Long Tails Vs. Superstars: the Effect of Information Technology on Product Variety and Sales Concentration Patterns

Home , Long tail

Research Commentary Long Tails vs. Superstars: The Effect of Information Technology on Product Variety and Sales Concentration Patterns

Erik Brynjolfsson Sloan School of Management, Massachusetts Institute of Technology, Cambridge, Massachusetts 02142, [email protected] Yu (Jeffrey) Hu Krannert School of Management, Purdue University, West Lafayette, Indiana 47907, [email protected] Michael D. Smith Heinz School of Public Policy and Management, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213, [email protected]

he Internet and related information technologies are transforming the distribution of product sales across Tproducts, and these effects are likely to grow in coming years. Both the Long Tail and the Superstar effect are manifestations of these changes, yet researchers lack consistent metrics or models for integrating and extending their insights and predictions. In this paper, we begin with a taxonomy of the technological and nontechnological drivers of both Long Tails and Superstars and then define and compare the key metrics for analyzing these phenomena. The core of the paper describes a large and promising set of questions forming a research agenda. Important opportunities exist for understanding future changes in sales concentration patterns; the impact on supply chains (including cross-channel competition, competition within the Internet channel, implications for the growth of firms, and the balance of power within the supply chain); implications for pricing, promotion, and product design; and, ultimately, the potential effects on society in general. Our approach provides an introduction to some of the relevant research findings and allows us to identify opportunities for cross-pollination of methods and insights from related research topics. Key words: Long Tail; Superstar; product variety; sales concentration; information technology History: Vallabh Sambamurthy, Senior Editor. This paper was received on July 6, 2010, and was with the authors 20 days for 1 revision. Published online in Articles in Advance November 18, 2010.

1. Introduction products and services and the concomitant increase Improvements in information technology (IT) are in product variety available to consumers (Bresnahan transforming the way consumers learn about goods and Gordon 1997, Brynjolfsson et al. 2003). In recent and services, as well as the way producers develop, years, IT has played a central role in this increase in distribute, and deliver them. These technological the supply and demand of niche products, creating 1 advances are not simply the result of radical increases a “Long Tail” in the distribution of product sales. in the capacity of digital communications, computing At the same time, many markets can be increas- and storage but also represent a qualitative transfor- ingly described as “Superstar” or “winner-take-all” 2 mation in search, recommendation tools, and social markets where blockbuster products dominate sales. network technologies. 1 See, for example, Brynjolfsson et al. (2003), Anderson (2004), Bryn- The implications of these changes for business and jolfsson et al. (2006, 2007), Cachon et al. (2008), Tucker and Zhang society are large and will likely grow as technology (2009), and Oestreich-Singer and Sundararajan (2009). advances. For instance, signiﬁcant increases in societal 2 See, for example, Rosen (1981), Noe and Parker (2005), and Fleder welfare have come from the introduction of new and Hosanagar (2009).

Such changes in production and consumption pat- such as print and broadcast media advertising and in- terns across niche and blockbuster products portend store displays toward broader Internet search channels profound effects on competition and market structure. with greater targeting and broader product selection. In this paper, we argue that although both the As a result, an estimated 57% of all online shoppers Long Tail and the Superstar3 literatures have largely use a search engine to begin their online shopping been separate, they can and should be analyzed as research (Internet Retailer Magazine 2010). Moreover, part of an integrated research agenda that studies by enabling consumers to specify their desired prod- shifts in product variety and concentration patterns uct features, prices, characteristics, or locations, Inter- caused by information technology. Each of these lit- net search engines can influence the importance of eratures has made important contributions, and in brands, and potentially increase the relative share of our view, they are connected by common fundamen- niche products that cater to specific needs. Similarly, tal drivers, methods, and metrics. We first discuss search tools at sites like Google Products, eBay, or some of the common technology drivers and describe shopping aggregators like PriceGrabber make it pos- a unified measurement framework. We then go into sible for consumers to find obscure products with rel- more detail on a set of open research questions that atively little effort. On the whole, search engines are can form an agenda for the next decade. Ultimately, becoming more sophisticated in the way they handle a better understanding of how IT can affect product queries, which may portend the ability to match con- sales concentration may not only provide guidance to sumers even more precisely to products and services entrepreneurs, managers, and policymakers but also with specific or even unique attributes. could provide a lens for analyzing related questions Personalization Technologies: Personalization and such as the balkanization of ideas and innovations. recommendation technologies go one step beyond search, to predict what consumers may be interested 2. Drivers of Product Variety in based on observed actions and to display these pre- and Concentration dictions to consumers in the course of their shopping. These technologies can take various forms, from sim- As discussed below, the drivers of these changes in ple “top 10” lists of recommended items other users product variety and concentration can be divided have selected to sophisticated collaborative filters that along two axes: technological and nontechnological infer preferences with little or no active effort by the drivers, and demand- and supply-side effects. consumer. Engines that disproportionately help con- 2.1. Technological Drivers sumers find obscure products that they would not have otherwise known about should reduce product 2.1.1. Demand-Side Drivers. There are at least