinal data regarding TCS adoption site downstream complementary assets for commer- and evolution. We introduce a hand-collected data set cialization, the more attractive is a competitive com- tracking all entrants into the automatic speech recog- mercialization strategy. This is especially true if the nition (ASR) industry from its inception in 1952 appropriability regime is weak so that the entrant’s through the end of 2010. ASR is an attractive indus- exposure to disclosure risks when bargaining over try for TCS analysis because its commercialization deal terms with industry incumbents is high. environment leaves open a variety of possible strate- By contrast, the literature on the direction of inno- gies. The data allow us to follow technology com- vation in an industry has started with the organi- mercialization strategies on an annual basis, including zational effect of such innovations on incumbents. when firms change from their initial TCS. Further- Tushman and Anderson (1986) classify innovations more, our long time horizon of observing industry into those that are competence-destroying (requiring entrants allows us to study the relationship between new organizational skills to successfully commercial- innovation characteristics (e.g., disruptive technology ize) and competence-enhancing (those that build on status) and their commercialization strategies. existing organizational know-how). Across a variety Our analysis reveals that ASR entrants who intro- of industrial settings, researchers have found that duce disruptive technologies are more likely to adopt competence-destroying innovations are more likely to a two-stage commercialization strategy in which they be initiated by new entrants, whereas industry incum- initially compete with incumbents but later cooper- bents tend to originate competence-enhancing discon- ate with them. This result calls into question the tinuities (Tushman and Anderson 1986, Christensen  Marx, Gans, and Hsu: Dynamic Commercialization Strategies for Disruptive Technologies Management Science 60(12), pp. 3103–3123, © 2014 INFORMS 3105 and Bower 1996). This pattern reflects the behavior of both conceptually and in reality possible that having established firms, which are typically eager to invest chosen one commercialization path, an entrant may and support innovations that sustain and extend rates subsequently switch to another. Building a dynamic of improvement along the dimensions demanded by theory of commercialization choice involves consid- their mainstream consumers. ering what changes might occur after an entrant’s Although entrants constrained to choose coopera- initial commercialization choice and, importantly, tive commercialization paths may themselves pursue the changes that will occur because of the choice Downloaded from informs.org by [131.175.147.147] on 24 January 2025, at 01:32. For personal use only, all rights reserved. a competence-enhancing innovation, they have strong (Gans 2012). incentives to originate competence-destroying inno- The Christensen line of research describes a class vations because they do not fear product cannibaliza- of technologies called “disruptive technologies.” As tion and typically do not have vested positions in a already noted, such technologies poorly serve the preexisting complementary asset infrastructure. existing customers of incumbents in key dimen- From this perspective, the incumbent’s costs of inte- sions. But, importantly, what gives them their disrup- grating new technology will impact the surplus that tive power is that this underperformance is eroded can be generated from cooperative commercialization over time and in the long run, such technologies with an entrant’s technology. If those integration costs may outperform existing technologies along dimen- were high with regard to incumbent market reposition- sions valued by mainstream customers. For example, ing and complementary asset reorientation (as would Christensen and Bower (1996) show that the lower be the case under competence-destroying innovation), capacity, slow access speed, and high cost of 5.25-inch cooperative arrangements would be less likely to take disk drives compared to existing 8-inch disk drives led place. By contrast, if those costs were low, there would to their rejection by minicomputer original equipment be no incumbent-side barrier to integration, and coop- manufacturers (OEMs). By contrast, “sustaining” tech- erative commercialization would be favored.1 nologies would improve capacity and access speed. In an influential line of research, Bower and Chris- Thus, the 5.25-inch disk drive was not attractive to tensen (1995), Christensen and Rosenbloom (1995), incumbents. However, the 5.25-inch disk drive had a Christensen and Bower (1996), and Christensen (1997) path of improvement along those traditional metrics as describe a set of technologies which are, initially, less its use case became better understood over time. Con- compatible with incumbent products and processes. sequently, small drives came to dominate the market. This is because they perform poorly on dimensions What are the drivers and implications of an that are currently valued