The superhit effect and long tail phenomenon in the context of electronic word of mouth

Depósito de Investigación de la Universidad de Sevilla https://idus.us.es/ This is an Accepted Manuscript of an article published by Elsevier in Decision Support Systems, Vol. 125, on October 2019, available at: https://doi.org/10.1016/j.dss.2019.113120 Copyright 2019 Elsevier. En idUS Licencia Creative Commons CC BY-NC-ND The super-hit effect and the long tail phenomenon in the context of electronic Word-of-Mouth M. Olmedilla1, M. R. Martínez-Torres2 and S. L. Toral3 1Léonard de Vinci Pôle Universitaire, Research Center, 92 916 Paris La Défense, France 2Facultad de Turismo y Finanzas, University of Seville, Seville, Spain 3E. S. Ingenieros, University of Seville, Seville, Spain The super-hit effect and the long tail phenomenon in the context of electronic Word-of-Mouth The proliferation of online retailers combined with recommender systems has facilitated the availability of a wide variety of products and services through the Internet. However, the effect of this availability depends on the products’ categories: in some cases, the web promotes even more the super-hit products while keeping a short tail, but in other cases, a long tail effect emerges. This paper deals with the coexistence of the super-hit effect and the long tail phenomenon from a quantitative perspective and proposes two different methods to mathematically prove the presence of both phenomena in 28 different product categories collected from an eWOM (electronic word-of-mouth) website. The aim is to test whether the Internet promotes best-seller products, niche product or both. Findings reveal that eWOM promotes either the long tail phenomenon or the super-hit effect depending on the different product categories and that they can coexist together. Keywords: Electronic word-of-mouth (eWOM); user-generated content; long tail; power law fitting; elbow criterion; 1. Introduction The effect of Internet and the increase in information available to the consumer is redefining a shift in the distribution of the sales of products that are reviewed and purchased online. This subject has caught much attention and discussion among research related to information systems, marketing or operation management [56][29][8]. In this respect, online markets have propitiated that non-hit or niche products represent an important part of the overall product sales [29]. Such transformation of the distribution of product sales is what Anderson called the long tail [1]. The long tail phenomenon can be explained because, although consumers can now find everything effortlessness in the Internet, their choices are based on preferences that are not similarly distributed [17]. Moreover, different search technologies [33] or peer-based recommendations associated with electronic commerce [55] are strongly related to the phenomenon of the long tail of demand, as they lead to a redistribution of the demand toward niche products as well as toward popular products. Actually, the influence of recommender systems in this point seems to be contradictory as they can promote both niche and popular products, but this is an open debate in the current literature. In this regard, earlier studies drive a wedge between two differing directions. On the one hand, research in favor of the long tail idea uphold that boosting the collection of products accessible over online channels (e.g. eWOM communities) will strengthen sales of niche products [10][23]; while others advocate the sales of popular products with even high ratings [65]. In both cases, the role of online channels is the same: to help consumers finding the products they are looking for, no matter if they are less popular or super-hit products. The only difference is that in some cases, the tail is extended while in others the tail is shortened [41]. Many existing studies on the topic focus on e-commerce sites [8][9][10], Web search distributions [35] or cultural products such as music [23], books, films or video games [66]. This paper expands the border of the long tail quantitative research by providing a new perspective on the understanding of both the super-hit effect and long tail phenomenon of consumption patterns of online users within an eWOM community. Specifically, it is studied the presence of super-hit or niche products across the entire spectrum of product categories distribution within an eWOM community. This analysis is based on a unique data set of 105,918 online reviews of 68,650 products posted on a popular eWOM community by 44,352 different users. The accessibility to online reviews posted by users within the proper product category and subcategory allows differentiating the type of distribution across products. The overall aim of the paper is to test whether the existence of the super-hit effect is compatible or not with the long tail phenomenon. To that end, this study compares two methodologies to mathematically measure a heavy-tailed and a long tail distribution: the power-law fitting method of the distribution of the number of products (supply side factor) per number of online reviews (demand