three important demand-side technological drivers of sales concentration. However, it is also possible that changes in product variety and product sales concen- these tools could be tuned so that they lead to dis- tration: (1) changes in search and database technolo- proportionate gains in sales of popular products rel- gies, (2) changes in personalization technologies, and ative to niche products (Fleder and Hosanagar 2009). (3) changes in online community and social network The net effect on product concentration will depend technologies. not only on the recommendations of the engine, but Search and Database Technologies: Search tech- also on the information sources and decision-making nologies like Google have transformed how con- process on which consumers would rely on in the sumers find the products they buy. Consumers’ absence of a recommender tool. The ultimate outcome information sources are shifting away from channels is unlikely to be purely a function of technological forces but may also depend on how tool develop- ers and the sites that pay them choose to drive pur- 3 Although terms “Long Tail” and “Superstar” can be used in var- chase considerations. In turn, this may depend on ious contexts, in this paper, we capitalize the terms Long Tail and Superstar to refer specifically to the contrasting phenomena the nature of product market competition and other describing changes in product variety and concentration patterns. factors. Brynjolfsson et al.: Research Commentary: The Effect of IT on Product Variety and Sales Concentration Patterns 738 Information Systems Research 21(4), pp. 736–747, © 2010 INFORMS

Online Communities and Social Networks: Perhaps 2.2. Nontechnological Drivers the most basic function of digital technologies has In addition to these technological drivers, research been to connect people at reduced communication must also consider important nontechnological factors costs, and an example of this can be seen in the influencing changes in consumption. For instance, explosion of dedicated online communities and social as early as 1776, Adam Smith noted, “specializa- networking sites over the past decade. Although the tion is limited by the extent of the market,” sug- monetization of these sites is low relative to search, gesting that larger markets—such as those supported the potential to influence purchases is significant. For by the Internet—could support more niche produc- instance, to the extent that demand-side economies ers. Conversely, starting with Rosen (1981) and Frank of scale are important and consumers communicate and Cook (1995), a number of papers in the Super- as part of a single interconnected network, these star literature have advanced supply-side explana- tions for why companies may prefer to produce a few technologies may increase the importance of fads, Superstar products: economies of scale may allow blockbusters, and the Superstar effect. However, if Superstar products to have higher profitability than people use these tools to self-segregate into smaller niche products do; Superstar products may command groups, or to learn about and develop idiosyncratic price premiums either because of increasing returns tastes, the net effect of online communities and social to quality or because of a highly skewed distribu- networks might be increased balkanization, favoring tion of talent, which in turn leads to highly skewed niche products. distribution of quality. Likewise, increasing returns to 2.1.2. Supply-Side Drivers. On the supply side, advertising expenditures can increase the profitability IT-enabled markets may increase producers’ incen- of Superstar products (Noe and Parker 2005). tives to create niche products while also increasing Papers in the Superstar literature have also pro- retailers’ incentives to stock these products. Consider- vided a set of nontechnological demand-side expla- nations for why consumers would prefer to consume ing retailers’ incentives, technology changes the cost Superstar products: consumers may want to have of stocking products. To make one additional product social interactions with other consumers; they may available to all the consumers in the United States (or use product sales as a signal of (otherwise difficult to in the world), Internet retailers only need to add the obtain) product quality or they may limit their choice inventory of that product to their centralized ware- set to economize on cognitive costs. houses, which in the case of digital products may sim- After reviewing the various drivers of product con- ply be a central server and product database. Even centration, it is clear that theory alone cannot predict for physical goods, these costs can be minimized with an inevitable trend in product concentration. Instead, technology. For example, by forming drop-shipping these are primarily empirical questions requiring care- agreements with book wholesalers, Amazon lowers ful research and models tailored to identifying and the cost of stocking an additional book. In addition, a quantifying the specific metrics and drivers of inter- variety of IT systems have reduced the costs to create est. In the following sections we will review relevant niche products. empirical measures of these phenomena and describe Finally, these technological drivers may create a set a program of research for the IS community to pursue of secondary supply-side effects: by creating nation- in addressing these questions. ally and globally interconnected markets as opposed to local markets, technology may create incentives for 3. Measures Matter retailers to disproportionately pursue Superstar prod- Given that technological and nontechnological fac- ucts. Conversely, the ability of consumers to locate tors can drive sales toward either the head or tail products that otherwise would not have been stocked of the sales distribution, the next logical question is: in physical stores may allow producers to pursue dif- “How should the research community measure these ferent projects, projects that apart from technology shifts?” A review of the literature suggests that the would not have been profitable. Long Tail can be defined and measured in at least Brynjolfsson et al.: Research Commentary: The Effect of IT on Product Variety and Sales Concentration Patterns Information Systems Research 21(4), pp. 736–747, © 2010 INFORMS 739 three different ways, which can lead to confusion apply a relative metric that is scale invariant and thus because these metrics can lead to seemingly contra- can be used to compare different markets. dictory outcomes: However, the relative metric will not always give • The Absolute Long Tail measures changes in the the same results as the absolute metric, even when absolute number of products sold, typically making applied to the same situation. For instance, suppose a comparisons across channels, time periods, or cate- retailer introduces a large number of new niche prod- gories. The measure of sales above an absolute cutoff ucts, each of which has very low sales. By the absolute of 100,000 titles, the typical size of large bricks-and- metric, the Long Tail has grown longer: consumers mortar bookstore, is an example of an absolute mea- have more choice, and on an absolute basis the num- sure used in Brynjolfsson et al. (2003). ber of sales above a certain rank cutoff may increase • The Relative Long Tail focuses on the relative even if the sales of any individual product is relatively share of sales above or below a certain rank percentile. low. However, it’s quite possible that sales will now be more concentrated when viewed by the relative met- The Gini coefficient (e.g., Brynjolfsson et al. 2007) is ric: the top X% of products SKUs may well account an example of this metric, and the classic Pareto prin- for a larger share of overall sales than before, which ciple that the top 20% products often generate 80% of seems to suggest that the Long Tail is now less impor- sales is an application of it. tant. Figure 1 provides a simple illustration of this. • Finally, because the ordinal rank to cardinal sales Fortunately, the relative metric is consistent with the relationship often follows a power law distribution, absolute metric when the number of products does the exponent (i.e., the slope of the log-linear relationship) not change. In this case, a Lorenz curve, or Gini coef- provides an indication of the relative importance of ficient can be used to illustrate the general importance the head versus the tail of the sales distribution. of the tail, or various rank percentiles (e.g., 20%) can Each of these metrics has strengths and weaknesses, be compared. and they are not interchangeable. A weakness of The third frequently used metric is the slope of the absolute metric is that it is not always intuitive the log-linear relationship between rank and sales. to apply it across different markets. The Long Tail This exploits the fact that many sales distributions phenomenon might be important for understanding can be characterized by a power law (see, for exam- changes in competition among camera retailers, even ple, Huberman 2001 with respect to website visits and though they have orders of magnitude fewer SKUs other online metrics). Of course, this metric is not par- than, say, booksellers do. This makes it appealing to ticularly useful when this is not the case. For instance,

Figure 1 Increasing the Number of Available Products Can Increase Sales Concentration

Sales Sales Case I Case II

Rank Rank Notes. Case I: 100 products are available and the top 50% of products account for 75% of total sales. Case II: Add a “longer tail” of 100 niche products with minimal sales, while leaving the sales of existing products unchanged. Now 200 products are available, and the top 50% of products account for 95% of total sales. Brynjolfsson et al.: Research Commentary: The Effect of IT on Product Variety and Sales Concentration Patterns 740 Information Systems Research 21(4), pp. 736–747, © 2010 INFORMS

Brynjolfsson et al. (2009) provide evidence that the Following the drivers and metrics discussed above, sales-rank slope is not constant, but changes over dif- we have identified four major areas of inquiry for ferent parts of the distribution. future research in the Long Tail: the nature of the Each of these metrics has its uses, but as with Long Tail phenomenon, its impact on supply chains, many academic fields, progress can be hampered its impact on pricing and other marketing strategies, when different researchers use different and poten- and its impact on society. tially conflicting metrics. Reference to the above tax- 4.1. The Existence of Long Tail and Superstar onomy can help assure that researchers are speaking Outcomes in Different Environments the same language when they contribute new the- As noted above, a particularly salient research ques- ory or empirical results. Emphasizing this point, we tion surrounding increasing product variety is note that the papers concluding that the Long Tail is whether this is an important long-term shift or a pass- important or is growing have typically used an abso- ing fad. The mere fact that Internet markets can sup- lute metric—sales of products above a certain sales ply more product variety than traditional markets can rank4—to measure its size, whereas papers finding does not necessarily mean that they will. It is possi- reduced importance of the Long Tail are more often ble that consumers like to consume the same products based on the concentration of sales or sales above that their friends are consuming and, as argued in the a certain proportion of products. There may not be literature, that this characteristic will lead to Superstar a single best measure, and it is quite possible that effects in some product markets. different measures should be used in different set- In this regard, it would be useful for the liter- tings. For example, the absolute measure adopted by ature to measure how the Long Tail varies across Brynjolfsson et al. (2003, 2009) may be more appro- product categories. Given differences in tastes and priate for measuring the consumer surplus gain from consumption patterns across product categories, it is the Long Tail, whereas the relative measure adopted possible that Long Tail markets will be more likely by Elberse and Oberholzer-Gee (2008) and Chellappa to develop in some product markets than in others. et al. (2007) may be more appropriate for analyzing For example, Elberse (2008) argues that music and producer incentives for targeting a particular area of movies are more likely to be enjoyed in a social con- the distribution curve. text and thus have natural characteristics that will lead to “winner-take-all” outcomes. Is this true, and 4. Toward a Research Agenda for if so, what other product categories will experience Understanding the Long Tail and similar outcomes? Much of the current research stud- Superstar Effects ies information goods (e.g., books, music, movies); As discussed in §§2 and 3, both the Long Tail and thus, future research could look into noninformation Superstar effects can be seen as the two sides of a categories. common set of questions about product variety and Similarly, Brynjolfsson et al. (2006) predicted that sales