by the majority of con- enlarged choice-set in which an entrant may alter its sumers in the market. These represent a good exam- initial commercialization strategy? The first issue is ple of technologies that would be costly for incum- generic uncertainty regarding the innovation’s future bents to integrate into their existing product lines. value. For example, it may be profitable for an incum- Thus, we will use this metric in our empirical work bent to incur the costs of integrating a technology as a proxy for technologies that have high initial costs and improving it in-house if the incumbent were of integration and could be a technological driver of assured of the innovation’s future value. But if there the choice between competition and cooperation. is uncertainty in that regard, the incumbent may be reluctant to cooperate initially.2 The second issue is 2.2. What Drives Changes in Commercialization what happens to the incumbent’s cost of integra- Choice by Entrants? tion over time. If the incumbent chooses to cooperate The static TCS literature assumes that commercializa- initially, those costs are incurred and then sunk, so tion is a one-time choice for the entrant. However, it is they are irrelevant from the perspective of subsequent decisions. However, in situations where the technol- 1 There are likely to be heterogeneous incumbent firm responses in ogy is disruptive, one expects that following market the face of radical technologies, however. Mitchell (1989) finds that tests, a technology may improve along all dimensions, the degree of industry rivalry and prior organizational investments including those that the incumbent’s customers value. in specialized assets shape the likelihood and timing of incumbent firm entry in emerging subfields of medical imaging technologies. If such an improvement was anticipated, an incum- King and Tucci (2002) document that in the hard disk drive indus- bent may prefer to wait before engaging in cooper- try, market entry in the face of radical technical change depends ation. For disruptive innovations, we may therefore on firms’ production and sales experience (and so is not simply a observe competition initially followed by cooperation function of demand-side forces). More generally, Iansiti (2000) pro- at a later stage. vides evidence that both evolutionary and revolutionary responses by firms in navigating technological transitions can achieve compa- 2 rable performance, and so there is not necessarily a “best response” Arora et al. (2001, p. 430) allude to this possibility: “0 0 0 [s]ometimes strategy by incumbents to technical transitions. Evolutionary and self-production is a necessary condition for successful licensing. For revolutionary response strategies each have different precursors for instance, self-production could help assess the true value of the use along the dimensions of experimentation and project versus technology or could help identify potential bottlenecks in technol- research experience. ogy transfer.”  Marx, Gans, and Hsu: Dynamic Commercialization Strategies for Disruptive Technologies 3106 Management Science 60(12), pp. 3103–3123, © 2014 INFORMS The interplay between uncertainty and expectations or disruptive in the sense that they underperform on regarding future integration costs for incumbents is traditional metrics, it will be the entrant who has an complex. In Appendix A, we provide a dynamic advantage in conducting that test.3 model of commercialization that formally investigates One of the main claims of Christensen (1997) is that these effects, taking into account the fact that com- disruptive innovation is often associated with replac- mercialization strategy is not a choice of the entrant ing incumbent firm market leadership despite (ini- per se but is the outcome of a negotiation between tial) technical underperformance in the predominant Downloaded from informs.org by [131.175.147.147] on 24 January 2025, at 01:32. For personal use only, all rights reserved. the entrant and incumbent. In this case, because that performance dimension. However, an entrant strat- negotiation may take place both in the present and egy of initially competing followed by later cooperat- potentially in the future, examining the equilibrium ing would suggest that, in some cases of disruptive outcomes is not trivial. technology, incumbent market leadership might still The model confirms the intuition expressed here. be preserved. Bower and Christensen (1995), in dis- It demonstrates that the more uncertain is the future cussing managing disruptive technological change, do value of the entrant’s innovation in the market place, consider an incumbent acquisition strategy (although the more likely the entrant will undertake competitive not a technology in-licensing one). While the authors commercialization initially. However, the model also acknowledge and give examples of how such acqui- demonstrates that a switch in commercialization strat- sitions have helped preserve incumbent market lead- egy from competition to cooperation does not depend ership, they point to both the innovator’s possible on that uncertainty even if it depends on its resolu- reluctance in