Market share, profit margin, and marketing efficiency of early movers, bricks and clicks, and specialists in e-commerce

Transcription

1 Journal of Business Research 58 (2005) Market share, profit margin, and marketing efficiency of early movers, bricks and clicks, and specialists in e-commerce Sungwook Min*, Mary Wolfinbarger Department of Marketing, College of Business Administration, California State University, 1250 Bellflower Boulevard, Long Beach, CA 90840, USA Received 18 November 2002; accepted 26 February 2004 Abstract We examine three strategic issues that have been frequently discussed and debated in e-tailing: early mover advantages, bricks and clicks combinations as compared with pure-plays, and generalizing versus specializing. We develop hypotheses and test them utilizing annual and quarterly financial data from 42 public online retailers. We use a multilevel repeated-measures model to analyze our data because of its flexibility and ability to handle unbalanced repeated measures. The results indicate that early movers in e-commerce do not have a significant advantage in market share, profit margin, or marketing efficiency compared with later entrants. The bricks and clicks combinations in our sample possess higher market share and higher marketing efficiency than do pure-plays. Last, while specialists have lower market share and lower marketing efficiency, they nevertheless can be successful because they command higher profit margins than do generalists. The managerial implications of our findings are discussed. D 2004 Elsevier Inc. All rights reserved. Keywords: E-commerce; E-tailing; Early mover advantage; Bricks and clicks; Online specialists; Repeated-measures design 1. Introduction Presently, empirical research on the strategic levers that drive e-tailer profitability and success is limited (Grewal et al., 2003). The few studies that are reported are largely managerial and descriptive and generally do not use sufficient controls or measures of variables such as size of firm and product category that can impact the success of a particular strategy (Bughin and Zeisser, 2001). Case studies and stories about individual stars, such as Amazon.com and Dell abound, are suggestive but hardly generalizable (Zhu and Kraemer, 2002). At the same time, existing research on offline retailing is likely to apply to e-tailing in a limited fashion, given that there are unique operational issues, such as website design, privacy, and security when selling online (Grewal et al., 2004). Thus, research studying strategic marketing issues in the online setting is likely to uncover findings unique to e-tailing environments. In this research, we address three important strategic marketing issues in e-commerce simultaneously. First, we * Corresponding author. Tel.: ; fax: address: smin2@csulb.edu (S. Min). examine whether there are early mover advantages in the e- commerce context. Writers have argued for and against their presence in e-tailing. For example, Porter (2001) argues that the first mover advantage in the Internet is a myth, specifically suggesting that, switching costs are likely to be lower, not higher, on the Internet than they are for traditional ways of doing business...on the Internet, buyers can often switch suppliers with just a few mouse clicks (p. 68). On the other hand, Reichheld and Schefter (2000) insist that price doesn t rule the web, trust does and claim that they have observed the presence of significant loyalty effects in e-tailing. If first movers use their early entry to determine who their profitable customers are and what their customers want, and then successfully fulfill promises to customers, then, first movers will be the recipients of loyalty effects. The second strategic issue that we address in this paper is how bricks and clicks combinations perform in Internet commerce as compared with pure-plays. Many pure dotcom players viewed the relative newness of e-commerce as an opportunity to introduce a new business and brand to an innovative audience. They believed that their new businesses would be more efficient than the older offline businesses, offering lower overhead, better prices, and, thus, competitive advantages as compared with offline counter /$ see front matter D 2004 Elsevier Inc. All rights reserved. doi: /j.jbusres

2 S. Min, M. Wolfinbarger / Journal of Business Research 58 (2005) parts. Nevertheless, many industry analysts now predict that as the online channel evolves, incumbents [bricks and clicks combinations] will have a golden opportunity (Silverstein et al., 2001). Another key research question is Do specialists perform better than generalists do in e-tailing? Online specialists carry narrow product lines with a deep assortment within each line. Generalists carry a wide product mix with varied product lines. Both strategies have been adopted and implemented online abundantly. The broad product line of generalists facilitates cross selling for e-tailers who have developed a trustworthy reputation. However, there may be advantages to specializing on the Internet because the web aggregates niche markets and offers consumers products that they may not be able to buy easily offline. The main contributions of this study are twofold. First, we empirically test three widely discussed e-commerce strategies simultaneously. By doing so, we control a significant portion of variation in e-tailing performance. For instance, we test early mover advantage hypotheses, holding other important strategic variables constant, such as bricks and clicks, and specializing. Thus, we can more accurately assess the success of particular e-tailing strategies than do industry commentators. Second, we measure market share, profit margin, and marketing efficiency utilizing annual and quarterly financial data rather than focusing on intermediate performance measures, such as number of visitors or customer satisfaction, or on the penultimate outcome, survival. Thus, this study enriches the understanding of e-commerce strategies and their outcomes. 