concentration. That said, the research literature many Long Tail markets would exhibit important sec- on the Long Tail is in many ways more developed and ondary effects that would lengthen the tail over time. provides a useful lens for highlighting key research It will be important for future research to analyze the changes in the Long Tail over time (e.g., Brynjolfsson opportunities. Fortunately, most of our research sug- et al. 2009). gestions can also be applied to questions about the As noted above, sites like Amazon include recom- Superstar effect, or more generally, an integrated view mendation engines that help consumers find prod- of both phenomena. ucts in the Long Tail, as well as prominent links to lists of top sellers in each product category, which 4 Note that a metric like sales of the bottom 100 products (e.g., Tan tend to increase winner-take-all effects. What’s more, and Netessine 2009) is not the same as the absolute metric such as sales beyond the top 100 products. Whereas the latter metric reflects it is possible that both types of tools are effective changes in product variety, the former metric cannot be used in at the same time, which in principle could increase this way. the relative size of both the head and the tail of the Brynjolfsson et al.: Research Commentary: The Effect of IT on Product Variety and Sales Concentration Patterns Information Systems Research 21(4), pp. 736–747, © 2010 INFORMS 741 sales distribution at the expense of the middle. Finally, 4.2.1. The Impact of the Long Tail on Cross- both producers and retailers may have self-interested Channel Competition. Internet retailers typically motivations to drive sales toward the head or the tail offer a large selection of niche products and pro- of the sales distribution and may act on these moti- vide IT-enabled search and recommendation tools vations by tuning their recommendation and search to help consumers discover niche products, whereas engines to ensure these outcomes. In light of the con- brick-and-mortar retailers typically only stock pop- trasting effects discussed above, it would be interest- ular products and do not provide IT-enabled tools. ing for future research to study the effect of each tech- Thus, an Internet retailer may be able to gain an nology on consumers’ shopping behaviors in terms of advantage over its brick-and-mortar counterparts by how consumers search for information, evaluate alter- pursuing a Long Tail strategy. natives, and make purchase decisions. How will these We are not aware of any paper that has made this technologies affect the concentration pattern of con- point through the use of an analytical model. How- sumers’ purchases, in isolation and in tandem? How ever, Brynjolfsson et al. (2009) represent an empirical will the relative importance of tools and technologies approach to this question, finding that as the num- like these vary across product categories, consumers, ber of local brick-and-mortar stores increases, Internet and over time, as the tools evolve? consumers’ demand for popular products drops significantly, but their demand for niche products does 4.2. The Impact of the Long Tail on Supply not change much. However, they have only studied Chains one product category—clothing—where products do How should supply chain participants respond to the not have unique identifiers. Future research is needed opportunities and challenges afforded by increased in other high-SKU product categories such as books, product variety available in Internet markets? At the music, and DVDs. Competition between Internet and producer level, Long Tail markets afford producers an brick-and-mortar retailers could be stronger in these opportunity to sell products online that would not be categories, because unique identifiers should make it commercially viable in traditional brick-and-mortar easier to identify an exact offline substitute for an markets (Brynjolfsson et al. 2006). Should producers online product. Technology may even increase com- focus their efforts on niche products with a small petition here because a variety of mobile applications number of sales, or should they continue to focus on (e.g., RedLaser on Apple’s iOS platform) allow con- hit products with large sales (but with ex ante risky sumers to scan bar codes in brick-and-mortar stores outcomes)? and compare prices with online retailers. Given the At the retailer level, changes in Long Tail markets central role of such cross-channel competition and can affect competition. The economic literature shows the growing availability of data, there are enormous that consumers benefit from the introduction of new opportunities for further research. products and new product varieties (Hausman 1981). Thus, consumers will naturally be attracted to com- 4.2.2. The Impact of the Long Tail on Compe- panies that offer a large selection of niche products tition Among Internet Retailers. Researchers could and provide IT-enabled tools that help consumers dis- also study how Internet retailers could use the avail- cover these niche products. There is some anecdo- ability of niche products as a strategic tool in its com- tal evidence that firms are consciously pursuing such petition with other Internet retailers. Some Internet a Long Tail strategy in order to gain a significant retailers like Amazon and Zappos offer a huge selec- competitive advantage over their rivals and to grow tion of niche products, whereas others provide a more their market shares (Mendelson and Meza 2001). All targeted selection. To what extent can the availability told, firm strategies associated with Long Tail mar- of niche products drive consumers’ choice of which kets may impact competition and strategy in four Internet retailer to buy from? To what extent can the main areas: cross-channel competition, competition availability of niche products help an Internet retailer between Internet retailers, firm growth, and power to gain a competitive advantage? What are the strate- within the supply chain. gic, technological cost or other drivers that lead firms Brynjolfsson et al.: Research Commentary: The Effect of IT on Product Variety and Sales Concentration Patterns 742 Information Systems Research 21(4), pp. 736–747, © 2010 INFORMS to choose different strategies with respect to their little power over their creative product. It is possible Long Tail approach? to view this structure as arising in part from the abil- Furthermore, low search and switching costs on ity of the record labels to control access to a small the Internet may intensify price competition among number of promotion (e.g., radio) and distribution Internet retailers. It would be interesting for future (e.g., brick-and-mortar record store) channels (see, for research to study whether an Internet retailer can example, Meier 2000). Does the Long Tail reduce adopt a Long Tail strategy to move consumers from the scarcity of access to distribution and promotional shopping solely on price toward looking for niche channels and thus weaken the control publishers have products that fit their tastes well. Such a strategy had over these important resources or will publish- could lead to higher levels of differentiation among ers continue to be able to exercise strong control over Internet retailers and, in turn, potentially higher most creative artists, as they have historically? profitability. 