Platform Preannouncement Strategies: The Strategic Role of ...

MANAGEMENT SCIENCEVol. 67, No. 3, March 2021, pp. 1527–1545

http://pubsonline.informs.org/journal/mnsc ISSN 0025-1909 (print), ISSN 1526-5501 (online)

Platform Preannouncement Strategies: The Strategic Role ofInformation in Two-Sided Markets CompetitionRamnath K. Chellappa,a Rajiv Mukherjeeb

aGoizueta Business School, Emory University, Atlanta, Georgia 30322; bCox School of Business, Southern Methodist University, Dallas,Texas 75275Contact: [email protected], https://orcid.org/0000-0001-9522-8849 (RKC); [email protected],

https://orcid.org/0000-0002-8602-5983 (RM)

Received: August 17, 2019Accepted: December 14, 2019Published Online in Articles in Advance:May 11, 2020

https://doi.org/10.1287/mnsc.2020.3606

Copyright: © 2020 INFORMS

Abstract. The release of a new platform version is often preceded by prelaunch activitiesincluding a preannouncement of new features, improvements, and other innovations. Theinformation contained within these preannouncements not only shape expectations ofdistinct but connected sides, for example, users and developers in a video game platform,but also informs a rival platform in a competitive market. Through a game-theoretic analysisof three different preannouncement strategies (formal, informal, and no-preannouncement)in a duopoly, our research furthers the understanding of externality-related informationon expectation formation and the associated competitive dynamics. The literature on the roleof information in platform competition is limited, and as the first to model the use of thisinformation (through preannouncements) as a strategic lever, we characterize the firms’equilibrium preannouncement strategies under different market conditions including whenagents have an option to switch and/or the ability to multihome. Our findings show a clearrelationship between the equilibrium preannouncement strategies and agents’ strength oftaste preferences. In markets with weak preferences, firms pursue a no-preannouncementstrategy in equilibrium to a formal preannouncement strategy where users and developershave strong taste preferences. Overall, our welfare analyses reveal that in a competitivemarket, firms will prefer not to preannounce their platform features even if preannouncementmay increase expected network effects. Consumer and social welfare are both higher inmultihoming markets than in single-homing ones.

History: Accepted by Chris Forman, information systems.Supplemental Material: The online appendix is available at https://doi.org/10.1287/mnsc.2020.3606.

Keywords: preannouncement • platforms • two-sided markets • network effects • fulfilled expectations • duopoly • game theory • pricing

1. IntroductionIn several industries (e.g., automobile, consumer elec-tronics), it is common for firms to preannounce theforthcoming release or launch of their products. Pre-announcement is an important part of the prelaunchactivities of a firm and can potentially influence bothconsumers and competitors. Although product pre-announcements target one group (consumers or usersof the product), preannouncements of platforms (e.g.,video game consoles, mobile operating systems) areoften intended to inform and educate two (or more)distinct groups (e.g., game or application users anddevelopers). Research in IS and Economics have re-ferred to such markets as two-sided markets (Parkerand Van Alstyne 2005).

Preannouncements and prelaunch activities con-tain information that may be critical to both sides ofthe platforms (and perhaps the competitor) as theyhelp in forming expectations about the forthcomingplatform. In addition, many technology-based plat-forms also have elements to them that create network

externalities wherein users may gain more value notonly from the increased number of users (called same-side network effects) but also from a greater num-ber of developers (called cross-side network effects)(Eisenmann et al. 2006). Therefore, one can imaginethat preannouncing any such network effects en-gendering (NEE) aspects in the new version of theplatform is an important decision for thefirm.More sobecause expectation formation by all parties (users,developers, and competitors) also needs to take intoaccount the potential future user/developer base ofthe platform before purchasing the new version.Extant research describes preannouncement as a

“formal, deliberate communication” (Eliashberg andRobertson 1988, p. 282) that helps a firm to fully in-form its existing and potential consumers regardingthe details of its future release. Preannouncementactivities typically involve several channels includ-ing traditional media (television, trade shows, in-terviews, billboards, special events, etc.) and socialmedia (e.g., Twitch, Facebook, Twitter, and specialized

1527

http://pubsonline.informs.org/journal/mnsc

mailto:[email protected]

https://orcid.org/0000-0001-9522-8849

https://orcid.org/0000-0001-9522-8849

mailto:[email protected]

https://orcid.org/0000-0002-8602-5983

https://orcid.org/0000-0002-8602-5983



discussion forums). In fact, a recent survey points outthat Facebook has now become the second-most usedsource for preannouncement behind TV commercials(Schneider 2015). Perhaps, one would think that iffirms are adding newNEE elements to their platform,then they should make every effort to transmit thisinformation to the market. However, even if such afull-fledged preannouncement was meaningful froma user/developer point of view, firms still have toincur significant costs in engaging in such prelaunchactivities. These costs are not trivial, for example,setting up a booth at CES1 and ancillary activitiesassociated with informing the market about a newversion of a video gaming console costs millions ofdollars2 even formarket leaders such asMicrosoft andSony. Indeed, we observe that not everyone in theindustry preannounces or invests in the same fashion.

Sometimes firms do pursue a formal/full infor-mation preannouncement of the new versions of theirplatforms thereby informing the market of all theirNEE elements. For example, consider Nintendo Wiithat camewith an advancedmotion-controller, whichenabled gamers to enjoy group activities such assports, dance, tournaments, workout, etc. Given thenovelty of this feature and its importance to enjoyingnetwork effects (primarily with regards to group-relatedgames/activities), prior to releasing Wii, Nintendomade significant investments (Sterlicchi 2013) to en-sure that both sides (gamers and developers) of theplatform understood the value of the new platform.As a part of its prelaunch activities, Nintendo recruitedrepresentatives to provide interactive demos (at home-shows, fitness expos, and major malls) and sponsored“Wii parties” to inform potential consumers regard-ing the experience and ease of playing games in a groupusing Wii. By their nature, many NEE aspects exhibitcharacteristics of experience goods (Yang andMai 2010),and in the absence of actual consumption, significantinvestments may be necessary for consumers to fullyunderstand and form expectations on the value of theplatform (Draganska and Klapper 2007, p. 1).

Even on the developer side, well before the actualrelease, Nintendo invested in providing extensiveinformation on the APIs (Application ProgrammingInterface) and toolkits that informed themarket aboutthe new NEE aspects through their Nintendo’s De-veloper Portal. Overall, the trade-press attributedsuch prelaunch initiatives toWii’s success in retainingits user base, capturing newmarkets of “alpha moms,pre-teen girls, and senior citizens,” as well as attractingdevelopers resulting inanumberofnewgames launchedon the day of platform release (Sterlicchi 2007, p. 1).

Interestingly, firms do not always invest in suchcomplete set of prelaunch activities for all their plat-forms. For instance, many experts attribute Nintendo’sstruggle with Virtual Boy to its inability to inform the

two sides of the market regarding its key NEE aspectsdespite being ahead in the VR headset trend in thevideo game industry (Mulkerin 2016). In practice, wedo witness situations where firms may not engagein any prelaunch activities at all (called a no pre-announcement strategy in our paper). Further, thereare cases where firms do not want to suffer the fullinvestment in all the prelaunch activities but per-haps incur a part of this cost. The latter case is calledinformal/partial preannouncement in our paper wheresuch a half-measure leads to the market becoming onlypartly aware of all the NEE aspects of the new platform.Microsoft’s Xbox One preannouncement can be

characterized as such a case where not all networkeffects engendering aspects were fully communicatedto the marketplace before launch. For example, al-thoughmany features of the Xbox One platform, suchas hardware specifications and some API supportwere preannounced, some key features like Xbox Livesupport and Kinect enhancements that allowed formultiplayer online gaming, were not promoted beforethe actual launch. Indeed, the industry publicationswarned that gamers are left with “far too manyunanswered questions and vague details (regardingXbox One)” (Gallaway 2013), suggesting that gamersmay not be able to have a full understanding of thenetwork benefits from this platform before its release.On the other side of themarket, however, evenwith

this less-than-a-full prelaunch, Microsoft itself seemedto be sure that game studios and developers who fol-lowed their communication on developer forums willunderstand their preannouncement (Charla 2016). In-deed, there is some evidence that though partial, frominformation provided in the API support and hard-ware specifications, developers realized that Micro-soft was planning to enhance multiplayer experiencethroughNEEaspects of its features. Thiswas borne out byintricate comments in development forums (Staff 2013)that reveal a sophisticated understanding of specificfeature-sets likely in the new release (e.g., enhance-ments made to Kinect for group play, support fromMicrosoft’s Azure for intensive gaming computationsfor multiplayer games). Such observations lead to anunderstanding that perhaps the two sides of the markethave different potentials to form expectations based onthe same amount of a firm’s investment in prelaunch.From a managerial perspective, it is indeed im-

portant to understand the drivers and implicationsof preannouncement strategies in a platform setting.How do market characteristics impact preannounce-ment strategies? What is the impact of a loyal con-sumer base on preannouncement? How is the pre-launch decision affected by the presence of platformagnostic developers?Such managerial questions abstract as a signifi-

cant area of exploration for platform competition.

Chellappa and Mukherjee: Platform Preannouncement: The Strategic Role of Information in Two-Sided Markets1528 Management Science, 2021, vol. 67, no. 3, pp. 1527–1545, © 2020 INFORMS

Essentially, it translates into studying the role of in-formation available in the marketplace in the platformcompetition. Although there are a limited numberof papers (Suleymanova and Wey 2012, Hagiu andHałaburda 2014) that have explored this context,there is no research that has investigated the strategicuse of informationwhereinfirmsmake choices in howthey communicate information about their platformsto the marketplace. Combinedwith elements of price-setting, switching/loyalty aspects in these marketsand the possibility ofmultihoming by developers, ourresearch provides the foundations for a rich under-standing of platform competition. Before we developour model in section 2, we first provide a brief reviewof literature below.

1.1 Review of Relevant LiteratureWe primarily draw upon the following streams ofliterature in this paper. First, we refer to research oncompetition in two-sided markets (platforms), par-ticularly those that examine the role of information.Second, we refer to the research on preannounce-ments that has primarily addressed one-sided mar-kets (products) to motivate the role of information inplatform preannouncement. We also refer to litera-ture related to multihoming and switching to supportour modeling extensions in this paper.

In recent years, competition amongst firms inmanyindustries such as video game consoles, mobile com-munication, and personal computers, has evolvedfrom single-sided products to platforms characterizedby two-sided markets (Parker and Van Alstyne 2005,Eisenmann et al. 2006, Eisenmann 2006, Choi 2010,Anderson et al. 2013, Hagiu and Spulber 2013, Adneret al. 2016). Although product markets have beenextensively studied in the context of network effects(Katz and Shapiro 1985) where an agent’s utility of aproduct depends upon its consumption by other agents,in platform settings (Economides and Katsamakas2006, Zhu and Iansiti 2012), such externalities be-come even more prominent in the presence of same-side and cross-side network effects.