2. Hypotheses In our development of hypotheses, we use three measures of online retailing performance: (1) market share, (2) profit margin, and (3) marketing efficiency. Market share has been widely used as a performance measure to assess first mover advantages in empirical (e.g., Kalyanaram and Urban, 1992; Urban et al., 1986; Robinson and Fornell, 1985) and theoretical studies (e.g., Bohlmann et al., 2002; Freshtman et al., 1990). We define profit margin as (sales revenue cost of goods sold)/sales revenue. Profit margin is the proportion contributed to profit with one dollar of sales revenue; thus, lower profit margins are associated with lower profit levels. Finally, we define marketing efficiency as sales revenue/marketing and selling expenses. Marketing efficiency is the sales revenue response to one dollar of marketing and selling spending Early mover advantages in online retailing The excellent survey of Lieberman and Montgomery (1988) of the literature suggests that first mover advantage can arise from several sources, including a pioneer s preemption of assets, learning and patent protection, buyer preference for the pioneering brand, and buyer switching costs. Ample empirical studies on first mover advantages find that early movers have greater market share (e.g., Urban et al., 1986; Robinson, 1988; Robinson and Fornell, 1985) and enjoy higher consumer trial and repeat purchase rates than later entering brands do (Kalyanaram and Urban s, 1992). Early mover advantages in e-tailing can be both customer- and producer-based (Golder and Tellis, 1993). Potential customer-based advantages in e-commerce include (1) customer familiarity with a specific website and its features and (2) the trust that results when a customer successfully places an order and receives items (Jarvenpaa and Tractinsky, 1999; Reichheld and Schefter, 2000; Urban et al., 2000). An additional possible customer-based benefit for early e-commerce movers is the ability to gain superior knowledge about customers. If this information is used in a way that respects consumer interests, time, and privacy, early movers will be able to leverage their accumulated customer knowledge to a greater degree than later entrants can. In addition to the customer-based benefits, there are possible producer-based benefits. These benefits include desirable website names, partnerships with major portals and content sites, and affiliate relationships. Early entrants may have more access to financial resources than do later movers. Economies of scale and learning can lead to lower costs for early entrants (Robinson and Fornell, 1985; Urban et al., 1986). Finally, staying at the forefront of technology may enable early movers to offer better website functionality and more efficient and effective fulfillment than their competitors. But there may be disadvantages for early entry as well (Golder and Tellis, 1993; Lieberman and Montgomery, 1988). At least three disadvantages are directly relevant to e-commerce. First, in e-commerce, late movers can free ride on early movers investments, including improved buyer education and awareness and increased security of transactions and other infrastructure development. The effect of free riding might be substantial, as early movers in e- commerce faced skeptical customers who were hesitant to make an online purchase (Greenspan, 2002). Second, late movers can enter after technological uncertainties in e-commerce are resolved. In many new-product markets, technological uncertainty is resolved through the emergence of a dominant design (Suarez and Utterback, 1995; Christensen et al., 1998). While the standardization of web design principles is still not complete, each year of Internet time has brought increased understanding of web design principles and improved customer conversion rates (Nielsen, 2001; Nielsen and Tahir, 2002; Pastore, 2000a). Third, in e-commerce, customer needs and expectations when shopping online often have not been well understood, especially by early entrants. Only in recent years

3 1032 S. Min, M. Wolfinbarger / Journal of Business Research 58 (2005) have academics and market researchers started to conceptualize and measure the factors that lead to customer judgments of satisfaction and quality when buying online (see Zeithaml et al., 2002; Wolfinbarger and Gilly, 2003, for reviews). In addition to the fact that the online environment has unique selling challenges, early movers who were pure-plays often did not understand customer needs because they lacked previous retail experience altogether (Grewal et al., 2004). In addition to these three plausible factors that would be associated with late rather than early mover advantages in e- commerce, low switching costs for customers on the Web may help late movers compete against early movers in e- commerce. Online buyers can easily compare prices and products and switch suppliers with just a few mouse clicks; new Web technologies and web design standardization are reducing switching costs even further (Porter, 2001). While it is an empirical question whether the advantages of early moving outweigh the disadvantages, we posit that early movers in e-commerce will leverage their brand recognition, marketing partnerships, and learning about technology and customers to obtain higher performance than later entrants will in market share, profit margins, and marketing efficiency. H1: Holding bricks and clicks and specializing constant, early online retailers will have higher market share than later entrants will. H2: Holding bricks and clicks and specializing constant, early online retailers will set higher profit margins than later entrants will. H3: Holding bricks and