side factor), previously proposed by Clauset et al. [16] and developed by Olmedilla et al. [56], and the elbow criterion demarcated by the power-law fitting. The main contribution of the proposed methodology is its ability to mathematically determine the presence of super-hit and niche products. The rest of the paper is organized as follows. The next section discusses the related work and conceptual framework by conducting a literature review to develop the two proposed research questions. Then, the following section describes the research methodology. Here, the case of study and data collection is detailed as well as the two methods used to fit the powerlaw distribution. The results are next presented. Finally, the last section concludes the study with the discussion and the implications of the study. 2. Related work and conceptual framework 2.1 Super-hit effect vs long tail phenomenon As aforementioned, the power-law distribution curve presents a small group of high frequency events at the head followed by a large set of low frequency events in the tail. In the sales distribution, that acquires the form of a power-law distribution, the head represents the super-hit products and the long tail the niche products [39]. Figure 1 depicts an example of a power-law distribution with all probable events ranked by their frequency. To the left (blue) are the super-hits that dominate and to the right (orange) is the long tail with niche content. In this example there is a red cutoff selected so that areas of both the head and the long tail are equal. Actually, according to Elberse [23], the orange area under the curve may become bigger than the blue area over time. [FIGURE 1] The super-hit effect and the long tail phenomenon are two contrasted theories still studied among many researchers. The long tail theory advocates that a big part of consumers prefer popular content while only a minority pursue niche content [28]. In this regard, Anderson [1] theorizes that the tail at the end of the distribution curve is practically different from the head. He adds that consumers are moving toward the niches in the tail since the tail satisfies rare interests better [2]. Today, the Internet is allowing consumers to obtain information related to low-volume products from a massive, geographically dispersed group of people, which facilitates this long tail effect [33][56]. Besides, in some areas such as music, movies or blogging, the tail is theoretically unlimited since the Internet offers infinite storing space. Since Anderson [1], the studies presented until now about the long tail have provided a theoretical framework or empirical details about the sales of rare products not available in traditional brick- and-mortar stores. For instance, some studies by Brynjolfsson et al. [8][9][10] draw attention to Internet’s long tail by analyzing data collected from a clothing retailer or from sales distributions in product markets such as Amazon. They consider that consumers find it easier to search for and discover niche products because (1) the Internet has eliminated communication barriers as well as stocking and distribution costs; and (2) due to the proliferation of Internet discovery tools such as recommendation engines. Similarly, the author Elberse [23] argues that the long tail of the sales distribution is getting longer and fatter using data from Quickflix or Rhapsody Music. Other researchers studying power-law distributions of the consumers’ actions in the Internet find that much of the action occurring within power-laws is placed at the head of the distribution where there are less but with high frequency events [20], in essence, the super-hit effect. Such effect was originally presented to describe the few top performers reaching most of the audience and achieving the majority of profits [62] and currently still thrives among online markets [9][71]. In this regard, Hennig-Thurau & Houston [30] provide empirical evidence that the super-hit effect still leads in entertainment and it is guiding consumers away from traditional forms of entertainment to new ones such as social media. Dellarocas [20] reveals through a Gini coefficient analysisas measure of concentration of demandthat the weekly volume of movie online reviews is more skewed towards popular movies. Consequently, thanks to the Internet and the digital technologies and networks [11], a product becomes more appreciated as the more users it has, meaning that the super-hit effect is affecting more products than before. 