Rochet and Tirole (2003) and Armstrong (2006)provide an elaborate understanding of outcomes inthe platform competition. In modeling competition,both papers incorporate cross-side network effectswhile considering markets with heterogeneous tastepreferences. Through a stylized Hotelling model,both papers examine the competing platforms’ pric-ing strategies for the two sides of themarket.Armstrong(2006) models a duopoly with single-homing agentson either side of the platform and finds that theplatform will more aggressively target that side ofthe market where agents face a lower disutility ofmoving away from their ideal taste preferencesand cause higher cross-side network effect benefits.

Although we examine similar pricing strategies in ourpaper, unlike Armstrong (2006), our model also in-corporates same-side network effects providing ad-ditional insights.Although Rochet and Tirole (2003) and other au-

thors (Caillaud and Jullien 2003, Choi et al. 2005,Hagiu 2009, Boudreau 2010) consider that all agentsare informed about prices, and therefore its impact ondemand. Some othermodels (Gabszewicz andWauthy2014, Hagiu and Hałaburda 2014) have analyzedscenarios where some agents may not be fully in-formed about prices and form strong expectationsthat the platform recognizes in its pricing. Further,consumer expectations (Suleymanova and Wey 2012,Hagiu and Hałaburda 2014) about network effectsand prices may themselves be dependent upon re-lated information available.The role of information in platform competition is

particularly tricky because it can impact five actors—afirm’s users and developers, the competitor, and thecompetitor’s users and developers. In the extant lit-erature,Hagiu andHałaburda (2014) have consideredthe partial awareness of certain information (e.g.,price) on the part of agents (users and developers),whereas others (Suleymanova and Wey 2012) haveconsidered this as agents’ characteristics where theymay form strong or weak expectations. In both thesepapers, they have observed a lack of full informationabout some platform characteristics (e.g., price) maylead to very different outcomes compared with tra-ditional platform duopolistic competition.Through our abstraction of platform preannounce-

ment strategies, we take this investigation of informa-tion to the next level by allowing this element to be apart of the firm’s strategy space. In other words, weendogenize the use of this information as a strategiclever to examine optimal behavior and equilibriumoutcomes. We believe that this is a significant con-tribution to both theory and practice.Product preannouncement has been extensively

studied in Economics and Marketing. In particular,“(preannouncement) creates an awareness of theforthcoming product, spreads information about itscharacteristics, and enables journalists and opinionleaders to evaluate its performance” (Le Nagard-Assayag and Manceau 2001, p. 206). Preannounce-ment may include “information based on the tech-nical attributes and capacities of the new hardwareplatform, on its quality level and market potential, aswell as all the technical information necessary todesign compatible programs” (Le Nagard-AssayagandManceau 2001, p. 206). Particularly, in the contextof experience goods that we study, besides adver-tising (Narayanan et al. 2005), firms may also engagein immersive demonstrations, participate in trades-hows, press conferences, interviews, etc.

Chellappa and Mukherjee: Platform Preannouncement: The Strategic Role of Information in Two-Sided MarketsManagement Science, 2021, vol. 67, no. 3, pp. 1527–1545, © 2020 INFORMS 1529

The extant literature suggests that “preannounce-ments play an important role in (consumer’s) ex-pectation management” because consumers makingadoption decisions are “guided not only by a prod-uct’s current installed base but also by expectations offuture installed base” (Lee and O’Connor 2003, p. 244).This work goes on to reiterate that for “network ef-fect products,” preannouncing has a positive influenceon product’s extrinsic (network effects) value perceivedby customers. Thus, the primary purpose of a pre-announcement is to aid expectation formation ofprices, demand, features, etc. of a future version of aproduct before its actual release (Bayus et al. 2001,Gerlach 2004).

Preannouncements are routine and prevalent inthe technology industry. Choi (2010, p. 314) observethat “. . . product preannouncement has been prom-inent in industries characterized by network effects,such as computer software industry.” Expectationformation regarding network effects as a result of apreannouncement enable consumers to “make a de-cision before the actual launch, and buy the productas soon as it is available” (Le Nagard-Assayag andManceau 2001, p. 206). Su and Rao (2010, p. 670)emphasize the importance of examining the effect ofpreannouncement on the demand side and study how“customers’ perceptions, preferences, and choices”are affected by the “extent to which the details of newproduct are revealed” in the preannouncement. Be-cause a firm’s preannouncement to the market isavailable to its competitors as well, it impacts thefirm’s choice of the level of information that it pro-vides through the preannouncement strategy.

The preannouncement literature has focused onproducts, and although this stream of literature hasprimarily studied monopolies, we begin our analysisby discussing a competitive model for platforms.Although not focusing on two-sided markets, an-other stream of IS literature has focused on emergingbusiness models and optimal pricing strategy in newproduct introduction (or preannouncement) in thecontext of versioning of software products (NiculescuandWu 2014). To the best of our knowledge, this is thefirst paper that addresses the impact of the level ofinformation in platform preannouncement strategieson expectation formation by the two sides of themarket, and how firms, in turn, consider this tochoose a preannouncement strategy and set com-petitive prices. Beyond studying the platform pre-announcement strategies in a duopoly, we are alsointerested in examining such strategies in the contextof both switching and multihoming.

The extant literature on two-sided markets, al-though primarily focused on single-homing agents

has also analyzed the impact of multihoming (byagents in one or both sides of the market) on pric-ing strategies (Armstrong and Wright 2007, Rasch2007, Athey et al. 2016). Armstrong (2006) findsthat “competitive bottlenecks” arise when the mul-tihoming side has their network benefits extractedfully while the single-homing side is subsidized toparticipate in the market. Extant literature is sparse inits analysis of the role of switching of installed bases inplatform competition, although such analysis hasbeen pursued for product competition. In our paper,we also explore the role of switching by the installedbase and multihoming by one-side of the market onequilibrium preannouncement strategies of duopo-listic platforms.

2. ModelWe consider a duopoly characterized by platformcompetition where both firms (i, − i) develop a newversion of their two-sided platform such that theyacquire revenue from (i) selling to a user side (bycharging a price) and by (ii) licensing (by charging afee) to developers on the other side. For the sakeof exposition, we shall develop our model using agaming platform as an example and hence will al-ternately refer to one-side as users or gamers and theother-side as developers. We refer to the entire set ofconsumers of the platform as agents (users and de-velopers). The firms compete in markets where theagents are heterogeneous in their taste preferences; onone-side gamers may differ in their preference forcolor, design of the product (e.g., console, controller,etc.), whereas on the other side, developers areheterogeneous in their preferences for program-ming languages, API management, and other softwaredevelopment environments. Each firm’s platform isidentical in quality and features but differs in thesetaste preferences and other horizontally differenti-ating factors.We abstract the distribution of these taste prefer-

ences on both the user (subscript u) and developerside (subscript d) through a Hotelling line such thatthe platforms i and −i are horizontally differentiatedin their product offering and are located on the ex-tremes. The users and developers are uniformly dis-tributed on their respective taste preferences, and aswith extant location models, each point representsthe ideal product for the user (developer), and theysuffer a disutility (tx)when obtaining a product that islocated some distance x from their ideal point. In ourcase, the parameter t(t> 0), called the transport costcoefficient in locationmodels, captures the strength oftaste preferences. In other words, greater this value,stronger is the users’ (developers’) preference for their


ideal product and thus greater is the disutility theysuffer when they have to choose a product away fromtheir preferred one.

We develop our base model along the lines ofArmstrong (2006) where both sides of the platformderive cross-side network effects, that is, developersgain value from more gamers and gamers gain valuefrommore developers. Similarly, we also assume thatthere is a same side network effect only on the userside, for example, gamers derive more value as moregamers play the same console. Note that while cross-side externalities are present on both sides, we do notinclude same-side externalities on the developer side.Even if more developers can create positive exter-nality in the form of potential labor supply, usuallysuch benefits are canceled out due to competitiveimpacts from more developers. We assume that thecross-side externality coefficient is the same for theuser and developer-side and is given by β and a pa-rameter α captures the same-side externality on theuser-side. In other words, α abstracts the marginalvalue of having an additional gamer to other gamerswhile β abstracts the marginal value of having anadditional developer for gamers and of having anadditional gamer for the developers.

We could also think of the above parameters ascapturing the strength of network effects in that theydetermine the network effects that can be enjoyed byan agent for a given network size. An importantdistinction is that although these parameters are staticin prior work, our model allows for these parametersto be influenced by the platforms’ preannouncementstrategy and therefore are part of the strategy spacefor the platforms. Further, we choose to examineidentical firms with limited differentiation on otherparameters to isolate and specifically examine het-erogeneity in the choice of platform preannounce-ment strategies.

2.1 Platform Preannouncement StrategiesOur main aim in this paper is to examine a firm’schoice of preannouncement of the new version of itsplatform. Literature in marketing states that evenfor products (let alone platforms), in addition toR&D related investments, new product developmentstrategy should also include methods to communi-cate to the market about the firm’s course of action(Eliashberg and Robertson 1988, Le Nagard-Assayagand Manceau 2001, Ofek and Turut 2013). In ourpaper, we define three preannouncement strategies—aformal/full strategy, an informal/partial strategy, and ano-preannouncement strategy as the strategic choice setfor a platform. Fundamentally, the strategies capturethe “extent” of product details (Su and Rao 2010)released to the marketplace. Preannouncement of suchNEE aspects of the platform has a positive influence on

the platform’s network effects value as perceived byconsumers (Lee and O’Connor 2003) and enablesconsumers to “make adecisionbefore the actual launch,andbuytheproductas soonas it is available” (LeNagard-Assayag and Manceau 2001, p. 217). We specificallymodel the mechanism through which preannounce-ments impact platforms with network effects. In ourmodel, the extent and quality of information in thepreannouncements help gamers and developers esti-mate the strength of network effects and thus influencethe overall expectation of the total network effects to beenjoyed from purchasing the platform.

2.1.1. Formal (or Full) Preannouncement Strategy. Afull/formal preannouncement strategy is one wherethe firm preannounces the release of its platformwitha variety of prelaunch activities wherein it fully in-forms the marketplace about the network effect en-gendering aspects and other characteristics of the newversion prior to actual launch. Prelaunch activitiesmay include everything from advertisements to immer-sive demos of NEE features such as multiplayer gam-ing, etc. To fully convey information about such NEEelements, the firm has to not only invest in obtainingexpensive CES booths and other advertising channelsbut may also have to develop demos and giveawaysthat can be sustained through its partners. We denotethe net investment to fully inform the marketplace bya one-time cost c.Although the firm is burdened by this cost upfront,

it also stands to benefit from this strategy as this al-lows both users and developers to fully take intoaccount all the NEE aspects of the new version andconditions of the new marketplace, in setting theirexpectations and allocating their budgets. From amodeling perspective, this strategy implies that whenthe full extent of information about the platform isconveyed, gamers/developers can estimate the fullstrength of the network effects that the platformcan offer.We commonly observe preannouncements predat-

ing new releases during Christmas and other holi-days (Dranove and Gandal 2003, Poulter 2013). Akey reason is that firms are keen to influence bud-getary decisions and other consumer expectations inadvance of actual purchase. All activities under thisstrategy are denoted with a subscript F.