clicks and specializing constant, early online retailers will have higher marketing efficiency than later entrants will Pure-play versus bricks and clicks Bricks and clicks combinations have both producer- and customer-based advantages. Key producer-based advantages include prior customer base and utilization of established distribution and retail networks (Grewal et al., 2003; Silverstein et al., 2001). Customer-based advantages for bricks and clicks businesses include the reputation and trust already established by traditional retailers as well as consumer interest in the benefits offered by multichannel shopping. The reputation provided by the backing of a traditional bricks and mortar business is especially important online because customers cannot easily examine products before purchase. For consumers, bricks and clicks combinations offer the benefits of multichannel shopping. Pure-plays cannot offer prepurchase trial or experience, nor can they offer immediate gratification, except for downloadable digital products (Grewal et al., 2004). Additional advantages from integrating online and offline channels include increased shopping convenience, easier return of products purchased online, and the ability to offer deep information on products that may motivate a consumer to buy offline. Increased shopping convenience and deep information are valued highly by online consumers (Bellman et al., 1999; Jarvenpaa and Todd, 1997; Lohse et al., 2000; Wolfinbarger and Gilly, 2001). Moreover, bricks and clicks businesses can augment their land-based businesses by offering customers extended selections online (Burke, 1997), as Nike has done with their Nike ID line and Landsend.com has done with custom-tailored pants. Thus, bricks and clicks businesses can offer consumers what they want whenever they want it, the flexibility of extended selection and customization, and all with the trustworthiness of a known brand. Moreover, having a known brand also lowers the cost of customer acquisition. Boston Consulting Group reported that multichannel retailers paid US$38 in 1999 to acquire each new customer, while pure-plays spent US$82 (Pastore, 2000b). Thus, we predict that bricks and clicks combinations will garner greater performance than do pure-plays in market share, profit margin, and marketing efficiency. H4: Holding entry timing and specializing constant, bricks and clicks will have higher markets share than pure-plays will in online retailing. H5: Holding entry timing and specializing constant, bricks and clicks will set higher profit margins than pure-plays will in online retailing. H6: Holding entry timing and specializing constant, bricks and clicks will have greater marketing efficiency than pureplays will have in online retailing Specialists versus generalists The strategic advantage of generalizing is the ability to leverage a company s reputation and investment in branding, website design, and fulfillment capabilities across a number of product categories, as has e-commerce innovator Amazon.com. Many commentators originally touted the low barriers to entry in the Internet space due to the ease of setting up a website compared with building a physical store. However, the barriers have turned out to be higher than predicted (Barsh et al., 2000). In sum, generalists can leverage their brand name and spread the costs of warehousing and website development and maintenance across a broader product mix. On the other hand, specialists can offer products online that are often not easily available offline. Specialists have the ability to offer deep selection by aggregating marketplaces that would be too small to support a local business. Importantly, consumers report that one of their motivations for shopping online is to find an increased selection of products (Wolfinbarger and Gilly, 2001). Specialists are not

4 S. Min, M. Wolfinbarger / Journal of Business Research 58 (2005) likely to garner the sales that are possible for generalists, nor is it their strategy to do so. Nevertheless, specialists offer some of the most interesting possibilities for e-tailing s future (Grewal et al., 2004). In particular, offering a clearer value proposition (Porter, 2001) to a niche marketplace is likely to result in the ability of specialized sites to draw higher margins than do generalists. How will the marketing efficiency of specialists and generalists compare? Marketing efficiency, especially as reflected in the burn rate or cost of acquiring a new customer, was problematic for online businesses in the years that our analysis covers. Offering a general product assortment, as does Amazon.com or JCPenney.com, leverages the reputation of a company and results in increased opportunities to garner repeat purchases and revenue from the same customer base. Offering a narrower line of services makes it harder to build and leverage reputation (Nayyar, 1990); yet, the intangible nature of buying online means that reputation is important to consumers. On the other hand, search engines, shopping bots, specialized portals, and sites that aggregate smaller businesses (such as Yahoo.com) are making it increasingly easy for customers to find specialized products (Grewal et al., 2004); these online marketing vehicles may result in relatively efficient marketing for specialists. But one challenge remains for these specialists. Trust has proven to be especially important in creating sales online (Reichheld and Schefter, 2000; Urban et al., 2000), and generalists are more likely to build well-known brands because they sell products in multiple categories. Thus, on balance, we predict that each marketing dollar spent will make more sales for a generalist than for a specialist. We offer the following three hypotheses based upon our discussion: H7: Holding entry timing and bricks and clicks constant, online generalists will have higher market share than online specialists will. H8: Holding entry timing and bricks and clicks constant, online specialists can set higher profit margins than online generalists will. H9: Holding entry timing and bricks and clicks constant, online generalists will have higher marketing efficiency than do online specialists. 