2.2 electronic Word-of-Mouth (eWOM ) Consumers are progressively taking more part in sharing their experiences with a product by writing online reviews, and eWOM is widely recognized as an influential information source on consumer decision-making, which is well established in academic literature [36][59]. Besides, the emergence and impact of user-generated content within the Web has made classic WOM moving towards eWOM [31], which has been enabled by the Internet and online communication [50]. Actually, the eWOM—considered as an inexpensive online medium—is currently decreasing the barriers of entrance in the markets of the products [49]. eWOM often has a strong impact on product judgements because the information received is more accessible. Reviews are published on a great variety of products and services and have become part of the decision-making process for consumers [52]. Evidence from research suggests the capability of eWOM to find niche products and thus facilitate the long tail phenomenon [31][37], and the way eWOM is decreasing the cost of the information search about products [54]. For instance, a study on this topic has concentrated on assessing the relationship between long tail and product sales [9]. Besides, a more discerning approach is that “the effect of eWOM is not monolithic”, but it differs according to product characteristics [48]. Existing studies on eWOM have principally dealt with information on products such as movies [67][68], books [8][15] or music [51]. In this regard, online reviews within eWOMs are very significant when consumers are choosing products that they do not have first-hand experience with. Online reviews are also generally used to reduce uncertainty regarding service quality from the consumer perspective [60]. Besides, the study by Park and Lee [58] describes how eWOM information’s direction and website’s reputation contribute to the eWOM effect focusing on the moderating role of the product type. 2.3 Fitting power-law distributions online Not always the measured processes reach their maximum point around a typical value, sometimes those diverge across a wide range [53]. For instance, when the variance and/or the mean is not finite any longer, the Central Limit Theorem does not predict Gaussian distributions, but it predicts distributions that look like power-laws. Besides, unlike normal distributions, observations in power-law distributions are far to the left of the mean and some outliers represent an uneven quantity of the total distribution’s output [18]. In short, large events happen more often in systems that exhibit power-law distributions. This means that the events in the tail of the distribution are more likely to occur in a power-law distribution. That is why power-laws are also called heavy-tailed [3][6]. Thus, the extreme tails are important because they give an idea of how often, on average, the largest events might occur. Consequently, power-law distribution is considered a valuable tool to measure these information uncertainties since Gaussian distributions cannot handle when happening at a certain probability. Former and current research has examined whether the Internet phenomena are better described by power-law distributions [4][7][43][46]. Findings in related fields show that power-laws distributions usually emerge in the Internet social systems [34], where users express their preferences among many choices (e.g. YouTube [27], eWOM communities [56], Amazon [14]). Similarly, Brynjolfsson, Hu and Smith [8] argues that a power-law distribution can be used to describe the relationship between a product’s sales rank and sales quantity. As can be observed, the interaction of online users exchanging information is producing complex dynamics in the demand for products. Consequently, although there are other heavy-tailed distributions (e.g. lognormal, t-distribution, exponential or chi-squared distribution), this paper focuses on the power-law distribution because the super-hit effect and the long tail phenomenon prevail in online consumption patterns in which a relatively small number of very popular products account for the majority of sales [9] and the tail of the distribution encloses niche products [2]. To fit the power-law distribution within this paper the number of reviews is used, nor is the content. Research has shown that the message content of the online reviews has less impact when users look at many reviews if they are about to make any purchase decision [5]. In this regard, Shao [64] argues that a higher number of reviews, either positive or negative, is more probable to gather consumers looking for information and consequently increases product awareness. The author [44] defended no relationship between the valence of the online review (negative or positive value assigned to an online review) and sales. In this regard, many authors defended that a higher number has a greater influence on sales [13] [44] [22]. Interestingly, though, review volume does not necessarily mean the reviews are positive. Numerous studies analyze the effect of online reviews on product sales or consumers’ choices by also considering the characteristics of the reviews. In this respect, Chen et al., [12] argued that “negative WOM information has a greater impact on product sales than positive WOM information”. In fact, many authors have found that the negative reviews have a better impact on purchase intention than positive reviews [19]. This positive effect of negative reviews has been evidenced in numerous studies. For instance, Ghose and Ipeirotis [26] associate reviews of products rated negatively with product sales. This author’s results are consistent with Berger et al. [5] who show that negative reviews have a positive impact on sales since they can