2.1.2. Informal (or Partial) Preannouncement Strategy. Wedefine informal strategy as one where the firm choosesnot to fully inform the agents regarding the NEE as-pects of its new platform version. In such cases, firmsmight even eschew the use of traditionally expensivechannels (e.g., trade shows with immersive demos)and perhaps largely rely on free/informal channels totransmit information regarding the platform’s NEE


aspects besides other features. For example, an in-formal preannouncement may provide such infor-mation through pictures, videos, and technical detailson fan pages, discussion forums, social media chan-nels, etc. instead of an expensive hands-on andimmersive experience that characterizes a formalpreannouncement (as discussed in Section 1).

Thus, a firm pursuing an informal preannounce-ment strategy may only incur a lower prelaunch costin informing the market regarding the NEE aspects ofits platform, compared with a formal preannounce-ment strategy. In our model, we capture this ad-vantage to the firm through a cost discount parameterk(k ∈ [0,1]) such that the total preannouncement costincurred in pursuing this strategy is kc.

The second element of this strategy pertains to theexpectation formation of the two sides correspondingto the preannouncement. When the platform pursuesa partial preannouncement strategy providing lessthan the full extent of information, agents estimatethe strength of network effects to be correspondingto the “the extent to which the details of new prod-uct are revealed” (Su and Rao 2010, p. 670). In otherwords, as a result of the platform’s preannouncementstrategy, the agents end up underestimating thestrength of network effects that the platform couldhave provided. This in turn implies that the agents’expectations of the platform’s network benefits arealso attenuated. At one end, there are cost advan-tages to this strategy, but it also potentially limitsthe expected surplus on the agents’ side. It is also notclear as to how competition may influence the choiceof this strategy and hence further examination is re-quired. All activities under this strategy are denotedwith a subscript I.

2.1.3. No-Preannouncement (or No Information) Strategy.We define this strategy as one where the firm choosesnot to release any information related to the networkeffects of the new version of the platform to thegamers or developers—prior to actual release. Notonly does the firm not incur any cost in pursuingprelaunch activities, but it also provides no new basisfor estimating the strength of networks effects in themarket. As a result, agents rely purely on the intrinsicvalue of the platforms in setting their expectationsand allocating their budgets for purchase/licensingof the new version of the platform.All activities underthis strategy are denoted with a subscript N.

2.2. Relevance of Switching and HomingThe examination of platform preannouncement strate-gies will be incomplete without taking into accountcertain market conditions commonly observed in plat-form competition. Note that our investigation seeksto shed light on expectation formation because of

platform’s externality related information availablein the marketplace. In this context, installed baseplays an important role if the market conditions donot allow for switching to the other platformwhen thenew version is released. In suchmarkets, the installedbase may be considered as loyal agents. However,when the installed base also considers switching, theyend up affecting agents’ expectations regarding totalnetwork effects.Similarly, in platform models of competition, even

when defined by Hotelling-type conditions, manyauthors (Armstrong and Wright 2007, Rasch 2007,Athey et al. 2016) have examinedmarketswhere someagents end up serving both sides of the market.Commonly known as multihoming, this refers tosituations when developers might find it optimal todevelop games for both platforms even if they arepositioned at the extremes of the linear market. With-out rigorous analysis, it is not clear as to how pre-announcement strategies are influenced by such marketconditions. In the gaming context, it is rather commonto observe multihoming by some developers who de-velop games for both consoles, whereas other devel-opers choose one platform (single-home). For example,when technology or developer’s innovation is easilyportable between platforms (gaming consoles), thensuch developers will multihome. On the other hand,developers may choose to single-home when tech-nology is proprietary (e.g., video games such as NFL,NBA, etc.) and/or porting development to a com-peting platform is difficult due to technology in-compatibility and/or steep learning curve.To examine preannouncement strategies under mar-

ket conditions of switching and multihoming, we firstdevelop the base case where the agents are single-homing and where their installed base does not switchwhen a new version of a platform is released.We denotethis by SH-NSW. Following this base case model, wedevelopourmodels for the case,where agents still single-home but where now their installed base considersswitching when a new version is released, denoted bySH-SW. In Section 3.2, we introduce the option tomultihome and examine both cases where agents donot switch (MH-NSW) and consider switching (MH-SW).Independent development of preannouncement

strategies in each of the four cases allows us to explicitlyisolate the relationship between strategic choices andtypical market conditions. Of particular importanceto modeling preannouncement is the element of in-formation (Suleymanova and Wey 2012, Gabszewiczand Wauthy 2014, Hagiu and Hałaburda 2014) andwhen it is available to influence expectation forma-tion. Hence, our choice of the base case, where agentssingle-home and do not switch, to illustrate the dif-ferent informational nature of the three preannounce-ment strategies.


2.3. Timeline When Agents Do Not Switch Platformsand Only Single-Home (SH-NSW)

Tomaintain a general understanding of the role of theinstalled base, we develop the initial model with firm-specific identifiers. However, note that throughoutthis paper, we shall focus on results for identical firmsfor two reasons: First, examination of symmetricfirmsis the norm in platform competition and followingthis standard allows us to easily compare our resultswith extant literature. Second, our interests are pri-marily regarding preannouncement strategies andthe role of market conditions such as allowance forswitching and multihoming. However, developingthe base model in a generic fashion should allow thereader to see the differential role played by this factorwhence switching is allowed and not (it is easyenough to derive closed-form solutions evenwith thisheterogeneity though expressions are unwieldy).

The timeline of the new platform release precededby prelaunch activities (including preannouncement)is illustrated in Figure 1 and developed as follows:

Stage 0: Firm i starts with an installed base ofgamers/users (mi

u) and developers (mid) of the initial

version of its platform that are each normalized to 1,that is

miu +m−i

u � 1; mid +m−i

d � 1. (1)

Stage 1: In stage 1, both firms develop a version oftheir platforms with new features including NEEaspects that create and benefit from network effects.We assume that these additional features increase thesize of the overall market by bringing in new agents.

This assumption captures the fact thatmany productsin the high-tech industry undergo continuous inno-vation, and improved versions introduce technolog-ical advances that bring in a new market of agents.For example, in the video game industry, the market(and the number of games) increased substantiallywith the introduction of new features like motionsensing, faster CPU, collaborative gaming, etc. (Hall2015). The proportion of this newmarket garnered bythe firm i can be expressed as the price/market-dependent user and developer demand functionsniu and nid. Although the new markets may expand toany arbitrary size, we assume a unit market as givenby the following equation:

niu + n−iu � 1; nid + n−i

d � 1. (2)

Further, because these platforms (e.g., video gameconsoles) have experience-good like qualities, con-sumers can get an adequate understanding of thesenew aspects either because (i) they purchase and expe-rience the product in Stage 2, or (ii) firms have engagedin preannouncement activity in Stage 1 to help theconsumers understand as much as possible about theseaspects, before the platform is released. As the firmshave to set prices before agents can purchase/license inStage 2, the platforms have to make choices regard-ing their preannouncement strategies here in Stage 1.They can choose from among a formal strategy(F), aninformal strategy(I) or a no-preannouncement strat-egy(N), and they incur marketing and preannounce-ment costs of c, kc and 0, respectively.Corresponding to the preannouncement strategies

(that differ in the level of information released to themarket), gamers and developers estimate the strength

Figure 1. Timeline of the Preannouncement Game


of network effects, and therefore form expectationsregarding the eventual network effects from the newversion that would be released in Stage 2.

In related literature that has considered the role ofinformation in expectation formation, typically con-sumers are a priori assumed to vary in their informed-ness of prices (Hagiu and Hałaburda 2014) or possessweak or strong expectations about market share be-forehand (Suleymanova andWey 2012). In ourmodel,consumers that are homogenous in all aspects maystill have different expectations of platforms as aresponse to the firms’ choice of preannouncementstrategy. It is therefore important to develop the ex-pected utilities under each strategy separately.

First, let each firm’s new version provide someintrinsic or standalone value (θ), to the agents, that isindependent of network effects (e.g., single playergames that may be shipped along with a console). Weassume that both firms produce platforms identical inthis value as we do not wish to create any bias towardone platform and will be able to fully tease out thedifferences created by the firm strategies.

The second part of the utility function is the ex-pected value from network effects. Agents makingadoption decisions are “guided not only by a prod-uct’s current installed base but also by expectations offuture installed base” (Lee and O’Connor 2003, p. 248).We shall first consider the case when agents are single-homing and do not switch (SH-NSW) and hence theinstalled base of users given in Stage 0 remain with thefirmwhen the new version is released. The network size,therefore, is the sum of the installed based plus any newagents that might arrive with the new version. Note,however, that the estimated strength of network effectsis still a function of the platform’s preannouncementstrategy. In our model, both the installed base and thenew market react similarly to the extent of informa-tion released during preannouncement. In other words,the installed base, even if loyal, is assumed to have nospecial understanding of the new version’s strength ofnetwork effects in comparison to the new market.