3. Data 3.1. Sample There are 2233 businesses with IPOs in the United States between May 1996 and June 2000 in Hoovers database. According to the database, the first e-business that went public was Yahoo.com, in May Each IPO was screened to identify online retailers based on Hoovers description of the business. This process yields 42 online retailers in 16 product categories, including books, music, travel and leisure, health, drugs and beauty (HBAD), food and beverage, computer and electronics, toys, pet products, clothing, cars and parts, and office supplies. Because the data cover public firms, all annual and quarterly financial information is publicly available (e.g., This makes the research feasible. The data cover major online retailers such as Amazon.com, Barnesandnoble.com, Buy.com, Peopod.com, and Travelocity.com in the United States. Yet, our data exclude all web media sites such as Yahoo.com, AOL, and MSN.com. Although some media sites sell both products and services, their business models are different from that of e-tailers. Most media sites aggregate their revenues from sales and various service fees; thus, revenues due only to e-tailing cannot be tracked utilizing publicly available data Key variables The lag time of entry is used to measure early mover advantage in e-commerce. It is the time lag between the e- tailer s entry and the first public e-tailer s entry in the market. The entry dates are when the online retailers first began selling products through the Internet. In most cases, the date of online retailing entry is reported in the company s annual report (10Ks) or IPO proposal (S1s). Because IPO proposals report the revenue history of a business since its inception, we cross checked the start dates by verifying that the online retailer actually made sales in the quarter that they first claimed to do business. A business is identified as bricks and clicks when the online retailer or its parent company owns bricks and clicks stores. An increasing number of retailers have become bricks and clicks combinations in recent years. However, when the business aggregates the revenues from their offline operations with the revenues from online retailing in their reports, they were excluded from our sample. Four area experts evaluated each online retailing business to rate the degree to which an online business was a specialist versus a generalist. The experts relied on their knowledge, visits to the website of e-tailers, and/or read investor s business descriptions, and annual reports. A ninepoint scale was used, with generalist and specialist serving as anchors for the scale. A specialist is defined as an e-tailer that carries a narrow product line with a deep assortment within the line, just like specialty stores do in offline retailing. A generalist carries a wide product mix with varied product lines, as do department stores in offline retailing. Ratings for the experts were then averaged for each e-tailer; this average score was used in the analysis to indicate the degree to which an online retailer was a specialist or a generalist. To measure market share, the value of annual sales revenue of each e-tailer is divided by category sales. Annual

5 1034 S. Min, M. Wolfinbarger / Journal of Business Research 58 (2005) category sales in 1999 and 2000 are available in E-Stats by U.S. Census Bureau. We use E-Stats product classification to define market boundaries. Because four e-tailers (Amazon.com, Barnesandnoble.com, Value America, and Buy.- com) sold multiple product categories in , their market boundaries are broadly defined accordingly. For instance, the market boundary of Barnseandnoble.com includes not only books but also CDs and videos. Because the marketing efficiency and profit margin vary across product categories, we use relative performance measures of marketing efficiency and margin. Marketing efficiency is thus measured as the marketing efficiency of the focal e-tailer divided by the average marketing efficiency of e-tailers in their product category. Similarly, relative margin is the margin of the focal e-tailer divided by the average margin of e-tailers in the product category. 4. Repeated-measures model Analysis of variance (ANOVA) has been often used for analyzing repeated-measures data in various behavioral and social sciences (see the brief review by Girden, 1992, pp. 4 6). The most important limitation of it is the requirement of balanced data. Our repeated-measures data are unbalanced due to business failures that occurred within the time period of our observation window. While most e-tailers survived for year 1999, by the fourth quarter of the second year (2000), about 25% of public e-tailers had exited the market. We use a multilevel repeated-measures model to analyze data (Goldstein, 1987; Kenny et al., 2002). The model can easily handle unbalanced repeated-measures data and designs with continuous predictor variables. Let Market Share ij represent the market share of e-tailer j in the ith period. Then, we can express the variation in Market Share ij using a pair of linked models: within-e-tailer (the Level 1 model) and between-e-tailers models (the Level 2 model). At Level 1, we express e-tailer j s market share in the ith period as follows: Market Share ij ¼ p 0j þ p 1j Marketing Expense ij þ p 2j Startup Period ij þ r ij ; ð1þ where r ij f N(0,r 2 ). The market share variation within e-tailer is modeled as a linear function of the marketing expense spending and a dummy variable indicating a startup period. Investing more dollars in marketing