pursuing a cooperative strategy as well tion. Instead, switching strategy depends on the real- as the difficulty of successfully executing acquisitions. ized changes in the incumbent’s cost of integration. If The end result is the predominant conclusion in the these are large and the entrant’s innovation turns out existing literature that disruptive innovation over- to be valuable in the marketplace, a switch will occur. turns incumbent market leadership. We now explore In the empirics, innovations that turn out not to be how an innovator’s commercialization strategy of ini- valuable may be short-lived, so we are more likely tial cooperation followed by later cooperation might to observe changes in commercialization strategy for temper this view. long-lived innovations. We predict that an observed switch from competition to cooperation will be asso- ciated with technologies that initially underperform 3. Data We test our hypothesis using a new, hand-collected but have a strong path of improvement along tra- data set of the automatic speech recognition (ASR) ditional metrics; that is, disruptive technologies. For industry from its inception in 1952 through the end such technologies we may see entrepreneurs switch of 2010. ASR technology converts spoken language their commercialization strategy. That is, competition into text by modeling the sound waves generated by may precede cooperative commercialization strate- the human vocal tract. It is a science-based industry gies (e.g., licensing or acquisition), as was the case whose technology was incubated for many years in with Qualcomm. By contrast, innovations that per- corporate and university research labs before coming form well initially and/or do not have a strong path to market. The earliest recorded ASR research effort of improvement along those metrics (i.e., sustaining was in 1952, when scientists at AT&T Bell Labora- technologies) will not be associated with switches in tories built a machine that could recognize the dig- commercialization choice. Thus, our hypothesis is as its zero through nine when spoken in isolation. Sim- follows: ilar projects sprang up shortly thereafter at nearby Hypothesis. Disruptive technologies will be associated RCA Laboratories and Lincoln Laboratories in the with a higher level of competition initially followed by a United States as well as internationally at London’s switch to cooperation 4either licensing, acquisition or both5. University College, Kyoto University, and NEC. The early 1960s brought the entry of Texas Instruments It is useful to stress here that when there is uncer- and the founding of IBM’s T.J. Watson Research Cen- tainty over an innovation’s value, there are two paths ter, which invested in ASR. The industry’s first com- to a market test to resolve that uncertainty. First, the pany dedicated exclusively to ASR was Threshold incumbent could license or integrate the technology Technology, spun out of RCA Labs. Since then, ASR into its own products and test it in the market. Sec- ond, the entrants could enter the market themselves 3 and test the innovation’s value. When the incumbent It is precisely because the incumbent has an option to negotiate for and entrant negotiate initially over cooperation versus an entrant switch to cooperative commercialization that the incum- bent has an incentive for the entrant to bear those risks and carry commercialization they are, in effect, choosing who out the initial market test. In the absence of that option, the poten- would be more efficient in conducting that market tial for disruption may see incumbents acquiring technologies just test. For technologies that are competence-destroying to put them on the shelf.  Marx, Gans, and Hsu: Dynamic Commercialization Strategies for Disruptive Technologies Management Science 60(12), pp. 3103–3123, © 2014 INFORMS 3107 has been used for myriad applications including radi- Inc.’s 1988 MEDTRANS radiology dictation system ology dictation, plush toys that respond to voice, tethered dedicated hardware to a Sun Microsys- remote access to personal computers, 411 directory tems workstation, which provided the user interface. assistance automation, personal telephone assistants, Although the move to software promised both cost and podcast transcription. reduction and convenience as dedicated hardware ASR is an attractive industry for this study for was eliminated, these came at the expense of per- at least two reasons. First, it represents a com- formance trade-offs in vocabulary size and accuracy. Downloaded from informs.org by [131.175.147.147] on 24 January 2025, at 01:32. For personal use only, all rights reserved. mercialization environment where cooperating with Consequently, many firms were reluctant to abandon incumbents does not strongly dominate competing hardware acceleration. in the product market or vice versa. Technology is (2) Word-spotting. Speech recognizers generally strongly excludable, with ASR firms having filed operate by attempting to decode all words spoken more than 3,000 patents. Although complementary by the user, as is necessary in a dictation program. assets are often needed to bring innovations to mar- For some