We first develop the utility function for users anddevelopers under the formal preannouncement strategycase. In case of a formal preannouncement by firm i, theusers’ utility from purchasing from this platform canbe written as

Uiu,F � θ + α

[mi

u + niu(pi, p−i, li, l−i

)]+ β

[mi

d + nid(pi, p−i, li, l−i

)]− pi, (3)

where pi and li are the prices and licensing fees. Thesecond term captures the same-side externality ben-efits, whereas the third term captures the cross-sidebenefit as a result of the developer demand. As dis-cussed earlier in this section, α and β are the same-sideand cross-side network effect coefficients or the strength

of network effects, that is, they represent the marginalvalue to a user from additional user and developer, re-spectively. Also, note that the new demand faced by thefirm is a sumof thedemand fromits installedbase aswellas the new demand introduced in Equation (2). Thisnew demand that the firmwill acquire is a function ofits prices and licensing fees as well those of itscompetitors. Similarly, we can write the utility of thefocal platform’s developer as

Uid,F � θ + β

[mi

u + niu pi, p−i, li, l−i)]

− li.(

(4)

Note that the same-side externality is absent from thedeveloper side for reasons discussed earlier.The “extent to which the details of new product are

revealed” in a preannouncement is known to influ-ence consumers’ response to the new product (Suand Rao 2010, p. 670); in the informal or partial pre-announcement regime, the limited nature of the pre-announcement provides an incomplete understand-ing of the platforms’NEE aspects to the agents. In ourmodel, we introduce a parameter µ∈ [0,1] that cap-tures the underestimation of the strength of net-work effects due to the lesser extent of informationabout NEE in the partial/informal strategy. Thus,the same-side and cross-side strength of network ef-fects become αµ and βµ, respectively, and we canwrite the total expectation of network effect fromthe market (of installed base and new consumers) asαµ[mi

u + niu (pi, p−i, li, l−i)] from the user sided andβµ[mi

d + nid (pi, p−i, li, l−i)] from the developer side.Thus, a user’s utility under the informal regime canbe written as

Uiu,I � θ + µα

[mi


)]+ µβ

[mi


)]− pi. (5)

Similarly,we introduceaparameterδ∈ [0,1] to capture anequivalent understanding on the developer side suchthat the developers’ utility in Stage 1 can be written as

Uid,I � θ + δβ

[mi


)]− li. (6)

Although not explicitly modeled in this paper, we caneasily see that the parameters µ and δ can be thoughtof as being proportionally related to the discountfactor k, that is, when k � 1 or when the full set ofprelaunch activities are pursued (and all of the costsincurred), µ and δ will also converge to 1 thus re-ducing Equations (5) and (6) to Equations (3) and (4),respectively. To allow for the fact that developers—a technically advanced and industry-tuned group,understand firms’ announcements better than end-users, we assume δ≥µ.Following the discussion, we can see that a no pre-

announcement strategy corresponds to the situationwhen k, µ and δ are all zero, that is, no announcement


is made, no cost is incurred, and there is no specificunderstanding of the strength of network effects.Thus, the utilities derived by users and developerscan be written as follows:

Uiu,N � θ − pi (7)

Uid,N � θ − li. (8)

Stage 2: Stage 2 corresponds to the release of the newversion where firms also reveal their prices and li-censing fees. Along the lines of literature in economicson network effects, in our model also, agents “willbase their purchase decision on expected networksizes,” and “must make their purchase decisionsbefore the actual network sizes are known” (Katz andShapiro 1985, p. 426). In other words, there is norevision to their expectations formed in Stage 1, andhence firms have to take that into account in settingthe prices and licensing fees in Stage 2. Under thefulfilled rational expectations equilibrium, the pricesset at this stage should match the expectations of theusers and developers in Stage 1. All agents adopt thenew version released by the platform.

We can now write the objective functions of thefirms under the three different preannouncementstrategies, formal, informal, and no-preannouncementas follows:

max{pi,li}

πiF � pi

[mi


)]+ li

[mi


)]− c (9)

max{pi ,li}

πiI � pi

[mi


)]+ li

[mi


)]− kc (10)

max{pi ,li}

πiN � pi

[mi


)]+ li

[mi


)]. (11)

2.4. The Preannouncement Game (SH-NSW)Wemodel the duopoly as a simultaneous move gamewhere a firm i chooses its strategy Si � {F, I,N} inStage 1 and sets prices and licensing fees in Stage 2taking into account the agents’ expectations formed inStage 1. Consistent with the tradition in platformcompetition literature, we are only interested in pure-strategy subgame perfect Nash equilibria (SPNE)—mixed strategies have limited explanatory ability andmanagerial meaningfulness.

For the base model, we largely follow Armstrong(2006). Our market is developed as a Hotelling seg-ment such that the users and developers are eachuniformly distributed (8~ [0,1]) over a unit line. Thefirms are located at the extremes (maximally differ-entiated in their platform offerings), and each agent

location represents the ideal platform for that user ordeveloper. Therefore, in using or developing for aplatform that is not at the ideal location, the agentssuffer a disutility as defined by their transport costs,as a function of how far away they are from the offeredplatform and the transport cost coefficient t. Through-out the paper, along the lines of extant literatureon platform competition (Choi 2010, Hagiu and Lee2011), we examine a variety of symmetric equilibriafor a duopoly of symmetric firms.

2.4.1. When Firms Engage in Formal PreannouncementStrategies. Consider the case where the firms pursuethe strategy of preannouncing their new platformrelease. To determine the equilibrium prices and li-censing fees, we need to first find the price-dependentdemand that each firm faces in this market. Given theHotelling specification, from Equation (3) a user lo-cated at a distance x from firm i derives a net utilityUi

u,F − tx. In equilibrium, the indifferent user is locatedwhere the market is split between the two firms, andtherefore we have

Uiu,F − tniu,F � U−i

u,F − t[1 − niu,F

]. (12)

Now, substituting for the utility functions and denotinga difference in installed bases of users and developersbetween firms i and −i as Δmu and Δmd, respectively,we can write the price/licensing fees dependent userdemand as

niu �12

+ −t [pi − p−i] − β

[li − l−i

] + [αt + β2

]Δmu + βtΔmd

2[t[t − α] − β2

] .

(13)

Along these same lines, we can also write the speci-fication for the demand function on the developer sideof the market as

nid �12

+ −β [pi − p−i] − [

t − α][li − l−i

] + βtΔmu + β2Δmd

2[t[t − α] − β2

] .

(14)

Substituting Equations (13) and (14) in the objectivefunction given by Equation (9), we can solve for theoptimal prices and licensing fees for this case.

2.4.2. When Firms Engage in Informal Preannounce-ment Strategies. When the firms pursue an informalstrategy in Stage 1, they provide partial informationto the market about their platform’s NEE aspects.Further, the estimates of the strength of network ef-fects at this stage may be different for the two sides ofthe platform, as captured by the extent of information


µ (for the users) and δ (for the developers) relayed inthe informal preannouncement. This eventually af-fects the expectation of network effects.

Therefore, in Stage 2 when the firm sets prices, ithas to take into account this lower extent of infor-mation available in the informal preannouncement,its impact on an agent’s estimation of strength ofnetwork effects, and their corresponding expectationof total network effects. Observe that none of theother aspects of the horizontally segmented marketare different from the formal preannouncement case,that is, consumers are single-homing and do notswitch. Similar to Equation (12), we can find the lo-cation of the indifferent user in the informal case aswell. Substituting Equation (3) for utilities, we canwrite the user and developer demand facing thefirm as

niu,{I,I} �12+−t[pi − p−i

] − βµ[li − l−i

]+ [

µαt + µδβ2]Δmu + µβtΔmd

2[t[t − αµ

] − β2δµ] ,

nid,{I,I} �12+−βδ [pi − p−i

] − [t − αµ

][li − l−i

]+ β

[δt + [

δ − µ]α]Δmu + µδβ2Δmd

2[t[t − αµ

] − β2δµ] . (15)

Substituting Equation (15) in Equation (10) we cansolve for optimal prices and licensing fees for firmswith identical installed bases.

2.4.3. When Firms Engage in No-PreannouncementStrategies. At the extreme, in the strategy space isthe option to not engage in any preannouncement atall. Although a simplistic view is that it saves on theprelaunch costs, we knowwell that a more significantimpact of this strategy is due to the lack of informationregarding the NEE aspects of the new version of theplatform. In the absence of such information, theusers and developers cannot form estimates regard-ing the strength of network effect (α, β) and theirexpected utility from the platform is strictly confinedto the platform’s intrinsic benefits. Therefore, alongthe lines of Sections 2.4.1 and 2.4.2, we can write theagent demands as

niu,{N,N} �12−[pi − p−i

]2t

, nid,{N,N} �12−[li − l−i

]2t

. (16)

Substituting these demands into the profit functiongiven in Equation (11), we can now derive the pricesand licensing fees for this case.

2.5. Equilibrium Outcomes in SH-NSW CaseThe first step is to derive the optimal prices and li-censing fees for a candidate equilibrium strategychosen by the firms. To do so, substitute the user anddeveloper demand functions that were derived from

the Hotelling line, in the profit function. Simulta-neously solving the first order conditions of the profitfunction of both firms with respect to piand li, we getthe prices and licensing fees as a function of the in-stalled bases. Because our focus is on identical firms,we can get candidate equilibrium prices and licensingfees for the focal firm by setting the installed bases ofboth firms to be the same (mi

u � m−iu , mi

d � m−id ). Note,

however, for these prices and licensing fees to beSPNE, not only should this preannouncement strat-egy dominate over others, but it should also be fea-sible in the region as defined by the various param-eter relationships. Thus, our candidate equilibriumprice-licensing fee pairs are pi{F,F}, l

i{F,F} for formal pre-

announcement strategy; pi{I,I}, li{I,I} for informal pre-announcement strategy; and pi{N,N}, li{N,N} for no-preannouncement strategy.For any of the pairs to be SPNE prices and licensing

fees, we require the profits from that strategy to behigher than the other two strategies. For example, inthe region, where {F, F} is feasible, pi{F,F} and li{F,F} areequilibrium price and licensing fees if and only ifπi*{F,F}>max[πi*

{I,I},πi*{N,N}] and π−i*

{F,F}>max[π−i*{I,I},π

−i*{N,N}].

Further, we must ensure that various constraintson parametersα, β,µ, δ, etc., are also satisfied besidesindividual rationality conditions. By imposing allparameters to be nonzero and keeping µ, δ ∈ (0,1), wenow consider other economic constraints and char-acterize them through the strength of preferenceparameter(t). Such constraints include individual par-ticipation constraints of users and developers wherebyall agents derive a nonnegative utility. Because agent’sutilities are decreasing in t, such constraints provide alower bound for t. An upper bound on t is provided byconditions ensuring that the firms’ prices, licensingfees and profits are nonnegative, and such profits areconcave in prices and licensing fees. Considering allthese constraints together, through reduction, we cancharacterize the potential bounds where formal, in-formal and no-preannouncement strategies are SPNE.In our paper, we assume that the intrinsic value (θ)

provided by the platforms is high enough so that cer-tain trivial situations, for example, where utilities maynot be positive, may be dismissed. Through lemmas andpropositions, we examine equilibrium preannounce-ment strategies and characterize their choice throughthe strength of preferences in the market. The agents’strength of preferences characterizes different marketrepresentations such as in markets where users/gamersare very particular about their tastes and/or where de-velopers are more partial to particular technical productdesigns (higher values of t). Alternatively, in somemarkets, agents may have a very minimal attachmentto their tastes and be readily willing to considerplatforms that are not their ideal ones (lowvalues of t).


Unlike existing work that primarily examines firms’pricing/licensing fees when they engage informed oruninformed consumers (or those that may hold weakvs. strong expectations), our goal in this work is toinvestigate the feasibility of using information as astrategic tool. Lemmas 1–3 tell us that the optimalityof these three different preannouncement strategiescan be characterized through the disutility sufferedby agents in this markets and hence the strength ofpreferences therein.