should help an e-tailer obtain higher market share. Urban et al. (1986) and other studies find that increasing promotion and advertising expenditures increases market share. A dummy for two years of the startup period is included to control for the unusually low level of sales in that period. We assume that the random error is distributed normally with mean 0 and standard deviation r. At Level 2 (between e-tailers), we express the e-tailerlevel intercepts as Eq. (2). p oj ¼ b 00 þ b 01 Ln Lag Time j þ b 02 Bricks and Clicks j þ b 03 Specializing j þ b 04 Number of Employees j þ u 0j ð2þ where u 0j f N(0,s 00 ). The hypothesis testing variables and a variable for the number of employees are included in the between-e-tailers model. See Table 1 for the definitions and the expected signs of the variables. As market share is a relative measure, the explanatory variables should also be relative measures (Urban et al., 1986; Robinson and Fornell, 1985; Robinson, 1988). Note that the natural logarithm transformation is used for entry lag-time to capture the diminishing marginal returns. The number of employees is included in our models to control for the effect of firm size. Larger firms with more resources should obtain higher market share than smaller firms. b 00 is a common intercept for all e-tailers, and u 0j represents between-group variation, which is assumed to follow a normal distribution with mean 0 and variation s 00. We do not include random errors in the slopes of the marketing expense variable and the startup period variable due to the small number of repeated measures within an e- tailer. That is, we let p 1j = b 10 and s 2j = b 20. Substituting Eq. (2) into Eq. (1) yields the multilevel model as follows: Market Share ij ¼ b 00 þ b 01 Ln Lag Time j þ b 02 Bricks and Clicks j þ b 03 Specializing j þ b 04 Number of Employees j þ b 10 Marketing Expense Share ij þ b 20 Startup Period ij þ u 0j þ r ij ð3þ where u 0j f N(0,s 00 ), and r ij f N(0,r 2 ). The model includes two random effects. That is, s 00 is the measure of unexplained variation in market share that occurs between e-tailers and r 2 represents the unexplained variation in market share that occurs within e-tailers across repeated measures. The models predicting relative margin and marketing efficiency are specified similarly to the model for market share. The relative margin for e-tailer j at time i is shown in Eq. (4). Relative Margin ij ¼ p 0j V þ p 1j V Ln t ij where rv ij f N(0,rV 2 ). þ p 2j V Startup Period ij þ r ij V ð4þ

6 S. Min, M. Wolfinbarger / Journal of Business Research 58 (2005) (10.81) Table 1 Variable definitions and expected signs Variable Definition Mean (S.D.) Expected signs Market share Relative margin Relative marketing efficiency Dependent variables Market share The percentage value of annual sales of the e-tailer divided by category sales. a Relative margin The margin of the e-tailer is divided by the average margin of the product category. The margin of the e-retailer is (quarterly sales quarterly COGS)/quarterly sales (0.66) Relative marketing efficiency The marketing efficiency of e-tailer divided by the average marketing efficiency of product category. The marketing efficiency is the value of quarterly sales divided by marketing and selling expenses of the e-retailers (0.64) Hypothesis testing variables Ln Lag time The natural logarithm of one plus the time lag in months between the e-tailer s entry and the first public e-tailer s entry in the market boundary. Bricks and clicks 1 if the e-retailer or parent company owns Bricks and Clicks stores, 0 otherwise. Specializing The degree of e-tailer s focus on its core product categories only is divided by the average value of product category (1.64) (H1) (H2) (H3) 0.14 (0.35) + (H4) + (H5) + (H6) 1.03 (0.32) (H7) + (H8) (H9) Control variables Marketing The percentage value of annual marketing and 5.43 (6.44) + Not included Not included expense share selling expenses divided by category sales. Number of The number of employees of the e-tailer divided 0.97 (0.73) + +/ + Employees by the average value of the product-category. Ln t The natural logarithm of the number of quarters 1.14 (0.67) Not Included + + since Startup period 1 if it is within two years after e-tailer s entry, (0.48) otherwise. Note that the means and standard deviations for market share, marketing expense share are based on annual data and includes 80 data points. The others are based on quarterly data and include 264 data points. a The category sales figures are from E-Stats by U.S. Census Bureau. To capture whether the profit margins of e-tailers have been increasing over time, Ln t is included in the Level 1 model. Ln t is the natural logarithm of the number of quarters since We take the logarithm to capture the diminishing return of the variable. A dummy variable for two years during the startup period attempts to control for abnormally low profit margins in the startup period. The between-group difference in the intercept term is modeled the same as the market share model. p oj V ¼ b 00 V þ b 01 V Ln Lag Time j þ b 02 V Bricks and Clicks j þ b 03 V Specializing j Most of the growth models for repeated-measures data allow random errors in the slope of time period variables (see the literature review and modeling details in Raudenbush, 2002). We allow random errors in the slope of Ln t, assuming p 1j V = b 10 V +uv 1j, where and let p 2j V = b 20 V. Substituting Eq. (5) into Eq. (4) yields the multilevel model as follows. Relative Margin ij ¼ b 00 V þ b 01 V Ln Lag Time j þ b 02 V Bricks and Clicks j þ b 03 V Specializing j þ b 04 V Number of Employees j þ b 04 V Number of Employees j þ u 0j V ð5þ þ b 10 V Lnt ij þ b 20 V Startup Period ij þ u 0j V þ u 1j V Lnt ij þ r ij V ð6þ where u 0j V f N(0,s 00 V ). where u 0j V f N(0,s 00 V ), u 1j V f N(0,s 11 V ), and r ij Vf N(0,rV 2 ).