4.3. The Impact of the Long Tail on Pricing and 4.2.3. The Long Tail and the Growth of Firms. Other Marketing Strategies Long Tail strategies may also affect the growth of How might niche products made available through firms. Even before the term “The Long Tail” was the Internet channel (i.e., place, one of marketing’s coined by Anderson (2004), a number of Internet com- “4 Ps”) impact the other “4 Ps” of marketing practice: panies had successfully pursued management strate- price, promotion, and product? gies seemingly based on a Long Tail concept. For 4.3.1. Pricing. Pricing in the context of the Long instance, early in the development of Internet com- Tail presents a variety of fertile areas for future merce, Jeff Bezos, the founder and CEO of Amazon, research. First, it is not obvious how companies observed that there were more than three million should price niche products. Should they charge book titles in print worldwide, whereas the largest a price premium for niche products that fit con- physical Superstores carry only about 175,000 titles sumer tastes well, or should they cut the price for (Mendelson and Meza 2001). Recognizing the inher- niche products to stimulate consumers’ exploration of ent advantages of Internet retailers in providing products that may have lower qualities and higher greater product variety, he decided to choose books uncertainties? as its first product category when he founded Ama- Likewise, should companies adopt uniform pric- zon in July 1995, and the company provided a ing within product categories? Without conducting large selection of books online, along with tools that in-depth research, Apple initially adopted a uniform enabled consumers to search for books. Likewise, pricing strategy for the music in its iTunes store. More Steve Barnhart, the CEO of Orbitz, commented that recently, Apple, along with other music retailers, has niche products present a great growth opportunity for shifted toward a more flexible tiered pricing scheme. travel companies (Joyce 2008). Under this new scheme, most songs will still sell for It is reasonable to expect that the impact of the 99 cents; a certain number of new hit songs will sell Long Tail phenomenon on managerial practices and for $1.29; and many older songs will go for 69 cents. firm growth will only increase as Internet commerce A researcher could utilize data from such a natu- accounts for a larger proportion of total consumer ral experiment to understand how consumer demand commerce in the future. responds to price changes and how this effect is dif- 4.2.4. The Long Tail and the Balance of Power ferent across popular products and niche products. Within the Supply Chain. Finally, researchers could Such an analysis could shed light on whether niche analyze whether Long Tail distribution and promo- songs should be sold at a higher or lower price than tion channels will impact the balance of power within hit songs and whether a tiered pricing scheme has the supply chain. For example, consider the context led to higher or lower profits for music retailers and of music where historically a concentrated set of large music labels. music labels have controlled the industry, leaving Moreover, the answers to these and related ques- artists (with the exception of a few Superstars) with tions will likely vary across product categories, Brynjolfsson et al.: Research Commentary: The Effect of IT on Product Variety and Sales Concentration Patterns Information Systems Research 21(4), pp. 736–747, © 2010 INFORMS 743 because consumers likely place high values on niche channel to connect to consumers. Given that tra- products in categories they need while placing low ditional media are often saturated by promotions values on niche products in categories that they may for popular products, it is possible that promotion not necessarily need. For instance, online booksellers via social media can skew sales toward niche prod- typically offer bestsellers at substantial discounts, ucts, driving the Long Tail phenomenon. On the whereas more obscure books are often sold at smaller other hand, the effect of promotions via social media or even zero discounts. depends critically on the number of consumers who Second, the above scenario assumes that consump- are connected to the company and its products. tion of product A is independent from product B. Because more consumers are likely to be connected to However, in reality the demand functions for differ- Superstar products than to niche products via social ent products could be correlated. A consumer may media, it is also possible that social media will expand initially want to buy a popular product, but could the gap between Superstar products and niche prod- end up buying a niche product to go with the pop- ucts. More research is needed to understand how ular product, either because of the retailer’s recom- social media will drive these competing effects.5 mendation or because of the retailer’s free shipping 4.3.3. Product. Finally, product design itself might hurdle. The reverse can also happen: a consumer be affected by increased product variety. For instance, may be attracted to a retailer because they offer incentives for the creation of new niche products will niche products and may end up buying a popular be increased if these products have successful chan- product in addition to the niche products. In other nels for distribution and sales. This may cause exist- words, the demand for niche products could have a ing producers to shift production toward Long Tail spillover effect on the demand for popular products, products or may cause new producers to pursue these and vice versa. An executive at Amazon told us that markets. even when obscure products are not profitable stand- Another product design question involves bundling ing alone, the company benefits from keeping them or unbundling products, especially