Lemma 1. In single-homing markets where the agents donot switch and where they possess strong taste preferences(defined by t*{F,F}), firmswill pursue formalpreannouncementstrategies in SPNE. Equilibrium prices and licensing feesare given by pi*{F,F} � t − α − β and li*{F,F} � t − β. ■

The prices and licensing fees are decreasing in thestrength of network effects even if the utilities ofusers/developers are increasing in them. This is spe-cific to the fact that this is a competitive model; in amonopoly, strength of network effects serves to increaseprices. Further, prices and licensing fees are increasingin the transport cost or taste preferences. This is indeedan artifact of the Hotelling abstraction and is consistentwith other works on platform competition as well. Wecan now contrast our results against other extents ofinformation when firms pursue informal preannounce-ment or eschew any prelaunch activity.

Lemma 1 refers to markets where agents have verystrong preferences for their platforms implying sig-nificant disutility costs are incurred by them inconsuming/developing for the nonideal platforms atthe extreme. The bounds of these preferences are givenby (Equation (A.5) in Appendix A)

t*{F,F} ∈(2θ3

+[µ + δ

]β

3,2θ3

+ 2β3

).

Our results suggest that the equilibrium strategy is aformal (or full) preannouncement of the newplatformfor such markets. The economic intuition is providedby the tradeoff between the loss of surplus from high tand the extraction of expected benefits from net-work effects through a complete understanding of thestrength of network effects offered by the platform’snew version. In other words, firms need to make aformal preannouncement even if it increases competi-tive intensity because they have to offset the disutilitycreated by the strong taste preferences.

Lemma 2. In single-homing markets where the agents donot switch and where they possess moderate taste prefer-ences (defined by t*{I,I}), firms will pursue informal pre-announcement strategies in SPNE. Equilibrium pricesand licensing fees are given by pi*{I,I} � t − µα − δβ andli*{I,I} � t − µβ. ■

First, note that the prices and licensing fees arehigher than those in the formal case. In other words,releasing less than full information about the plat-forms’NEE aspects lessens the competitive intensity.Not surprisingly, the extent towhich prices are higherthan those in the informal case is a function of theextent of information that the agents have about theNEE aspects of the forthcoming version of the plat-form. Lower the extent of this information conveyedthrough a preannouncement, higher are the prices,suggesting that perhaps eschewing all prelaunch activ-ity might be a better strategy for the firm. However,we know that no-preannouncement is not an equilib-rium option for markets whose strength of taste pref-erences are given by (Equation (A.6) in Appendix A)

t*{I,I} ∈(2θ5,2θ3

+[µ + δ

]β

3

).

Further, observe that although prices are a function ofboth extent of information parameters (µ, δ), licensingfees is directly a function of the level of understandingon the user-side—the reason being that there are nosame-side network effects on the developer side.Although a lower extent of information regardingNEE aspects of the platform leads to underestimationof the strength of network effects, it also attenuatesthe impact of strength of network effects on the pricesand licensing fees.

Lemma 3. In single-homing markets where the agents donot switch and where they possess weak taste preferences(defined by t*{N,N}), firms will pursue no preannouncementstrategies in SPNE. Equilibrium prices and licensing fees aregiven by pi*{N,N} � t and li*{N,N} � t. ■Lemma 3 points out that there is a pure-strategy

equilibrium where firms eschew preannouncementand where, in the absence of estimates regarding thestrength of network effects, agents form expectationsbased on the platform’s intrinsic benefits alone. Thus,there is no impact of the strength of network effects onthe prices and licensing fees in this strategy. Also,note that among the three cases, the strategy of no-preannouncement allows the firms to charge the highestprices and licensing fees and therefore the firms alsoderive higher profits compared with the other twostrategies. The bounds on agent’s strength of tastepreference that characterize such markets are givenby (Equation (A.7) in Appendix A)

t*{N,N} ∈(0,2θ3

).

The result echoes recent findings in the economics oftwo-side platformswherein firms in a duopoly derivehigher profits even when only one side has less in-formation (Hagiu andHałaburda 2014). In comparing


Lemmas 1–3, we not only observe that prices andprofits are highest when there is no informationabout the future release, but we also see that theexternality benefits from the new features only serveto lower prices for the firms. Earlier work (Armstrong2006, Hagiu 2009) also alludes to similar results—thegeneral understanding is that a combination of theHotelling abstraction and competitive dynamics causesthis outcome. Note that the agents’ surplus is increas-ing in the strength of network effects. However, thefirms are unable to extract this surplus due to com-petition, which in turn is intensified by network ef-fects. Because such network effects are a function ofthe agents’ estimates of strength of network effects,which are formed as a result of information aboutthe platform’s NEE aspects conveyed through thepreannouncement, the economic intuition behind thesuperiority of the no-preannouncement strategy be-comes evident.

Thus, in markets where the agents are not veryparticular about their ideal platforms, that is, wherethey can consume/develop for a platform other thantheir ideal ones with little disutility, our results suggestthat no-preannouncement is the SPNE. Firms prefer notto preannounce the NEE aspects of their new platformversion as this information only serves to increase thecompetitive intensity and further the firms can chargehigher prices/licensing fees when the transport costparameter t is high. However, we also know that theagents’ utility is decreasing in this parameter so theremay be markets where such taste preferences aremoderately strong enough such that the agents do notderive positive utility. In such markets, our resultssuggest that the SPNE is informal preannouncementwherein the agents will only have a partial under-standing of the new version’s NEE aspects. Eventhough the transportation cost parameter is relativelyhigher in these markets, equilibrium prices and li-censing fees are lower (relative to no preannounce-ment case) due to the underestimation of strength ofnetwork effects by agents that result in lowered ex-pectation of total network effects. Finally, in marketswhere agents incur an even higher disutility frommoving away from their ideal taste preferences, firmsbenefit from providing full information (operation-alized through a formal preannouncement) regardingtheir new version to the agents.

In summary, Lemmas 1–3 form the crux of ourinquiry into platform preannouncement strategies,which allows us to conclude that when firms canchoose the extent of the information they put outabout their platforms, even if the information pro-vides value to the agents, itmay not always be optimalto do so in a competitive market.

3. Preannouncement Strategies in theCase of Switching and Multihoming

As a part of our investigation, we are also interestedin understanding how the optimality of these pre-announcement strategies may be affected by othermarket settings. In particular, in this section, we relaxtwo major assumptions incorporated in Section 2—those about the actions of the installed base and theoption to develop for both platforms. Throughout, weconsider SH-NSW (discussed in Section 2) as the basecase against which we shall first examine the role ofloyalty, that is, what happens when the installed baseof agents do not automatically buy their originalplatform but in-fact also consider the competitor’snew version. Following that, in Subsection 3.2, weshall examine the case when the developers con-sider the additional possibility of developing for bothplatforms as opposed to having to choose between thetwo. Although the former has not been commonlyincorporated into the platform competition literature,the latter is calledmultihoming and is often examinedin the context of such duopolies.

3.1. Single-Homing Markets Where the InstalledBase of Agents Can Switch (SH-SW)

The installed bases of agents for each firm has so farrepresented a set of loyal agentswho always purchasethe new version of the platform they were earlierusing/developing for. However, akin to the newagents who come into the marketplace as a result ofthe new platform versions, the installed-base toocould re-evaluate their decision to choose a particularfirm. To examine this scenario, for brevity, we onlydevelop the informal preannouncement strategy inthis section as the other two strategies can easily beunderstood as extreme cases of this strategy.We now write the expected utility of a user in this

setting as

Uiu,I�θ+µα

[2niu

(pi,p−i,li,l−i

)]+µβ

[2nid

(pi,p−i,li,l−i

)]−pi .(17)

Note that in this scenario (SH-SW), all users (installedbase as well as new users) consider the tradeoff be-tween choosing one platform vs. the other. Becausethe entire market of users is in play, we ignore the sizeof the installed bases (mi

u) and derive the price-dependent expected demand considering both theold and new market of users due to the new version.Similar to Equation (5), under informal preannounce-

ment strategy in this setting there is only partial infor-mation regarding the NEE aspects of the new version ofthe platform, and therefore unlike the formal strategywhere agents estimate the full strength of network effects


(α, β), agents underestimate the strength of networkeffects. This underestimation of the strength of net-work effects corresponds to the extent of informationregarding NEE aspects in the preannouncement.Thus, users and developers underestimate theircorresponding strengths of network effects to be µαand δβ respectively, and the developer’s expectedutility can be written as

Uid,I � θ + δβ

[2 niu

(pi, p−i, li, l−i

)]− li. (18)

As described in the timeline given in Section 2.3,agents form rational expectations of the total networkeffects in Stage 1 and platforms release their versionsin Stage 2 by taking into account these expectations.Therefore, we can write the objective function of aplatform that engages in information strategy whenall its agents single-home but where the installedbases are not loyal when the new version is re-leased, as

max{pi,li}

πiI � pi

[2niu

(pi,p−i, li, l−i

)]+ li

[2nid

(pi,p−i, li, l−i

)]− kc.

(19)

As before, for this to be a fulfilled expectations equi-libria, firms will set prices and licensing fees such thatthe expectations of the agents are met. Furthermore,we can also develop the other two strategies fromEquations (17), (18), and (19) by settingparametersµ, δ,and k to be 1 (for formal preannouncement) or 0(for the no-preannouncement strategy). Following thegame development in Section 2.3, we can also developprices and licensing fees in this case as given inLemma 4, following which we present the implica-tions for platform profits in Proposition 1.

Lemma 4. (SH-SW). In single-homingmarkets where agentsmay switch, firms will pursue a no-preannouncement strategy,an informal preannouncement strategy and a formal pre-announcement strategy in markets defined by weak, moderateand strong agent preferences respectively. The correspondingSPNE prices and licensing fees

pi*{N,N} � t

li*{N,N} � t

pi*{I,I} � t− 2µα−2δβ

li*{I,I} � t− 2µβ

pi*{F,F} � t− 2α− 2β

li*{F,F} � t− 2β. ■

Proposition 1. When platforms choose to preannounce(with a formal or informal preannouncement), platformprofits are strictly higher in markets where the installed baseof agents does not switch platforms. In this case, both pricesand licensing fees are strictly higher than in markets wherethe installed base can switch platforms.

First, note that our results show that the appro-priateness of the preannouncement strategy is stillcharacterized by the strength of preferences. The boundson t for formal, informal, and no preannouncement

strategies are given by Equations (B5), (B6), and (B7) inOnline Appendix B. We find that our recommendationfrom Lemmas 1–3 continues to hold except that theSPNE feasibility regionswhen installed base is not loyal,are different. From the prices referred to in Lemma 4,we can see that unlike the case where there was noswitching, firms are forced to lower their prices in thesemarkets. Given that, we do not allow for subsidies(negative prices) in these models; our results tell us thatthere is aminimumstrengthofpreference that is requiredbefore we can even examine a pure-strategy equilibriumin preannouncements.Much of the differences between markets with and

without switching can be attributed to the increasedcompetitive intensity that switching brings to bear.Although firms could comfortably rely on the loyaltyof the installed base in the earlier context, with thisgroup also at play, firms need to lower their prices tobe attractive. A second aspect unique to models withnetwork effects stems from the fact that with a loyalinstalled base, some surplus from network effect isreadily built-in for the agents (except in the no-preannouncement strategy), whereas on the otherhand, total network effects expectations are fullyprice-dependent with switching. Given that prices,licensing fees and profits are always higher with no-preannouncement, in markets where the installedbase is not loyal, firms pursue this strategy for evenrelatively higher values of t compared with marketswith the loyal installed base.Because markets with switching possibilities en-

dow benefits to the agents, it behooves us to examineconsumer and social welfare in market settings withandwithout switching.However, beforewe do that inSection 3.3, we present an analysis of multihomingmarkets that are often of great interest to platformresearchers.