7 1036 S. Min, M. Wolfinbarger / Journal of Business Research 58 (2005) In a similar way, we specify a multilevel repeatedmeasures model for marketing efficiency of e-tailer j at time i in Eq. (7). Marketing Efficiency ij ¼ b 00 W þ b 01 W Ln Lag Time j þ b 02 W Bricks and Clicks j þ b 03 W Specializing j þ b 04 W Number of Employees j þ b 10 W Lnt ij þ b 20 W Startup Period ij þ u 0j þ u 1j Lnt ij þ r ij where uw 0j f N(0,sW 00 ), uw 1j f N(0,sW 11 ), and rw ij f N(0,rW 2 ). Larger firms, as measured by the number of employees, should have better marketing efficiency than smaller firms have because their marketing dollars are better leveraged. In general, as customers become more familiar and comfortable with online shopping, e-tailing marketing dollars are expected to become more effective. This line of reasoning implies a positive sign for Number of Employees. A dummy variable for the startup period is included to account for the inefficiency of marketing efforts in the startup period. Two years (1999 and 2000) of annual data are used to estimate the parameters in the market share model. Quarterly data are used for the profit margin and marketing efficiency models. From 1999 to 2000, online sales and marketing efficiency increased; the average cost of customer acquisition fell from US$38 to $29 (Greenspan, 2002). 5. Results The full information maximum likelihood method is used to estimate the parameters in the repeated-measures models. Table 2 reports the maximum likelihood results for market share models. Column I in Table 2 shows the results (7) without the hypothesis testing variables. All three coefficients for the control variables have the expected signs. In the startup period, market share is lower. Increasing the marketing and selling expenses helps increase market share significantly. As well, larger e-tailers tend to have higher market share than smaller ones do. The estimated value of s 00 is 42.10, and the estimated value of r 2 is This suggests not only that there is significant variation of market share within an e-tailer in 1999 and 2000 but also that e- tailers do differ in the average values of their market shares. Note that the between-e-tailers variance is more than two times larger than the within-e-tailer variance. As the hypothesis testing variables are added to the base model, the estimates for the between-e-tailer variation, s 00, becomes smaller. The coefficients for the hypothesis testing variables are reported in the columns labeled II, III, IV, and V. Adding the Bricks and Clicks variable to the base model improves the model s fit. The chi-square in the log likelihood ratio test, 11.00, indicates a significant improvement in the model fit. But, adding Ln Lag Time and Specializing does not improve the model fitness significantly. The chi-square statistics are 0.40 and 2.4, respectively. Hypothesis testing is based on the results for the full model specification shown in the column V. Adding all three testing variables improves the estimation fit significantly. The coefficients of Ln Lag Time are not significant, indicating that early e-tailers do not have higher market share than do later entrants. H1 is not supported. The positive coefficients for Bricks and Clicks indicate higher market share for bricks and clicks combinations as compared with pure-plays. H4 is supported. The negative signs for Specializing support H7. Generalizing helps increase market share more than specializing in e-commerce does. Table 3 reports the results for the relative margins of e-tailers. Column I reports the results for the model specification without the hypothesis testing variables. The positive coefficient of Ln t indicates that the profit margin of e-tailers increased significantly during 1999 Table 2 Repeated-measure model results explaining market share of e-tailers (n = 80) Variable I II III IV V Constant 0.50 (0.26) 1.88 (0.70) 1.27 ( 0.71) 6.77 (1.65) 5.74 (1.24) Ln lag time 0.54 ( 0.74) 0.18 ( 0.26) Bricks and clicks (3.57)*** (3.42)*** Specializing 5.67 ( 1.70) * 5.88 ( 1.94)* Marketing expense share 0.61 (4.05)*** 0.58 (3.74)*** 0.60 (4.27)*** 0.64 (4.28)*** 0.62 (4.33)*** Startup period 4.29 ( 2.71)*** 4.11 ( 2.58) ** 3.56 ( 2.32) ** 4.47 ( 2.85)*** 3.68 ( 2.42) ** Number of employees 6.11 (3.99)*** 5.84 (3.73)*** 6.35 (4.67)*** 5.44 (3.54)*** 5.55 (3.89)*** Log likelihood Chi-square *** *** s (3.62)*** (3.61)*** (3.37)*** (3.55)*** (3.26)*** r (4.37)*** (4.36)*** (4.39)*** (4.36)*** (4.40)*** * Statistic in the parentheses is significant at the 10% level. ** Statistic in the parentheses is significant at the 5% level. *** Statistic in the parentheses is significant at the 1% level.