increase publicity for lesser known products. Likewise, the results by Gavilan et al. [25] indicate that the negative reviews are more used for holiday decision-making than the positive ones. In general, users consider negative reviews more useful [63] or demonstrative for decision-making purposes [42] than positive reviews, since negative information is attention catching and incite curiosity [61] and controversial reviews can provoke a lot of discussion [64]. Consequently, as the proportion of negative or positive online consumer reviews increases, this means that there is a highinvolvement of consumers. Moreover, the research by Zhang et al. [70] find that the consumers that notice a high number of reviews arouse their purchase intention of a product. Besides, the authors do not explicitly distinguish the number of reviews into positive or negative. Consequently, the more online product reviews, the more likely a consumer will be informed about a product. There are other several relevant articles that have verified a positive impact of online reviews on sales by taking into account the volume of reviews. For instance, a recent article by Maslowska et al. [47] find that products that have a high price benefit from a big number of reviews since potential customers detect that others might have bought the product. In this sense, Park et al. [59] also affirm that the number of reviews can specify a product’s popularity due to the fact that costumers assume a relation between the number of reviews and the consumers who have bought the product. Likewise, findings by Zhou and Duan [72] illustrate that a larger volume of internal WOM, which depends on the number of reviewers, has a positive impact on sales. Extant studies, such as the one by Chevalier and Mayzlin [15] indicate that book sales between two sites (Amazon.com and Barnesandnoble.com) are related to the difference in the number of reviews per book throughout the sites. Additionally, Zhang et al. [69] find that the quantity of online reviews, among other characteristics, has an important influence on the sales of digital cameras as well as on sales of search goods. threshold, that is, the number of subcategories that are not part of the fitted power-law distribution. The fifth and sixth columns represent the areas of the head and the tail of the distribution, respectively, calculated also using the 𝑥𝑚𝑖𝑛 value. The values of 𝑥𝑚𝑖𝑛 distinguish between those subcategories that are part of the head of the power-law distribution (super-hit products, above 𝑥𝑚𝑖𝑛 ) and those that belong to the tail (niche products, below 𝑥𝑚𝑖𝑛 ). The ∝ exponent determines the shape of the power-law distribution. The fitting has an associated p-value, so that in those cases where the calculated p-value is considerably lower than 0.05, the null hypothesis is rejected, which means that such category does not to follow a power-law distribution and, therefore, there is no super-hit effect. But if the resulting p-value is greater than 0.05, that means that the null hypothesis cannot be rejected and hence the distribution is likely to follow a power-law distribution, showing then a super-hit effect. Finally, the head area percentage column denotes the area contained to the left of the cut-off point of the 𝑥𝑚𝑖𝑛 , which includes the dominating products or best sellers. The tail area percentage column embodies the tail, that is, the region to the right of the cut-off point of the 𝑥𝑚𝑖𝑛 . [TABLE 1] The same analysis was conducted but using the elbow method. Table 2 describes the results for the same 28 main categories. The first column displays the 𝑥𝑚𝑖𝑛 values, which correspond to the turning point, that is, the point from which the function goes from decreasing very fast to decreasing very slowly. Accordingly, the function is traversed from the right to the left, from 𝑥𝑚𝑎𝑥 to -1, until the first point/value that meets the search condition. This is the point of average slope between the two asymptotes of the graph. The second column shows the length of the tail, which was calculated using the calculated turning point. Finally, the third and fourth columns represent the areas of the head and tail of the distribution, respectively, (calculated using the obtained turning point). [TABLE 2] Both methods, the power-law fitting method and the elbow criterion, were compared in Table 3 to test their coincidence or not when identifying the presence of the long tail and considering the area and length criteria. The result of this comparison is shown in the last column of Table 3. A True value means that both methods agree about the presence or absence of the long tail, a False value means that both methods disagree and Uncertain means that both methods generate contradictory results. This is the case of Ciao Café, Electronics, Shopping, or Travel. The decision making was conducted by applying the joint-ratio 80:20 to the area under the tail and the length of the tail. Whenever the area or the length of the tail is under the 20% of the maximum value, it is considered the there is no long tail. Otherwise, the presence of the long tail is accepted. [TABLE 3] Figure 6 and