3.2. Multihoming MarketsMultihoming, distinct from switching, refers to thesituation when an agent develops (or consumes) notjust for one but both platforms. In the context ofgames and other app development, multihoming ismore likely, particularly becausemany games or appshave embedded intellectual property and they canmaximize their return by developing for both parties.Although multihoming has typically been studiedwith the assumption that all agents on one sideof the market multihomes (general multihoming)(Armstrong 2006), there is limited research that hasexamined this phenomenon particularly where thedecision to multihome itself is endogenized (Rasch2007, Choi 2010) such that while some may multi-home, others single-home. We follow this approachand endogenize multihoming as it encompassesthe general multihoming situation. Therefore, in this


section, we investigate preannouncement strategies inmarkets where developers can multihome—we firstexamine markets where the installed base is loyal(MH-NSW) followed by the case where both multi-homing and switching is feasible (MH-SW).

Continuing along the lines of the Hotelling ab-straction introduced in Section 2.4, we now allow forsome developers who may potentially develop forplatforms at both ends. Not surprisingly, the oneswho would consider this opportunity are located inthe middle and representative of developers that aremore or less equally attracted to platforms at twoends. Unlike the earlier scenarios, we shall first de-velop themultihoming agents’ utility and then followup with the single-homing side. Note that we usesuperscript i, − i to denote the fact that the agentsdevelop for both platforms.

Ui,−id,I �θ + δβ

[mi

u +m−iu + niu


)+ n−i

u


)]− li − l−i. (20)

Equation (20) captures the fact themultihoming agentbenefits from the installed bases of both platforms(mi

u,m−iu ) as well as the expected demand of both plat-

forms from the new version (niu,n−iu ) and indeed these

developers pay the licensing fees of both platforms.On the other hand, the single-homing developer willderive the same utility as in the previous cases, forexample, as in Equation (6). Note that since the pro-portion of single-homing and multihoming developersis endogenously determined we know the market issplit into three types—developers of platform i(Ui

d,I>max(Ui,−i

d,I ,U−id,I)), platform−i(U−i

d,I >max(Ui,−id,I ,U

id,I)), and

multihomingdevelopers i, − i(Ui,−id,I >max(Ui

d,I,U−id,I)). In

other words, the firms need to take into account thissingle vs. multihoming decision of the developers insetting their licensing fees.

On the other side, even though a gamer only paysfor one platform i, he derives cross-side network ef-fects from both the single-homing developers ofplatform i as well from the multihoming developers.Therefore, we can now write the gamer’s utility as

Uiu,I �θ + µα

[mi


)]

+ µβ

[mi


)+ni,−id


)]− pi . (21)

Note that we also now need to rewrite the profitfunction as the firm now also gains from the multi-homing developers. Therefore, we have

max{pi,li}

πiI�pi

[mi

u+niu(pi,p−i,li,l−i

)]+li

[mi

d+nid(pi,p−i,li,l−i

)+ni,−id

(pi,p−i,li,l−i

)]−kc. (22)

Lemma5 (MH-NSW). In multihomingmarketswhere agentsdo not switch, firms will pursue a no-preannouncementstrategy, an informal preannouncement strategy and aformal preannouncement strategy in markets defined byweak, moderate and strong agent preferences, respectively.The corresponding SPNE prices and licensing fees aregiven by

pi*{N,N} � 2t

li*{N,N} �t4

pi*{I,I} � 2t − 2µα − δβ

4− δ

[3µ + δ

]β2

2t

li*{I,I} �t + 2

[δ − µ

]β

4

pi*{F,F} � 2t − 2α − β

4− 2β2

t

li*{F,F} �t4. ■

Once again, note that even in the case of multihom-ing developers, the appropriateness of the pre-announcement strategy continues to be characterizedby the strength of preferences. The bounds for thestrategies are given by Equations (B10) and (B11) inOnline Appendix B. We find that our observation fromLemma 1 still holds for this case allowing us to cat-egorically conclude that preannouncement strategieshave a consistent impact on the marketplace inde-pendent of loyalty of installed bases and/or the op-portunity to multihome. From a theoretical per-spective, our findings encompass results from the twoextant pieces of literature that has examined the roleof information, albeit regarding prices and consumerknowledge (Suleymanova and Wey 2012, Hagiu andHałaburda 2014). By examining the strategic role ofinformation across several market settings, we canconclude that information increases the competitiveintensity leading to lower prices and profits. How-ever, firms end up having to engage in preannounce-ment strategies (informal/partial or formal/full) andthus providing more information on their forthcomingplatforms when they face agents that have strong tastepreferences. The key intuition behind these resultscontinues to be the fact that information that leads tobetter estimates of the strength of network effects of-fered by the new version of the platform results inhigher expected network effects benefits. Such expec-tation of network effects increases surplus for theagents but is nonrecoverable by firms.One element stands out in themultihoming case, that

is, unlike in all the earlier cases licensing fees underthe informal preannouncement strategy are weaklyincreasing in the cross-side strength of network effects;in the single-homing cases licensing fees are weakly


decreasing in this parameter or independent of it. Theintuition behind this result is that multihoming reducescompetitive intensity even if firms lose their monop-olistic hold over their developers. A platform obtainslicensing fees from a combination of both single-homingand multihoming developers, and the surplus of thesedevelopers are increasing in the cross-side strength ofnetwork effect. Unlike earlier (single-homing) when thestrictly monopolistic separation of developers intensi-fied competition, multihoming allows the platforms toenjoy the benefits of the installed base of developers andextract the surplus they may enjoy proportional to thecross-side strength of network effect.

Just like in our analysis of single-homingmarkets, wealso investigate multihoming markets where the in-stalled base of agents are not loyal and may, therefore,consider switching with the release of the new platformversion. As the reader can surmise, the utility functionsof the agents, in this case, will be developed along thelines of Section 3.1 (SH-SW) and the current discussionon multihoming. For brevity, we directly write theutilities of the users and both the single and multi-homing developers (for the informal preannouncementcase) as follows.

Uiu,I�θ+µα

[2niu

(pi,p−i,li,l−i

)]+µβ

[2[nid

(pi,p−i,li,l−i

)+ni,−id

(pi,p−i,li,l−i

)]]−pi

Uid,I�θ+δβ

[2niu

(pi,p−i,li,l−i

)]−li

Ui,−id,I �θ+δβ

[2[niu

(pi,p−i,li,l−i

)+n−i

u

(pi,p−i,li,l−i

)]]−li−l−i.

(23)

Further, the platform’s objective function can bewritten as

max{pi ,li}

πiI � pi

[2niu


)]+ li

[2[nid


)+ni,−id


)]]− kc. (24)

Following approaches similar to the previous cases,we seek to identify SPNE by solving for equilibriumprices and licensing fees. This gives us Proposition 2.

Proposition 2. When competing firms and their platformsare identical in all aspects, there is no SPNE in preannouncementstrategies in multihoming markets when the installed base ofagents can switch platforms.

As discussed in further detail in Online Appen-dix B, multihoming in a Hotelling model requirescertain fundamental assumptions to be satisfied inequilibrium. First, there are the standard conditionsof market coverage that is true for both single andmultihoming. Then, there is the specific requirement

that there is a nonzero number of multihoming de-velopers and the sum of developers is equal to the sizeof the market. In the earlier case where the installedbase of agents was loyal, these conditions were metin equilibrium. However, we find that when theseagents can switch the conditions for a multihomingequilibriumare notmet.Note that in otherworks suchas Rasch (2007), authors model contexts similar to ourswitching case and yet find equilibriummultihomingoutcomes. A careful analysis of these models willshow that their base setup incorporates some dif-ferentiation between either the two platforms in theform of strength of network effects engendered byeach and/or differences between the two sides of themarkets. In our model, we do not introduce thesedifferences to independently examine the impact ofthe preannouncement strategies.We can attribute the nonexistence of equilibrium to

this lack of differentiation. In other words, when theagents are not loyal and consider switching, they areprocurable when the new version of the platform isreleased. The option to multihome further removesthe possibility of the two platforms to differentiatethemselves as their only source of differentiation wasby their location on the Hotelling line. These condi-tions are not conducive to the existenceof apure-strategyequilibrium. In a similar vein, Choi (2010, p. 621) ob-serves, “Multi-homing has an unintended effect ofmaking consumers more homogeneous,” supportingour earlier assertion that multihoming was reducingthe one main source of differentiation in our model.

3.3. Welfare AnalysisWe conclude our examination of preannouncementstrategies with an understanding of its implication onconsumer and social welfare. In most platform papers,there is always a brief discussion on welfare, but theimportance of these analyses is furthermore significant inthe context of preannouncement. Since the 1960s, it hasbeen alleged that that product preannouncements canpotentially be (Gerlach 2004, p. 192) “predatory actsin high-technology industries.” With regards to therole of information itself, Hagiu and Hałaburda (2014)observe that all actors in a monopoly—platform,users and developers are all better off in the case ofinformed users. However, in a competitive model theysuggest (p. 30), “both users and developers prefer thescenario with uninformed users.” It, therefore, be-hooves us to examine welfare because informationin our model is part of firm strategy rather thanan exogenous condition. In this paper, not only dowe examine welfare in each of the different pre-announcement strategies but we also investigate howthese might be affected by outside market conditionssuch as when consumers are not loyal and when


developers have the option to multihome. Therefore,we have Propositions 3 and 4.

Proposition 3. Consumer (all agents—users and devel-opers) welfare is increasing in information while socialwelfare is decreasing, i.e., markets with Formal equilibriumoutcomes have the highest consumer welfare while marketswith No-Preannouncement equilibrium outcomes have thehighest social welfare.

Proposition 4. Across different types of markets, (i) con-sumer (social) welfare is higher (lower) in markets that allowfor switching than those that do not and (ii) both consumerand social welfare is higher when developers can multihomethan when they only develop for a single platform.