applications, however, it is less important to ket, including custom application development, many understand everything the user said and more impor- ASR entrants integrated into those assets Qualcomm- tant to capture a few key commands. As an example, style to compete in the product market. This stands some automated telephone call routing systems are in contrast to other industries, such as automotive or designed to pick out the words “operator” and “col- biotechnology, where complementary assets such as lect call” while ignoring whatever else the user hap- clinical trials are so expensive and difficult for a start- pened to say. Word-spotting promised to be advanta- up to undertake that new entrants can hardly hope geous for a niche set of applications, but the so-called to “go it alone” (Baum et al. 2000). And there is lit- “garbage models” required to filter out unwanted tle risk that the algorithms can be expropriated when speech could be unreliable. Moreover, only a small included as part of an end-user product. number of keywords could generally be handled by Second, ASR is an industry where considerable such systems. uncertainty surrounds the value of new innovations. (3) Grammar-free recognition. Historically, speech At first glance this might seem surprising, because recognition systems were configured to recognize the performance of an algorithm would seem to be from a set of words or phrases called a “recognition verifiable. Indeed, many ASR companies have pub- grammar.” The internal phonetic lattices generated lished performance claims for many years. As early by a statistical “hidden Markov model” search are as September 1981, Interstate Electronics Corporation pruned by comparing them against the set of allowed claimed 85% accuracy for its speech recognition tech- word sequences within the grammar. In grammar-free nology. One month later, Weitek claimed 90% accu- recognition, the results are not strictly filtered by a racy; the following month, IBM claimed 91% accu- set of allowable phrases; the user may, in a sense, say racy. By February of the following year, Votan claimed anything. Of course, the system may not recognize 99% accuracy, matched that summer by Interstate unusual or nonsensical utterances, but if the acoustic Electronics and soon after by Verbex, NEC America, evidence is strong enough, it may override the prior Dragon Systems, Kurzweil, Integrated Wave, General word-sequence probabilities in the bigram/trigram Instrument, and others. Such claims made it difficult models. for potential licensees to discriminate among technol- In the analysis section, we present evidence sug- ogy suppliers, as reflected by the National Bureau gesting that these technologies indeed were disrup- of Standards’ observation that almost all vendors of tive in that they underperformed existing technolo- speech recognition technology claimed 99% accuracy gies initially but gradually improved over time. (Creitz 1982). The National Research Council echoed The data for our study comprise nearly 60 years these concerns, lamenting the lack of uniform pro- since the inception of the ASR industry. The orig- cedures for evaluating speech recognition systems inal archives consist of approximately 15,000 pages (Creitz 1984). of several monthly trade journals, variously span- Additionally, some ASR entrants employed disrup- ning the years from 1981 through 2010, as well as tive technologies. Such innovations may not perform a historical account of the industry from its incep- as well on traditional metrics and thus may be less tion in 1952.4 Although it is possible that some firms attractive to potential cooperation partners, who may have been omitted from the newsletters or histori- regard their value as suspect. Three such innovations cal documents, even obscure companies were cov- are listed below. ered in detail. These trade journals offer the ability to (1) Software-only. ASR involves intensive audio sig- characterize entrepreneurs’ backgrounds and choices nal processing, so early systems generally required algorithms to run on specialized DSP chips or stand- 4 Few firms were active in the 1970s and earlier, and results are alone processing units. For example, Speech Systems robust to omitting pre-1981 data.  Marx, Gans, and Hsu: Dynamic Commercialization Strategies for Disruptive Technologies 3108 Management Science 60(12), pp. 3103–3123, © 2014 INFORMS Figure 1 ASR Firm Entry and Exit Since the Inception of the Industry in 1952 250 200 Downloaded from informs.org by [131.175.147.147] on 24 January 2025, at 01:32. For personal use only, all rights reserved. 