Figure 7 depict some of the distributions for the main categories shown in Table 3. The graphs show a power-law distribution fitted to the curve of the distribution of subcategories of products. Accordingly, the Y axis corresponds to the volume of reviews (number of reviews written by the users) and the X axis represents the all the subcategories of products of each of the 28 main categories of products distinguished by Ciao UK. In both Figure 6 and Figure 7, the 𝑥𝑚𝑖𝑛 value is demarcated by a horizontal red line representing its value calculated by the elbow method, and by a horizontal blue line representing its value calculated by the power-law method. According to Table 3, 15 of the 28 main categories (Adult Products, Beauty, Books, Entertainment, Family, Fashion, Finance, Food & Drink, Games, Health, Household & Appliances, Music, Musical Instruments, Sports & Outdoors, Telecommunications) clearly exhibit a long tail behavior as both criteria are accomplished using the two proposed methods. Regarding the super-hit effect, he p-values from Table 1 reveal that 7 out of the 28 categories among the data sets (Beauty, Books, Cars & Motorcycles, Computers, Internet, Sports & Outdoors, Telecommunications) do not obey a power-law distribution (see Figure 7). Their p-values is small enough so that the power-law model can be firmly discarded, according to their goodness of fit. As a difference and according to their p-values, the remaining 21 categories might be consistent with a power-law distribution. Consequently, and answering RQ1, eWOM promotes both the super-hit and the long tail phenomena although the super-hit effect is more frequent across the 28 categories considered. Among all those 21 cases with the super-hit effect, there are 6 cases without a long tail (DVDs, Education & Careers, Cameras, House & Garden, Office Equipment, Software), 11 with long tail (Adult Products, Entertainment, Family, Finance, Food & Drink, Games, Health, Household Appliances, Music, Musical Instruments & Equipment, Fashion), and finally, 4 of them that it is not clear whether they exhibit a long tail or not (Ciao Café, Shopping, Travel, Electronics). These findings answer RQ2 by showing that it is possible that the super-hit effect coexists with the long tail phenomenon depending on the product categories. [FIGURE 6] [FIGURE 7] 6 Discussion and implications 6.1 Theoretical implications On the theoretical front, this paper contributes to the growing research on how the Internet is influencing the super-hit effect and the formation of long tail across Internet data and more specifically user-generated content. The results suggest that it is important to consider the impact of the eWOM communities on the consumers product discovery process, due to the fact that many products are only sold online and their consumers are more prone to use online reviews as the primary source for information. Existing literature related to the long tail has shown that eWOM has a positive impact on sales. For instance, Brynjolfsson et al. [10] claim that consumers are previously informed about products before buying and thus, niche products are gaining an important share of overall demand. Likewise, Elberse and OberholzerGee [24] show similar results doing a comparison of the offline and online sales of DVD and VHS video titles. The authors show that improvements in word of mouth can explain that the increase in the assortment of products offered through online channels shift to a much larger number of niche products. Inversely, Hervas-Draney [32] show that word of mouth benefits the most the winner-take-all or market products and mainstream consumers. As can be inferred, the interaction of users exchanging information about products within eWOM communities creates intricate dynamics in demand. More existing studies of eWOM have found evidence of the super-hits effect. For instance, Standifird [65] supports that eWOM promotes the sales of popular products, so the head part of the sales distribution becomes thicker. The premise of our study is to clarify these two schools of thought by using the distribution of numerous product categories online reviews of an eWOM community comprising a total of 283.240 subcategories of 68.650 products to identify in which one is promoting the superhit and in which other the long tail. In this regard, our study has found evidence of eWOM as a promoter of both the super-hit effect and the long tail phenomenon depending on the product categories. However, and following previous studies, the presence of the super-hit effect is more extended than the long tail. This result means that the long tail phenomenon is closely related to certain categories of products where reviews disseminate new knowledge about items or subcategories unknown for most users. In addition, we found evidence that the long tail phenomenon is closely related to certain categories of products (e.g. Beauty, Books, Entertainment, Food & Drink, Health, Games, Music, etc.). Findings are supported by two quantitative methods that mathematically provide evidence about the super-hit effect and the long tail phenomena. Moreover, our