information goods in stock so that customers are confident of finding (see, e.g., Bakos and Brynjolfsson 1999, 2000). This what they look for at Amazon. This increases cus- question may be particularly salient in the context tomer loyalty and reduces incentives for visiting com- of music, where traditionally an album was built peting retailers. Given this strategic consideration and around two or three hit songs, with other songs filling the ones outlined above, it is important to know how the remaining space on the album. As such, albums Internet retailers should price their popular and niche can be viewed as a bundled good, whereas the shift products in the presence of spillover effects. toward digital markets for an increased variety of sin- gles is driving the industry toward unbundled con- 4.3.2. Promotion. As with pricing, there are a tent. Massive bundling or unbundling may be partic- variety of ways Long Tail markets could impact ularly dramatic for services like Rhapsody that allow promotion. Historically, companies needed to access subscribing consumers to listen to all the songs in scarce and expensive promotional channels such as the catalog, or services like eMusic that allow sub- radio and television to reach an audience. Today scribers to download a certain number of songs after companies can connect to potential consumers via paying a monthly fee. These subscription plans can a variety of social media channels, such as MySpace, be viewed as an aggregation mechanism, similar to Facebook, and Twitter. In such environments, produc- bundling. Finally, even in cases where physical prod- ers can broadcast information on its products to a ucts are being sold, the practice of offering free ship- set of consumers who are connected to the company. ping discounts on large orders would almost certainly Anecdotal evidence suggests that these sorts of promotions in social media can significantly impact prod- 5 Whereas we suggest a study of how firms’ promotional activ- uct sales (e.g., Vincent 2007). ities in social media affect concentration patterns, Dewan and In this regard, social media has clearly given niche Ramaprasad (2007) investigate how users’ activities in social media products, and producers of those products, a new affect concentration patterns. Brynjolfsson et al.: Research Commentary: The Effect of IT on Product Variety and Sales Concentration Patterns 744 Information Systems Research 21(4), pp. 736–747, © 2010 INFORMS encourage consumers to bundle several items into one is that the increased information available online will order, a form of nonlinear pricing. allow consumers to become better informed than they In this context, the existing literature on Long Tails could before. For example, consumers interested in a and Superstars has mostly studied standalone prod- particular bill before the U.S. Congress can browse ucts. It would be interesting to study how such the details of that bill on government websites and bundling and subscription practices can shift the bal- explore commentary on both sides of the issues in ance of power toward either Long Tail or Super- a much more detailed way than could be covered star products. How would a uniform bundling or in a typical newspaper. This could lead to a better- subscription strategy (such as CD albums, Rhap- informed global village of citizens and, potentially, sody subscriptions) alter consumers’ consumption better outcomes for society. On the other hand, the patterns? How would a customized bundling or sub- Internet has led to the creation of sites catering exclu- scription strategy (such as eMusic or Netflix subscrip- sively to extreme views on either end of the political 6 tions) change consumers’ consumption patterns? spectrum (e.g., Sunstein 2002). As the Internet allows How should information-goods producers bundle (or consumers to filter contents based on tastes, interests, unbundle) niche content? Should niche content be sold and political viewpoints, it is possible that consumers with other similar content, or in broader bundles of who are fed an exclusive diet of news tilted toward “sampler” content? a particular political viewpoint might be less able to understand and engage in a healthy debate with indi- 4.4. The Impact of the Long Tail on Society viduals holding a different viewpoint. This view of a It is also important to understand how changes in balkanized online world could have a negative impact product variety and concentration patterns affect soci- on society, even as the individuals involved report ety as a whole. What are the social welfare implica- higher satisfaction with their chosen news source (Van tions as firms develop, distribute, and deliver more Alstyne and Brynjolfsson 2005). Again, although both niche products to consumers, and as consumption potential outcomes have been discussed in the litera- patterns swing toward niche (or popular) products because of advances in technologies? How much will ture, there has been little empirical research into the consumers gain in terms of consumer surplus? How actual outcomes as consumers get used to consuming much will firms gain in the form of producer surplus? niche products or niche information contents online, Moreover, consumer surplus gains may not be evenly or how these outcomes might differ across informa- distributed across different types of consumers, just tion categories such as traditional news, public health as producer surplus may not be equally distributed information, and consumer product information. across different types of firms. Technologies also It is important to note that the Long Tail lit- introduce new players into the supply chain. For erature and the Cyberbalkans literature mentioned instance, many niche products in IT-enabled markets above study topics that are inherently connected. At are provided by small niche retailers who are affili- first glance, the Cyberbalkans/Global Village litera- ates of large marketplace operators such as Amazon ture appears to focus on different constructs (infor- and eBay (Bailey et al. 2008). How profitable are these mation and interests) from the Long Tail/Superstar businesses to small niche retailers and large market- literature (consumer purchases). However, informa- place operators, respectively? tion and interests clearly affect purchases, and pur- Long Tail consumption patterns of products and chases can also affect information and interests. For information could also lead to either a global village instance, the consumption of wine can lead to knowl- of citizens who are well informed about both products edge of wine properties and even an evolution of new and policies, or to fractured communications between wine preferences. Just as sales and tastes are natu- balkanized groups of consumers. An optimistic view rally intertwined and affect each other, the Long Tail/ Superstar literature (relating to sales) and the Cyber- 6 Goh and Bockstedt (2008) and Elberse (2010) show that cus- balkans/Global Village literature (relating to interests tomized bundling may shift sales to Superstar songs. and tastes) are inherently intertwined. Brynjolfsson et al.: Research Commentary: The Effect of IT on Product Variety and Sales Concentration Patterns Information Systems Research 21(4), pp. 736–747, © 2010 INFORMS 745