In our model, consumer welfare refers to the netwelfare of all agents (users and developers) after theplatform is releasedwhile social welfare also includesthe profits from the two platforms. Although firms setequilibrium prices in Stage 2 to fulfill consumer ex-pectations, all agents get full information regardingtheNEE aspects of the platformupon its release. Thus,in the SH-NSW case when firms made an informalpreannouncement in Stage 1, by substituting µ � 1 inEquation (5) we compute the utility realized by theuser after the release of the platform as

Uiu,I � θ + α

[mi

u + niu(pi*{I,I}, p

−i*{I,I}, l

i*{I,I}, l

−i*{I,I}

)]+ β

[mi

d

+ nid(pi*{I,I}, p

−i*{I,I}, l

i*{I,I}, l

−i*{I,I}

)]− pi*{I,I} .

Similarly, if firms made an informal preannouncement,by substituting δ � 1 in Equation (6), the developer’srealized utility post release of the platform is

Uid,I � θ + β

[mi

u + niu(pi*{I,I}, p

−i*{I,I}, l

i*{I,I}, l

−i*{I,I}

)]− li*{I,I}.

Along the same lines, we can compute the realizedutilities for different equilibrium strategies in SH-NSW, SH-SW, and MH-NSW scenarios.

For the single-homing case, the consumerwelfare isgiven by w{s,s} � wu,{s,s} + wd,{s,s} where

wu,{s,s} �∫ x

0Ui

u,sdx +∫ 1

xU−i

u,sdx;

wd,{s,s} �∫ y

0Ui

d,sdx +∫ 1

yU−i

d,sdx (25)

with s∈ {F, I,N} signifying preannouncement strat-egy, and x, y, respectively indicating the user anddeveloper indifferent between the two firms. There-fore, the net social welfare can be written as

W{s,s} � w{s,s} + πi{s,s} + π−i

{s,s}. (26)

In the case of multihoming along the lines of thedevelopment in Section 3.2, we need to account for themultihoming developers in both the developer wel-fare as well as the profit functions of the two firms.

Proposition 3 makes it clear that a full and formalpreannouncement, that is, one where themarketplacehas full knowledge of all the NEE aspects of the newversion of the platform is the best for all agents.Perhaps this is not surprising in that utilities are in-creasing in the strength of network effects and it isthrough formal preannouncement that the agents areenabled to accurately estimate the strength of net-work effects. What is interesting is that social welfareis the lowest in this case and is higher in the case of theno-preannouncement equilibrium. This makes it ev-ident that the losses suffered by the platform (as aresult of lower prices and licensing fees) in the case ofcomplete information in the marketplace is more thanoffset by the gains when the agents are making de-cisions without awareness of the new versions’ fea-tures. Therefore, it is not a strict transference in surplus.Also, note that this observation about the role of infor-mation is true in all markets, those with and withoutswitching and where multihoming is observed.Given our analysis of different markets, we can also

comment on the impact of the market structures them-selves. For example, is switching beneficial? Alterna-tively, is multihoming attractive to platform firms? Al-though beyond the scope of this paper, the literature ontechnology standards (Farrell and Saloner 1986) pointsout as to how decisions of compatibility can changemarket conditions, that is, switching and/or multi-homing themselves can be manipulated by technol-ogy firms. Ourfindings suggest that, not surprisingly,consumer welfare is higher when switching is pos-sible but in this case also it is not a question of simpletransference as bothfirmprofits and social welfare arehigher when switching is restricted. This once againsuggests that the gains made by firms more thanovercome the losses by consumers when switchingis disallowed thus further emphasizing on loyalty-engendering initiatives on the part of the firms.Proposition 4 also points to the fact that making one’snew platform version to be incompatible with theinstalled base of the other’s is attractive to firms.In the case of multihoming, our findings suggest

that everyone is better off—consumers as well asfirms. Multihoming is where at least some developersfind it optimal to develop for both platforms. Thismay be due to a variety of reasons including technicalcompatibility, low licensing fees on both sides and/orlimited differentiation between platforms. Whateverthe source of multihoming is, our results show that itis beneficial at large.

4. ConclusionThe analysis in this paper contributes to our under-standing of the role of information in preannounce-ment strategies in the context of two-sided markets(or platforms). Such insights can be leveraged by


managers and strategists of platforms in preannoun-cing their new versions and designing pricing strate-gies for agents (e.g., users and developers of a videogaming console).

In platforms that are very similar in their intrinsicbenefits, platform preannouncement could aid usersin forming estimates about the strength of networkeffects of their new versions. Fully informing themarket regarding NEE aspects of a platform is ex-pensive because it involves educating agents aboutthe experiential value derived from such elementsof the platform. One may think that such signifi-cant investments are necessary for console makersbecause they influence agents’ estimates regardingthe strength of network effects and therefore impactsfirm’s pricing strategies.

Our analysis suggests that while not providing anyinformation regarding NEE elements to the market ismost beneficial for the firms, there is a clear rela-tionship between firms’ choice of equilibrium pre-announcement strategies and differentiation of theplatforms from the agent’s perspective. When thecompeting platforms are less differentiated (whenagent preferences are weak), the firms will avoid anypreannouncement. However, when competing plat-forms are highly differentiated (when agent prefer-ences are strong), firms will commit to a formalpreannouncement strategy in SPNE. Partial pre-announcement is an SPNE strategy for both firmswhere the platforms are moderately differentiated.Thus, platform owners must understand the impactof market characteristics before identifying a pre-announcement strategy.

We extend our analysis to understand the impactof loyalty of the installed base of agents on pre-announcement strategies. In markets where the in-stalled base of agents can switch platforms, we findthat the ordering of SPNE preannouncement strate-gies over taste preferences remain the same, but it isclear that prices, licensing fees and firm profits are allhigher when agents cannot switch. This insight in-dicates that platform strategists should not discountthe impact of policies such as loyalty programs thatdiscourage the installed base from switching.

Further, investigating these strategies in marketsthat allow multihoming offers interesting insights.First, there is no equilibrium outcome in marketswhere both switching and multihoming are possibleunless the firms are nonidentical (e.g., in their stand-alone value). However, when agents do not switch ina multihoming case, we find that the extent of in-formation in preannouncement plays a similar role asin the single-homing setting.

We also explore the role of information in pre-announcement strategy, switching ability of the in-stalled base andmultihoming on consumer and social

welfare. Our results indicate that although consumerwelfare increases with more information in pre-announcement, the social welfare decreases. This isbecause the increased utility of agents from moreinformation is not offset by the firm’s profits due toincreased intensity of competition. Hence, a socialplannermay suggest a ban on preannouncements.Wefind that the ability of the installed base of agents toswitch and the presence of multihoming developersare both beneficial to consumer welfare. Unlike theswitching case,multihoming developers also increasesocial welfare irrespective of the extent of informationin the preannouncement strategy. Thus, platformmanagers in the video game industry have an in-centive to enable the developer side tomultihome andport their games easily across the platforms.

Appendix A. Platform Preannouncement Strategies:The Strategic Role of Information inTwo-Sided Markets Competition

Proof of Lemmas 1, 2, and 3To derive the optimal prices and licensing fees for whenfirms preannounce formally, we take the user demandfunction that was derived from the Hotelling line, given byEquation (13) and the developer demand function given byEquation (14) and substitute them in the profit functiongiven by Equation (9). This gives us

max{pi ,li}

πiF � pi

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣miu +

12+

−t [pi − p−i] − β

[li − l−i

]+ [

αt + β2][mi

u −m−iu

] + βt[mi

d −m−id

]2[t[t − α

] − β2]

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦+ li

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣mid +

12+

−β[pi − p−i] − [

t − α][li − l−i

]+ tβ

[mi

u −m−iu

] + β2[mi

d −m−id

]2[t[t − α

] − β2]

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦ − c.

(A.1)

To derive the response function for firm i, we differen-tiate Equation (A.1) with respect to piand li. Similarly, wecan construct the objective function of the other firmmax{p−i ,l−i}π−i

F , and differentiating with respect to p−iand l−i,we get its response function. We solve these four first-orderconditions simultaneously to get the prices and licensingfees as a function of the installed bases. Because our focusis on identical firms, we can get candidate equilibriumprices (pi{F,F}) and licensing fees (li{F,F}) for the focal firmby setting the installed bases of both firms to be thesame(mi

u � m−iu , mi

d � m−id ). Note, however, in order for these

prices and licensing fees to be SPNE, not only should thispreannouncement strategy dominate over others, but itshould also be feasible in the region as defined by thevarious parameter relationships. In addition, for the exis-tence of our equilibrium solutions to hold true post plat-form release, it is ensured that the platform i’s indifferentconsumer’s utility continues to be greater than or equal towhat hemay have derived from the other platform. In otherwords, to establish that no agent would change their


platform adoption decision post platform release, the fol-lowing condition is ensured to be true: Ui

u,F(realized) ≥U−i

u,F(realized) and Uid,F(realized) ≥U−i

d,F(realized). Note thatthis only needs to be checked for this indifferent agent asutilities are strictly increasing away from this user, for otherusers of the focal platform. Further, because firms aresymmetric, the position of the indifferent agent does notchange post platform release. In the case of Formal pre-announcement strategy,

Uiu,F(realized) � U−i

u,F(realized) � θ + α[1 + µ

] + β[1 + δ

] − 3t2

Uid,F(realized) � U−i

d,F(realized) � θ + β[1 + µ

] − 3t2.

Along the lines and as developed in Section 2.4, we cannow develop candidate equilibrium price-licensing pairsfor each of the other scenarios.3 Thus we will have pi{I,I}, l

i{I,I}

for the informal case (using Equations (15) and (10)), andpi{N,N}, li{N,N} (using Equations (16) and (11)) for the no-preannouncement case. Now in order for any of the pairsto be SPNE prices and licensing fees, we require the profitsfrom that strategy to be higher than the other two strategies.Therefore, formal strategy prices and licensing fees derivedearlier are equilibrium prices (pi*{F,F}) and licensing fees(li*{F,F}), if and only if

πi*{F,F} >max

[πi*{I,I},π

i*{N,N}

]π−i*{F,F} >max

[π−i*{I,I},π

−i*{N,N}

]. (A.2)

We can easily see that not only are these conditional onthe various constraints on parameters α, β,µ, δ, etc., but theyalso need to satisfy relationships amongst themselves asimposed by simple individual rationality conditions andother words. In other words, explicitly imposing all pa-rameters to be non-zero and keeping µ, δ∈ (0,1), we nowconsider other economic constraints and characterize allof them through t. Once again, we develop these only forthe formal case.

From the Hotelling model, we require that all agents derivenon-negative utilities (they will not participate otherwise). En-suring this for the indifferent agent suffices and thereforesubstituting the utility function (Equation (3)) and user de-mand (Equation (13)) in the net utility (Equation (12)), wecan derive the condition for non-negative net utility ofuser as

t ≤ 2[θ + 2α + 2β

]3

.

Along the same lines, using Equations (4) and (14), thecondition necessary for non-negative net utility of the de-veloper can be given as

t ≤ 2[θ + 2β

]3

.

We also require that profits are non-negative, whichyields us the condition

t ≥ c + α + 2β2

.