study also has important implications for the literature on the impact of eWOM on different products. We demonstrate that the long tail phenomenon is able to coexist with the super hit effect depending on which product category. In this regard, research by Lee et al. [41] argued that eWOM can change the rule of the long tail theory by differentiating the cases across product types. The authors categorized product types based on the objectivity of product evaluation standards. Not surprisingly, Internet is making it easier for consumers to gather information at a low cost and allows them to relay more on the experience of others when they evaluate products. Again, the increased availability of online reviews about a lot of different products within eWOM communities is the explanatory factor for this phenomenon, given that this online channel allows customers to find more niche products that fit their preferences or to promote products’ popularity. The coexistence of super-hit and niche products also means that eWOM has probably expanded an original short tail. Cases such music, entertainment or fashion are examples of products categories where online markets bring the diffusion of new options otherwise unknown by customers. Additionally, online reviews provide not only the description of new options but also the quality of experience. Most of reviews focus more on the feelings, perceptions and satisfaction than on objective characteristics of products. Therefore, online channels offer more complete information than the one that can be found on catalogs or advertising. 6.2 Managerial implications In general, Internet is changing the way consumers behave, and our study contribute to a better understanding of such consumer behavior within online markets. In this regard, our study offers a new perspective on how online information, specifically within an eWOM community, influences the product awareness effect of both super-hit and niche products. This disparity leads to a rich-get-richer situation, favoring the products that are already very popular on certain product categories and also to a more egalitarian distribution facilitating the formation of the long-tail among other product categories. Consequently, as many assortments of products are only sold online and their consumers are more prone to use online reviews as the primary source for information, managers should understand that the underlying economic principles are finding out what kind of products are enclosed in the tail or in the head of the distribution. Also in this line, the study provides new insights about potential markets that can be open as a result of the tail expansion. Companies can shift towards a specialized offer, focusing on products belonging to the tail of the distribution rather than to the head. A prior discovery of those categories exhibiting a long tail can then be used to gain new markets advancing future consumer trends. Today, any brand or company can at zero-cost reach many potential consumers via Internet. This makes it profitable to invest time and effort to create products that might be of interest to even just a tiny segment of them. Therefore, a strategy of increasing product variety and product information about niche products, might permit consumers to discover products that otherwise would be unavailable. 6.3 Implications for academics and practitioners This work has implications for academics and practitioners. From the academic perspective, the study offers the possibility of advancing on the conditions that favor the appearance of the two studied phenomena. Also, the distributions of products across categories can change over time, modifying then the length of the tail or even the existence of the superhit effect. In general, the Internet is changing the pattern of behaviors of consumers, and this study contribute to a better understanding of consumer behaviors in online markets. Due to the current importance of eWOM in businesses and economics, our study has significant implications for multi-product retailers. A retailer needs to comprehend the impact of eWOM on the distribution of different products in order to offer the accurate product variety and manage inventory. In this regard, our results show what products of the distribution of product categories form the long tail or are super-hits, which allows specifying the cases in detail. This implies that practitioners need to be strategic in choosing the product assortment they are offering online. Consequently, online retailers may need to pay more attention if their objective is to capture the mass market, then, they should acquire and manage customers by using the super-hit or most popular products and focus on marketing those super-hit products as well. Contrariwise, if their objective is use long tail strategies to sell a wider range of goods in smaller quantities, they should focus on broaden their assortment with more niche products. Likewise, firms can gain leverage of the results to develop better marketing strategies on advertising and promotion of products that consumers are more likely to purchase in the future. 