8 S. Min, M. Wolfinbarger / Journal of Business Research 58 (2005) Table 3 Repeated-measure model results explaining the relative margin of e-tailers (n = 267) Variable I II III IV V Constant 0.83 (5.09)*** 0.99 (4.75)*** 0.79 (4.79)*** 0.31 (0.96) 0.36 (0.90) Ln lag time 0.06 ( 1.19) 0.02 ( 0.37) Bricks and clicks 0.25 (0.98) 0.22 (0.84) Specializing 0.48 (1.89) * 0.45 (1.71) * Ln t 0.11 (2.67)*** 0.11 (2.63)*** 0.11 (2.67)*** 0.10 (2.60)*** 0.10 ( 2.59) ** Startup period 0.13 ( 2.10) ** 0.13 ( 2.09) ** 0.13 ( 2.06) ** 0.14 ( 2.18) ** 0.14 ( 2.12) ** Number of employees 0.05 ( 0.47) 0.09 ( 0.81) 0.05 ( 0.44) 0.00 ( 0.01) 0.01 ( 0.12) Log likelihood Chi-square * 4.6 sv (4.05)*** 0.48 (3.89)*** 0.50 (4.07)*** 0.48 (4.03)*** 0.47 (4.00)*** s 01 V 0.09 ( 2.71)*** 0.09 ( 2.63)*** 0.09 ( 2.73)*** 0.09 ( 2.73)*** 0.09 ( 2.71)*** s 11 V 0.04 (2.85)*** 0.04 (2.85)*** 0.04 (2.85)*** 0.04 (2.85)*** 0.04 (2.85)*** rv (9.71)*** 0.05 (9.71)*** 0.05 (9.71)*** 0.05 (9.71)*** 0.05 (9.71)*** * The t statistic in the parentheses is significant at the 10% level. ** The t statistic in the parentheses is significant at the 5% level. *** The t statistic in the parentheses is significant at the 1% level in the United States. The coefficient for the startup period dummy variable is statistically significant, meaning that e-tailers have significantly lower margins during their startup period. The estimate of s 00 V ( = 0.51) in Table 3 indicates the differences across e-tailers in the initial status. This between-e-tailer variation is about 10 times greater than the within-e-tailer variation (rv 2 = 0.05). The estimate of s 11 V ( = 0.04) is the variability of margin growth across e- tailers, indicating that different e-tailers improve their profit margins at different rates. The covariance between initial status (intercept) and growth (slope) is significantly negative (s 01 V = 0.09). This indicates that e-tailers with lower profit margins improve at a faster rate on average than e-tailers with higher initial margin do. Adding the three predictive variables to the base model barely decreases the random errors in the intercepts. The estimates for s 00 V s values decrease marginally. The chisquare fitness test comparing the base model with the fully specified model is not significant. The three strategic factors appear to be unimportant in predicting margin. The coefficient of Ln Lag Time is not significant. H2 is not supported. Early movers in e-commerce do not earn higher profit margins than later entrants do. The coefficient for bricks and clicks is not significant either; bricks and clicks companies do not set higher margins than do pure-plays. Thus, H5 is not supported. The positive coefficient for specializing supports H8; online specialists appear to have higher margins than do online generalists. The relative marketing efficiency of e-tailers is explained in Table 4. The base model I in the table indicates that marketing efficiency is significantly lower in the first two years after an e-tailer s entry. The coefficient for the lag time variable, shown in column V, is not significant. Thus, H3 is not supported; early entry does not improve marketing efficiency. However, the coefficients for the bricks and clicks and specializing variables possess the expected signs; H6 and H9 are supported. Bricks and clicks and generalists have higher marketing efficiency than do pure-plays and specialists. Table 4 Repeated model results explaining the relative marketing efficiency of e-tailers (n = 267) Variable I II III IV V Constant 0.52 (3.79)*** 0.58 (3.27)*** 0.44 (3.19)*** 1.14 (3.94)*** 1.21 (3.50)*** Ln lag time 0.03 ( 0.55) 0.03 ( 0.69) Bricks and clicks 0.48 (1.51) ** 0.41 (1.93)* Specializing 0.53 ( 2.37) ** 0.58 ( 2.59) ** Ln t 0.06 (1.40) 0.06 (1.42) 0.06 (1.51) 0.05 (1.17) 0.05 (1.27) Startup period 0.23 ( 3.40)*** 0.22 ( 3.36)*** 0.21 ( 3.21)*** 0.25 ( 3.75)*** 0.23 ( 3.56)*** Number of employees 0.15 (1.45) 0.13 (1.22) 0.16 (1.65) 0.08 (0.79) 0.06 (0.60) Log likelihood Chi-square ** 5.0 ** 10.2 ** sw (3.71)*** 0.21 (3.71)*** 0.19 (3.65)*** 0.17 (3.46)*** 0.15 (3.33)*** sw ( 0.02) 0.00 (0.11) 0.00 ( 0.18) 0.00 (0.03) 0.00 (0.02) sw (2.69)*** 0.04 (2.70)*** 0.04 (2.71)*** 0.04 (2.69)*** 0.04 (2.71)*** rw (9.74)*** 0.06 (9.74)*** 0.06 (9.76)*** 0.06 (9.73)*** 0.06 (9.74)*** * The t statistic in the parentheses is significant at the 10% level. ** The t statistic in the parentheses is significant at the 5% level. *** The t statistic in the parentheses is significant at the 1% level.