And that the profit function is concave in the equilibriumprices and licensing fees, which yields us conditions

t>α and t[t − α]> β2.

Finally, because we do not consider subsidies in ourpaper, all equilibrium prices and licensing fees must bepositive. This yields us the conditions

t>α + β, t> β.

Considering all these constraints together, through re-duction, we can characterize the potential bounds whereFF is SPNE to be defined as:

t{F,F} ∈(max

(α + β,

α + 2β + c2

),2θ3

+ 2β3

). (A.3)

Similarly, considering the utility functions, profit func-tions concavity conditions and others, we can have thepotential bounds for the other two cases as:

t{I,I} ∈(max

(µα + δβ,

µ[α + β

] + δβ + kc2

),2θ3

+[µ + δ

]β

3

)

t{N,N} ∈(0,2θ3

). (A.4)

However, note that in order for these bounds to be SPNEregions, we need to ensure that Equation (A.2) (and itscounterparts for other equilibrium solutions II and NN)) isalso satisfied alongwith Equations (A.3) and (A.4). Thus, theformal strategy equilibrium bounds characterized over t is

t*{F,F} ∈(2θ3

+[µ + δ

]β

3,2θ3

+ 2β3

). (A.5)

Similarly, we can derive the bounds for the other twoequilibrium outcomes as

t*{I,I} ∈(2θ5,2θ3

+[µ + δ

]β

3

)(A.6)

t*{N,N} ∈(0,2θ3

). (A.7)

Therefore, we have Lemmas 1, 2, and 3.

Endnotes1Even startups or small sized firms have to invest hundreds ofthousands of dollars at CES. See https://www.inc.com/ilya-pozin/what-it-costs-to-exhibit-at-ces.html (last accessed on December 25, 2019).2 See http://www.playstationlifestyle.net/2017/06/01/sony-marketing-budget-was-set-at-13-3-billion-yen-for-marketing-deals-and-advertising-partnerships-in-fy-2016/ (last accessed on December 25, 2019).3Note that to avoid redundancies, we explicitly derive and developprices and licensing fees and corresponding conditions only for theformal case. The techniques to do so are the same for other cases aswell.

ReferencesAdner, R, Chen J, Zhu F (2019) Frenemies in platform markets:

Heterogeneous profit foci as drivers of compatibility decisions.


https://www.inc.com/ilya-pozin/what-it-costs-to-exhibit-at-ces.html

https://www.inc.com/ilya-pozin/what-it-costs-to-exhibit-at-ces.html

http://www.playstationlifestyle.net/2017/06/01/sony-marketing-budget-was-set-at-13-3-billion-yen-for-marketing-deals-and-advertising-partnerships-in-fy-2016/



Management Sci., ePub ahead of print October 22, 2019, https://doi.org/10.1287/mnsc.2019.3327.

Anderson EG Jr, Parker GG, Tan B (2013) Platform performanceinvestment in the presence of network externalities. Inform.Systems Res. 25(1):152–172.

Armstrong M (2006) Competition in two-sided markets. RAND J.Econom. 37(3):668–691.

Armstrong M, Wright J (2007) Two-sided markets, competitive bot-tlenecks and exclusive contracts. Econom. Theory 32(2):353–380.

Athey S, Calvano E, Gans JS (2016) The impact of consumer multi-homing on advertising markets and media competition. Man-agement Sci. 64(4):1574–1590.

Bayus BL, Jain S, Rao AG (2001) Truth or consequences: An analysisof vaporware and new product announcements. J. Marketing Res.38(1):3–13.

Boudreau K (2010) Open platform strategies and innovation:Granting access vs. devolving control. Management Sci. 56(10):1849–1872.

Caillaud B, Jullien B (2003) Chicken & egg: Competition among in-termediation service providers. RAND J. Econom. 34(2):309–328.

Charla C (2016) A letter from Chris Charla: ID@Xbox Updates andGDC. Accessed February 24, 2020, https://news.xbox.com/en-us/2016/2003/2014/letter-chris-charla-idxbox-updates-gdc/.

Choi JP (2010) Tying in two-sided markets with multi-homing.J. Indust. Econom. 58(3):607–626.

Choi JP, Kristiansen EG,Nahm J (2005)An economic analysis of productpre-announcements. CESifo Econom. Stud. 51(2–3):299–319.

Draganska M, Klapper D (2007) Retail environment and manufac-turer competitive intensity. J. Retailing 83(2):183–198.

Dranove D, Gandal N (2003) The DVD-vs.-DIVX standard war:Empirical evidence of network effects and preannouncementeffects. J. Econom. Management Strategy 12(3):363–386.

Economides N, Katsamakas E (2006) Two-sided competition of pro-prietary vs. open source technology platforms and the implica-tions for the software industry. Management Sci. 52(7):1057–1071.

Eisenmann T, Parker G, Van Alstyne MW (2006) Strategies for two-sided markets. Harvard Bus. Rev. 84(10):92.

Eisenmann TR (2006) Platform-mediated networks: Definitions andcore concepts. (Revised October 2007.) Harvard Business SchoolModule Note 807-049, Harvard Business School, Boston.

Eliashberg J, Robertson TS (1988) New product preannouncing be-havior: Amarket signaling study. J. Marketing Res. 25(3):282–292.

Farrell J, Saloner G (1986) Installed base and compatibility: Innova-tion, product preannouncements, and predation. Amer. Econom.Rev. 76(5):940–955.

Gabszewicz JJ,Wauthy XY (2014) Vertical product differentiation andtwo-sided markets. Econom. Lett. 123(1):58–61.

Gallaway B (2013) Confusion ruled the day at Microsoft’s Xbox Onereveal event. WIRED (May 22), https://www.wired.com/2013/2005/xbox-one-hands-on/.

Gerlach HA (2004) Announcement, entry, and preemption whenconsumers have switching costs.RAND J. Econom. 35(1):184–202.

Hagiu A (2009) Two-sided platforms: Product variety and pricingstructures. J. Econom. Management Strategy 18(4):1011–1043.

Hagiu A, Hałaburda H (2014) Information and two-sided platformprofits. Internat. J. Indust. Organ. 34:25–35.

Hagiu A, Lee RS (2011) Exclusivity and control. J. Econom. Manage-ment Strategy 20(3):679–708.

Hagiu A, Spulber D (2013) First-party content and coordination intwo-sided markets. Management Sci. 59(4):933–949.

Hall C (2015) PS4 and Xbox One install base 60 percent larger thanlast generation, NPD says. Accessed February 24, 2020, https://www.polygon.com/2015/2/13/8033983/ps4-and-xbox-one-install-60-percent-higher-npd-jan-2015.

Katz ML, Shapiro C (1985) Network externalities, competition, andcompatibility. Amer. Econom. Rev. 75(3):424–440.

Le Nagard-Assayag E, Manceau D (2001) Modeling the impact ofproduct preannouncements in the context of indirect networkexternalities. Internat. J. Res. Marketing 18(3):203–219.

Lee Y, O’Connor GC (2003) New product launch strategy for networkeffects products. J. Acad. Marketing Sci. 31(3):241–255.

Mulkerin T (2016) 13 video game consoles you’ve probably neverheard of. Bus. Insider (August 8), https://www.businessinsider.com/14-video-game-consoles-youve-probably-never-heard-of-2016-8.

Narayanan S, Manchanda P, Chintagunta PK (2005) Temporal dif-ferences in the role of marketing communication in new productcategories. J. Marketing Res. 42(3):278–290.

Niculescu MF, Wu DJ (2014) Economics of free under perpetual li-censing: Implications for the software industry. Inform. SystemsRes. 25(1):173–199.

Ofek E, Turut O (2013) Vaporware, Suddenware, and Trueware: Newproduct preannouncements under market uncertainty. Market-ing Sci. 32(2):342–355.

Parker GG, Van Alstyne MW (2005) Two-sided network effects:A theory of information product design.Management Sci. 51(10):1494–1504.

Polygon staff (2013) Xbox One: What we know. Accessed February24, 2020, https://www.polygon.com/features/2013/5/21/4350930/xbox-one-what-we-know.

Poulter S (2013) It’s game on in Christmas console war: Xbox Oneand PlayStation 4 to go head to head with November releasedates. Daily Mail (September 4), http://www.dailymail.co.uk/sciencetech/article-2411912/XBox-One-Playstation-2411914-November-release-dates-mean-Christmas-console-war.html.

Rasch A (2007) Platform competition with partial multihoming underdifferentiation: A note. Econom. Bull. 12(7):1–8.

Rochet J-C, Tirole J (2003) Platform competition in two-sidedmarkets.J. Eur. Econom. Assoc. 1(4):990–1029.

Schneider J (2015) 10 tactics for launching a product using so-cial media. Harvard Bus. Rev. (April 16), https://hbr.org/2015/2004/2010-tactics-for-launching-a-product-using-social-media.

Sterlicchi J (2007) Nintendo’s Wii console captures new gamemarket.The Guardian (October 10), https://www.theguardian.com/business/2007/oct/2010/usnews.internationalnews.

Su M, Rao VR (2010) New product preannouncement as a signalingstrategy: An audience-specific review and analysis. J. ProductionInnovation Management 27(5):658–672.

Suleymanova I, Wey C (2012) On the role of consumer expectations inmarkets with network effects. J. Econom. 105(2):101–127.

Yang J, Mai ES (2010) Experiential goods with network externalitieseffects: An empirical study of online rating system. J. Bus. Res.63(9-10):1050–1057.

Zhu F, Iansiti M (2012) Entry into platform-based markets. StrategicManagement J. 33(1):88–106.




https://news.xbox.com/en-us/2016/2003/2014/letter-chris-charla-idxbox-updates-gdc/

https://news.xbox.com/en-us/2016/2003/2014/letter-chris-charla-idxbox-updates-gdc/

https://www.wired.com/2013/2005/xbox-one-hands-on/

https://www.wired.com/2013/2005/xbox-one-hands-on/

https://www.polygon.com/2015/2/13/8033983/ps4-and-xbox-one-install-60-percent-higher-npd-jan-2015



https://www.businessinsider.com/14-video-game-consoles-youve-probably-never-heard-of-2016-8



https://www.polygon.com/features/2013/5/21/4350930/xbox-one-what-we-know

https://www.polygon.com/features/2013/5/21/4350930/xbox-one-what-we-know

http://www.dailymail.co.uk/sciencetech/article-2411912/XBox-One-Playstation-2411914-November-release-dates-mean-Christmas-console-war.html



https://hbr.org/2015/2004/2010-tactics-for-launching-a-product-using-social-media



https://www.theguardian.com/business/2007/oct/2010/usnews.internationalnews

https://www.theguardian.com/business/2007/oct/2010/usnews.internationalnews

Platform Preannouncement Strategies: The Strategic Role of ...

Documents

Transcript of Platform Preannouncement Strategies: The Strategic Role of ...