6.4 Research limitations The principal limitation of this paper could be that the methodology has been implemented in just one eWOM community. Nevertheless, it would be possible to extend the two methods (power-law fitting and elbow criterion) to fit the distribution of products on other eWOM communities. Other probable methodological limitation in the analysis of this study could be the sample of the data set. This is due to the fact that Ciao UK does not represent all the population but a particular sub-set of online users. Besides, users within Ciao might have multiple profiles, or even the users’ profiles might be employed by several people. Whereas this limitation does not bias the results, they have to be understood as applying to a certain set of online users. A further limitation of the study is the fact that a user does not have to provide any purchase proof of the reviewed product in order to be allowed to write a review. However, writing reviews reports the following benefits to a user: (1) the status benefit that can be achieved from writing reviews, receiving review ratings from other users, rating and commenting on other reviews and referring members to the site and (2) the economic benefit, which depending on the status a user has within the community and if he/she writes reviews frequently, a certain amount of money is paid by Ciao. Consequently, to obtain such benefits, the users might not be interested in being dishonest. Finally, another limitation of the data is the absence of sales data, since this paper draws on online reviews volume, which is the best available data within an eWOM community. Nevertheless, the context of this study analyzes product diversity of those supposed to be purchased before the user writes any online review. 6.5 Future work Further research can extend the findings by characterizing the niche products across the long tail and the super-hit products in the head of the distribution. The goal would be discovering some common patterns among niche products through the creation of social network models or association rules and discover clusters of niche products that can represent a profitable target for retailers. Furthermore, products can be classified as search and experience products, thus, as a future work it could be analyzed which type of product is more prone two exhibit the studied behaviors. It would be also interesting to be able to test the validity of presented the models on other eWOM data, where there is available specific information of sales (e.g. Amazon). 7 Acknowledgements This work was supported by Fundación Hergar under the Research Project titled “Caracterización del fenómeno de la “Cola Larga” en los portales de boca a boca electrónico” with reference AIFH2017/005. 8 References [1] C. Anderson, The long tail, Wired Magazine 12 (10) (2004). [2] C. Anderson, Long Tail: Why the Future of Business is Selling Less of More, New York: Hyperion Books, 2008 [3] H. Aguinis, E. O’Boyle Jr, E. Gonzalez‐Mulé, H. Joo, Cumulative advantage: Conductors and insulators of heavy‐tailed productivity distributions and productivity stars, Personnel Psychology 69(1) (2016) 3-66. [4] A.L. Barabsi, R. Albert, Emergence of scaling in random networks, Science 286 (1999) 509–512. [5] J. Berger, A.T. Sorensen, S. J. Rasmussen, Positive effects of negative publicity: When negative reviews increase sales, Marketing Science, 29(5) (2010) 815-827. [6] K. Bimpikis, M.G. Markakis, Inventory pooling under heavy-tailed demand, Management Science, 62(6) (2015) 1800-1813. [7] L. Breslau, P. Cao, L. Fan, G. Phillips, G., S. Shenker, Web caching and Zipf-like distributions: Evidence and implications, INFOCOM’99, Eighteenth Annual Joint Conference of the IEEE Computer and Communications Societies, Proceedings IEEE. (1) (1999) 126-134. [8] E. Brynjolfsson, Y. Hu, M. Smith, Consumer surplus in the digital economy: Estimating the value of increased product variety at online booksellers, Management Science 49(11) (2003) 1580-1596. [9] E. Brynjolfsson, Y. Hu, M. Smith, Research commentary-long tails vs. superstars: The effect of information technology on product variety and sales concentration patterns, Information Systems Research, 21(4) (2010) 736-747. [10] E. Brynjolfsson, Y. Hu, D. Simester, Goodbye pareto principle, hello long tail: The effect of search costs on the concentration of product sales, Management Science 57(8) (2011) 1373-1386. [11] E. Brynjolfsson, A. McAfee, M. Spence, New world order: labor, capital, and ideas in the power law economy, Foreign Affairs 93(4) (2014) 44-53. [12] Y. Chen, Q. Wang, J. Xie, Online social interactions: A natural experiment on word of mouth versus observational learning, Journal of marketing research, 48(2) (2011) 238-254. [13] P.Y. Chen, S. Y. Wu, J. Yoon, The impact of online recommandations and consumer feedback on sales. ICIS 2004 Proceedings, (2004) 58. [14] J. Chevalier, A. Goolsbee, Measuring prices and price competition online: Amazon.com and BarnesandNoble.com, Quantitative marketing and Economics 1(2) (2003) 203-222. [15] J. Chevalier, D. Mayzlin, The effect of word of mouth on sales: Online book reviews, Journal of