On a more fundamental level, the IT drivers affect- content—the words between the covers—and coming these trends may have commonalities, as may the pare its distance to a new book. Such a distance metric research methods employed. Although the research translates readily into product space, and could enrich literatures have largely evolved separately, joining the our understanding of true product variety and how theory and methods that have been developed can true product variety affects the distribution patterns lead to new insights and enhance the overall vigor of consumers’ purchases. of the research. Factors that are prominent drivers in one literature could be studied as drivers in another literature. For instance, network structure can drive 5. Discussion Search tools, recommender systems, online communi- the Cyberbalkans/Global Village literature, and it can also drive the Long Tail/Superstar literature. An ties, and social networks are all changing the level of example of this would be that the position of a musi- information and interest that consumers have regard- cian’s Facebook page within the Facebook network ing both niche and Superstar products. Continuing could affect the flow of information and attention, advances in IT portend even bigger changes in com- which in turn affects the sales of the musician’s music. ing years. The effects on consumer choices, retailer In addition, factors driving the Long Tail/Superstar strategies, producer profits, and societal welfare are literature could drive the Cyberbalkans/Global Vil- likely to be substantial. However, despite a growing lage literature. An example of this would be that the research literature, the implications of these changes limit of shelf space could affect product sales, trun- remain poorly understood. The goal of this paper has cating the tail. The limit of shelf space can also shape been to highlight the relevant literature driving our consumer interests toward a Global Village. understanding of both the Long Tail and Superstar Viewing these two literatures in parallel has another phenomena and to identify a research agenda to bet- benefit—tools used in one literature can be used in ter understand their impact. the other literature. Given that the methods and tools To start, there are widely differing measures of used thus far in these literatures have relatively lit- product variety and concentration patterns used in tle overlap, viewing them as part of the same phe- the literature, creating seemingly contradictory results nomenon may suggest some immediate possibilities even when none may exist, thus hampering the cre- for cross pollination. For instance, network measures, ation of a cumulative research tradition. In addi- widely used in the Cyberbalkans/Global Village lit- tion, the implications of these measures affect each erature, may be used in the Long Tail/Superstar lit- segment of the “value chain” from producers, to erature. As one example, consider how the similarity retailers, to consumers, to society. At the most basic measures used in Cyberbalkans literature—Euclidean level, both producers and retailers face uncertainty distance across multiple dimensions where each word about how to approach the Long Tail and Super- provides one component of the vector—may be star effects. Are Long Tail markets a passing fad that used in the Long Tail/Superstar literature. The Long will quickly recede toward Superstar and winner- Tail/Superstar literature only measures the share of take-all outcomes, as some predictions in the lit- sales, with each product being treated as a differ- erature would suggest? Does the Long Tail have ent identity as long as it had a different SKU num- longevity, giving producers profitable niches in which ber. Thus, whether products are similar or not does to focus their efforts? At the retailer level, should not make a difference—adding one book about pol- Long Tail markets impact retailer’s strategies such itics to a bookstore filled with economics would be that retailers should focus on particular niche mar- treated in the same way as adding a ninth edition of kets, or will the broader selection retailers such as a book about economics when the eighth edition was Amazon and iTunes continue to see “winner-take- already in stock. Instead, researchers could use simi- all” outcomes in this space? Further down the value larity measures to create a more fundamental estimate chain, how will consumers respond to the new vari- of underlying market variety. As a specific exam- ety of products available in markets, the new search ple, one could classify a book based on its actual and recommendation tools available online, and the Brynjolfsson et al.: Research Commentary: The Effect of IT on Product Variety and Sales Concentration Patterns 746 Information Systems Research 21(4), pp. 736–747, © 2010 INFORMS new promotional channels offered through social net- References works? Finally, will Long Tail consumption patterns Anderson, C. 2004. The long tail. Wired Magazine 12(10) 170–177. of products and information cause fractured commu- Bailey, J., G. Gao, W. Jank, M. Lin, H. C. Lucas, S. Viswanathan. 2008. The long tail is longer than you think: The surprisingly nication between balkanized groups of consumers or large extent of online sales by small volume sellers. Work- will the resulting outcome look more like a global vil- ing paper, Smith School of Business, University of Maryland, lage of citizens who are better informed about both College Park. Bakos, Y., E. Brynjolfsson. 1999. Bundling information goods: Pric- products and policies than they could be in traditional ing, profits, and efficiency. Management Sci. 45(12) 1613–1630. “head heavy” communication environments? Bakos, Y., E. Brynjolfsson. 2000. Bundling and competition on the Integrating these underdiscussed literatures into a Internet: Aggregation strategies for information goods. Market- ing Sci. 19(1) 63–82. common research agenda is important because each Bresnahan, T. F., R. J. Gordon. 