150 100 50 0 52 54 56 58 60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 –50 Ceased operations Exited ASR Acquired New de novo New de alio Total active Note. The connected line is the overall industry density (i.e., number of active firms). “as it happened” from third-party accounts rather trade journal. Firms that competed directly for end than relying on retrospective reconstruction of events. customers by offering products or services were Moreover, they offer detail regarding the strategy for- classified as having adopted a “compete” strategy. mulation process that is unavailable from business For example, Dragon Systems sold software that registers or other traditional data sources. enabled consumers to dictate onto their personal The first author, along with research assistants, read computers. Tellme Networks offered an advertising- and coded the monthly trade journals by hand. We supported 1-800 number for retrieving sports scores, noted in each article the ASR firms mentioned and stock quotes, etc. on its voice platform. Firms were coded them as “active” in that month. A firm was categorized as adopting a compete strategy if, using counted as having entered the industry as of its first information from the trade journals, they sold end- mention in the trade journals. A firm was coded as user products, built custom solutions, or provided having left the industry when a trade journal arti- an advertising-supported service. By contrast, ASR cle noted that it either ceased operations in the ASR firms that licensed technology or development tools industry or was acquired by another company. For were classified as having a “cooperate” strategy. As firms that were never noted to have left the indus- examples, BBN Technologies (originally Bolt, Beranek, try, we checked current corporate websites to ensure and Newman) licensed its ASR technology, and Voice- that they were still operating in the ASR industry Objects supplied toolkits that companies used to build as of December 2010. For the few that were not, we end-user applications. If both compete and cooper- attempted to determine their date of exit from pub- ate strategies were mentioned at entry, the firm was lic sources; when such information was not otherwise coded as having started with them simultaneously as available, we backdated their exit date to their final a “mixed mode” (Teece 1986). mention in the trade journals. Patterns of entry and A shift of commercialization strategy from com- exit are depicted in Figure 1. pete to cooperate or vice versa was coded as such only if an initial TCS was noted in the newslet- 3.1. Technology Commercialization Strategy ters, followed by a subsequent mention of a differ- (TCS) Variables ent TCS. The variable switched TCS was set to 1 for Perhaps most unique to our study, we coded com- a given firm-year observation if the firm had previ- mercialization strategies undertaken by the firm. The ously changed from its initial TCS, and 0 otherwise. adoption of a particular TCS was coded as hav- Subcategorizations of this variable were also noted ing taken place the month it was reported in the for firms switching from cooperate → compete and  Marx, Gans, and Hsu: Dynamic Commercialization Strategies for Disruptive Technologies Management Science 60(12), pp. 3103–3123, © 2014 INFORMS 3109 vice versa. As an example of a switch from a coop- In models where acquisitions are treated as coop- erate to a compete strategy, Nuance Communications eration, we count only “attractive” acquisitions, as initially embarked on a cooperative commercializa- opposed to the purchase of a company (or its assets) at tion strategy involving technology licensing and the a “fire sale” price resulting in little or no financial gain sale of development toolkits. But a December 2002 for shareholders. Following Arora and Nandkumar trade journal article described Nuance’s switch to a (2011), we classify an acquisition as attractive if it competitive TCS: “Nuance has in the past empha- meets the following criteria. First, for venture capi- Downloaded from informs.org by [131.175.147.147] on 24 January 2025, at 01:32. For personal use only, all rights reserved. sized sales through partners 0 0 0 contribut[ing] 82% of tal (VC)-backed ventures, the acquisition price must Q3 revenues. Nuance will develop and sell prepack- exceed the invested capital. Second, for non-VC- aged applications directly, and has formed an applica- backed ventures (or VC-backed ventures where the tions group to develop the applications. Nuance will acquisition price was not available), either evidence sell directly to end-user customers” (Meisel 2002, p. 23; from press releases and news stories that the founder emphasis ours). or chief executive officer (CEO) of the focal firm joined As an example of switching from compete → the acquirer or an upward sales and/or headcount cooperate, Vlingo Corporation began by integrating growth trend must exist. We implemented these cri- its speech recognition technology into a download- teria by retrieving acquisition values from Securities able application for smartphones, only later entering Data Company, Zephyr, and other public sources; by into OEM licensing agreements with device manu- reviewing press materials associated with the acqui- facturers. Vlingo was among the early adopters of sition; and by assessing headcount and sales trends grammar-free speech recognition for cellular phones, using data from Dun & Bradstreet (Walls & Associates which was a bold move that met with skepticism 2010). Using this method to determine whether sales regarding its feasibility. Vlingo began demonstrating and headcount grew or shrank in the year prior to the its grammar-free speech recognition for phones in