Marketing Research 43(3) (2006) 345–354. [16] A. Clauset, C. Shalizi, M. Newman, Power-law distributions in empirical data, SIAM Review 51(4) (2009) 661-703. [17] E.K. Clemons, How information changes consumer behavior and how consumer behavior determines corporate strategy, Journal of management information systems 25(2) (2008) 13-40. [18] G.C. Crawford, H. Aguinis, B. Lichtenstein, P. Davidsson, B. McKelvey, Power law distributions in entrepreneurship: Implications for theory and research, Journal of Business Venturing 30(5) (2015) 696-713. [19] G. Cui, H.K. Lui, X. Guo, X. The effect of online consumer reviews on new product sales, International Journal of Electronic Commerce, 17(1) (2012) 39-58. [20] C. Dellarocas, R. Narayan, R. Tall heads vs. long tails: Do consumer reviews increase the informational inequality between hit and niche products?, School of Business Research Paper No. 06-056 (2007). [21] C. Dellarocas, G. Gao, R. Narayan, Are consumers more likely to contribute online reviews for hit or niche products?, Journal of Management Information Systems 27(2) (2010) 127-158. [22] W. Duan, B. Gu, A.B. Whinston, Do online reviews matter?—An empirical investigation of panel data, Decision support systems 45(4) (2008) 1007-1016. [23] A. Elberse, Should you invest in the long tail?, Harvard Business Review 86 (7/8) (2008) 88-96. [24] A. Elberse, F. Oberholzer-Gee, Superstars and underdogs: An examination of the long tail phenomenon in video sales, Harvard Business School (2007) 07-015. [25] D. Gavilan, M. Avello, G. Martinez-Navarro, The influence of online ratings and reviews on hotel booking consideration, Tourism Management, 66 (2018) 53-61. [26] A. Ghose, P.G. Ipeirotis, Estimating the helpfulness and economic impact of product reviews: Mining text and reviewer characteristics, IEEE transactions on knowledge and data engineering, 23(10) (2011) 1498-1512. [27] P. Gill, M. Arlitt, Z. Li, A. Mahanti, Youtube traffic characterization: a view from the edge, Proceedings of the 7th ACM SIGCOMM conference on Internet measurement (2007) 15-28. [28] S. Goel, A. Broder, E. Gabrilovich, B.Pang, Anatomy of the long tail: ordinary people with extraordinary tastes., Proceedings of the third ACM international conference on Web search and data mining (2010) 201-210. [29] B. Gu, Q. Tang, A. Whinston, The influence of online word-of-mouth on long tail formation, Decision Support Systems 56 (2013) 474–481. Figure 1. Example chart representing the power-law distribution of products/events Figure 2. Level specification of the main category “Travel” Figure 3. Level specification of the main category “Beauty” Figure 4. Application of the elbow criterion to the power la distribution Figure 6. Distribution of reviews for some of the main categories in Ciao UK (I) Figure 7. Distribution of reviews for some of the main categories in Ciao UK (II) Figure 5. Distribution of posted reviews for the 28 Main Categories Figure 1. Example chart representing the power-law distribution of products/events Figure 2. Level specification of the main category “Travel” long tail short head more generic more specific frequency events Music Yes Yes Yes Yes True Musical Instruments & Equipment Strong Yes Strong Yes Strong Yes Strong Yes True Office Equipment No No No No True Shopping No Yes No Yes Uncertain Software No No Yes Yes False Sports & Outdoors Yes Yes Yes Strong Yes True Telecommunications Yes Yes Yes Strong Yes True Travel Yes No Yes Yes Uncertain Figure 3. Level specification of the main category “Beauty” Figure 4. Application of the elbow criterion to the power la distribution 0 500 1000 1500 2000 2500 AdultProducts Beauty Books Cameras Cars&Motorcycles CiaoCafé Computers DVDs Education&Careers Electronics Entertainment Family Fashion Finance Food&Drink Games Health House&Garden HouseholdAppliances Internet Music MusicalInstruments&Equipment OfficeEquipment Shopping Software Sports&Outdoors Telecommunications Travel Main Categories No. of sucategories No. of Subcategories Main Categories Figure 5. Distribution of posted reviews for the 28 Main Categories Figure 6. Distribution of reviews for some of the main categories in Ciao UK (I) 0 30 60 90 0 500 1000 1500 2000 2500 Subcategories of Beauty Number of reviews 0 50 100 150 0 200 400 600 Subcategories of Books Number of reviews 0 50 100 150 200 0 25 50 75 Subcategories of Family Number of reviews 0 50 100 150 200 0 200 400 Subcategories of Food & Drink Number of reviews 0 50 100 0 100 200 300 400 Subcategories of Games Number of reviews 0 50 100 150 200 250 0 20 40 60 Subcategories of Household Appliances Number of reviews Figure 7. Distribution of reviews for some of the main categories in Ciao UK (II) 0 20 40 60 80 0 20 40 60 80 Subcategories of Cars & Motorcycles Number of reviews 0 50 100 150 0 50 100 150 Subcategories of Computers Number of reviews 0 100 200 300 400 0 20 40 60 80 Subcategories of Telecommunications Number of reviews 0 50 100 0 10 20 30 40 Subcategories of DVDs Number of reviews 0 50 100 150 200 250 0 10 20 30 40 Subcategories of Electronics Number of reviews 0 100 200 300 0 10 20 30 40 Subcategories of House & Garden Number of reviews