1997. The Economics of New Goods. of these areas has made important contributions to the University of Chicago Press, Chicago. information systems field, and it seems likely that the Brynjolfsson, E., Y. J. Hu, M. Rahman. 2009. Battle of the retail channels: How product selection and geography drive cross- importance of these topics, both in the real world and channel competition. Management Sci. 55(11) 1755–1765. in the research literature, will increase in the coming Brynjolfsson, E., Y. J. Hu, D. Simester. 2007. Goodbye Pareto princi- years. Each of these transformative trends is enabled ple, hello long tail: The effect of search costs on the concentration of product sales. Working paper, Massachusetts Institute and driven by IT. As IT becomes more powerful, the of Technology Sloan School of Management, Cambridge, MA. economic and societal importance of these trends will Brynjolfsson, E., Y. J. Hu, M. D. Smith. 2003. Consumer surplus in increase as well. the digital economy: Estimating the value of increased product variety at online booksellers. Management Sci. 49(11) 1580–1596. What’s more, integrating methods across these lit- Brynjolfsson, E., Y. J. Hu, M. D. Smith. 2006. From niches to riches: eratures takes advantage of the fact that the same The anatomy of the long tail. MIT Sloan Management Rev. 47(4) data and technology trends that allow managers to 67–71. Brynjolfsson, E., Y. J. Hu, M. D. Smith. 2009. A longer tail? Esti- track micro-niche products in exquisite detail, that mating the shape of Amazon’s sales distribution curve in 2008. allow executives to leverage small advantages into Working paper, Massachusetts Institute of Technology Sloan dominating global blockbusters, that bring together School of Management, Cambridge, MA. Cachon, G. P., C. Terwiesch, Y. Xu. 2008. On the effects of consumer diverse peoples in unprecedented communities, and search and firm entry in a multiproduct competitive market. that filter out and control all but the most preferred Marketing Sci. 27(3) 461–473. communications, are also transforming the research Chellappa, R., B. Konsynski, V. Sambamurthy, S. Shivendu. 2007. An empirical study of the myths and facts of digitization in the enterprise itself. Specifically, in the coming decade, IS music industry. Workshop on Information Systems and Economics researchers will gain access to nanodata including lit- (WISE), Montréal, Québec. erally trillions of Google search queries, e-mail, and Dewan, S., J. Ramaprasad. 2007. Impact of blogging on music sales: The long tail effect. Workshop on Information Systems and Eco- instant messages, Web clickstreams, radiofrequency nomics, Montréal, Québec. identification (RFID), and other real-time data feeds, Elberse, A. 2008. Should you invest in the long tail? Harvard Bus. and myriad other fine-grained data (Wu and Brynjolf- Rev. 87(7–8) 88–96. Elberse, A. 2010. Bye bye bundles: The unbundling of music in sson 2009). That will make it possible to understand digital channels. J. Marketing 74(3) 107–123. the market and social micro-structures that undergird Elberse, A., F. Oberholzer-Gee. 2008. Superstars and underdogs: An the Long Tail in the context of the Superstar effect, the examination of the long tail phenomenon in video sales. Work- ing paper, Harvard Business School, Boston. Global Village, and Cyberbalkanization. As a result, Fleder, D., K. Hosanagar. 2009. Blockbuster culture’s next rise and we are optimistic about prospects for research in this fall: The impact of recommender systems on sales diversity. area to not only clarify and resolve these key ques- Management Sci. 55(5) 697–712. tions, but also to provide practical, well-grounded Frank, R., P. Cook. 1995. The Winner-Take-All Society: Why the Few at the Top Get So Much More Than the Rest of Us. Penguin, advice to managers and policymakers. New York. Goh, K. H., J. Bockstedt. 2008. Unbundling and the long tail: New evidence on the consumption of information goods. Work- Acknowledgments ing paper, Nanyang Technological University, Singapore. The authors are grateful for the helpful comments from the Hausman, J. 1981. Exact consumer’s surplus and deadweight loss. editor of this journal and four anonymous reviewers and for Amer. Econom. Rev. 71(4) 662–676. generous support from the MIT Center for Digital Business Huberman, B. 2001. The Laws of the Web: Patterns in the Ecology of and the iLab at Carnegie Mellon. Information. MIT Press, Cambridge, MA. Brynjolfsson et al.: Research Commentary: The Effect of IT on Product Variety and Sales Concentration Patterns Information Systems Research 21(4), pp. 736–747, © 2010 INFORMS 747

Internet Retailer Magazine. 2010. Most consumers start their online Rosen, S. 1981. The economics of superstars. Amer. Econom. Rev. shopping research with a search engine (June 1). Accessed 71(5) 845–858. September 4, 2001, http://www.internetretailer.com/2010/06/ Sunstein, C. 2002. Republic.com. Princeton University Press, 01/most-consumers-start-their-online-shopping-research-searc. Princeton, NJ. Joyce, S. 2008. Steve Barnhart, CEO President Orbitz: Economics Tan, T., S. Netessine. 2009. Is Tom Cruise threatened? Using Netﬂix of the Long Tail. Travel and Tourism Technology Trends (Jan- data to examine the long tail of electronic commerce. Work- uary 6), http://tourismtechnology.rezgo.com/2008/01/steve ing paper, University of Pennsylvania, Philadelphia. -barnhart-ceo-president-orbitz-economics-of-the-long-tail.html. Tucker, C., J. Zhang. 2009. How does popularity information Meier, G. 2000. BMG Entertainment. HBS Case 9-701-300, Harvard affect choices? A ﬁeld experiment. Working paper, Mas- Business School, Boston. sachusetts Institute of Technology Sloan School of Manage- Mendelson, H., P. Meza. 2001. Amazon.com: Marching toward prof- ment, Cambridge. itability. Stanford Graduate School of Business Case EC-25, Van Alstyne, M., E. Brynjolfsson. 2005. Electronic communities: Stanford, CA. Global village or cyberbalkanization? Management Sci. 51(6) Noe, T., G. Parker. 2005. Winner take all: Competition, strategy, 851–867. and the structure of returns in the Internet economy. J. Econom. Vincent, F. 2007. MySpace for Musicians. Thomson Course Technol- Management Strategy 14(1) 141–164. ogy, Boston. Oestreicher-Singer, G., A. Sundararajan. 2009. Recommendation Wu, L., E. Brynjolfsson. 2009. The Future of Prediction: How Google networks and the long tail of electronic commerce. Work- Searches Foreshadow Housing Prices and Sales. MIT Center for ing paper, New York University, New York. Digital Business, Cambridge, MA.