9 Discussion S. Min, M. Wolfinbarger / Journal of Business Research 58 (2005) Conclusion Our study finds that earlier entrants have neither higher market share, nor higher margins, nor higher marketing efficiency than do later entrants. The insignificant early mover advantages may be, in part, attributed to low consumer switching costs in e-commerce, given that competition is literally a mouse-click away. The frequent practice of outsourcing website design and supply management or buying standardized packages again increases homogeneity among competitors; thus, technological advantages of early movers may be minimal (Porter, 2001). The fact that early entrants fared no better on average than did later entrants is likely also due to the fact that many early entrants did not understand the key importance of reliable fulfillment and usable websites (Wolfinbarger and Gilly, 2003). This knowledge began to be widely disseminated over the time period of our research, and early and later entrants were apparently equally able to put this knowledge to use. Bricks and clicks advantages are significantly greater than that of pure-plays on two of the three outcome variables we studied. Bricks and clicks combinations resulted in higher market share and higher marketing efficiency than did pure-plays. In 1999, traditional businesses rushed online, hoping to prevent being Amazoned. Our data suggest that their efforts have largely been successful in terms of competing with pure-plays. In particular, bricks and clicks have higher marketing efficiency than do pure players. This finding is consistent with the idea that offline players can leverage the trustworthiness of their brand in the online environment. Interestingly, bricks and clicks businesses do not presently make higher margins than do pure-plays. However, our analysis does not include the value of a website to the offline sales of bricks and clicks businesses; academics and consulting firms have suggested that websites create many more offline than online sales (Wolfinbarger and Gilly, 2001). The performance metrics of specialists clearly reveal a different pattern than that of generalists. Specialists sell less, but set higher profit margins than do generalists. This finding indicates the feasibility of niche strategies in e- tailing. A weakness of the public firm data should be noted. The data exclude private firms. Private online businesses are typically smaller than public online retailers are. If only successful early online businesses could go public, which is a plausible guess, the early movers in our data should be the strongest and most successful online retailers. Moreover, a later entrant would have to be as successful, or even perhaps more successful than early movers to go public. Thus, any potential bias toward early mover advantages should be minimal, if there is a bias at all. Additionally, private online retailers should be more likely to specialize in one product category than public online retailers do. If so, our data overrepresent generalists as compared with specialists. Who prevails in Internet commerce? The immediate response might be that no one does. Indeed, the gross dollar margin minus marketing and selling expenses for an average public online retailer in the United States is negative for 1999 and 2000 in our sample. However, there is some good news: Profit margin improves steadily over time. Moreover, our research suggests that strategic decisions other than entry timing affect online e-tailing revenue, profit margin, and marketing efficiency. The study finds no evidence for early mover advantages in Internet commerce. Bricks and clicks have clear advantages over pure-plays. The advantages of brand strength, cross-promotional opportunities, and the ability to offer consumers multichannel shopping results in bricks and clicks being more successful than are pure-plays. Last, both generalists and specialists can be successful. Generalists can leverage their product name across multiple product lines, thus creating greater sales revenue and marketing efficiency, while specialists can charge higher profit margins for their products. Acknowledgements The authors thank Mary C. Gilly, Juyoung Kim, and the anonymous reviewers for providing comments. References Barsh J, Crawford B, Grosso C. How e-tailing can rise from the ashes. McKinsey Q 2000;3: Bellman S, Lohse GL, Johnson EJ. Predictors of online buying: findings from the wharton virtual test market. Commun ACM 1999 [December] pp Bohlmann JD, Golder PN, Mitra D. Deconstructing the pioneer s advantage: examining vintage effects and consumer valuations of quality and variety. Manage Sci 2002;48(9): Bughin J, Zeisser M. There is light at the end of the tunnel: profitable strategies in online retailing. MIT online e-commerce forum; 2001 [Working Paper, id = 174]. Burke RR. Do you see what I see? The future of virtual shopping. J Acad Mark Sci 1997;25(4): Christensen CM, Suarez FF, Utterback JM. Strategies for survival in fastchanging industries. Manage Sci 1998;44(12):S Freshtman C, Mahajan V, Muller E. Market share pioneering advantage: a theoretical approach. Manage Sci 1990;36(8): Girden ER. ANOVA repeated measures. Series: quantitative applications in the social sciences. Newbury Park, CA: Sage Publications; Golder PN, Tellis GJ. Pioneer advantage: marketing logic or marketing legend? J Mark Res 1993;30: Goldstein H. Multilevel models in educational and social research. New York: Oxford Univ. Press; Grewal D, Iyer G, Levy M. Internet retailing: enablers, limiters and market consequences. J Bus Res 2004;57(7): Greenspan R. Online Consumer Confidence, Spending Grows. Available at: 0,6061_ ,00.html. October Jarvenpaa SL, Todd PA. Consumer reactions to electronic shopping on the World Wide Web. Int J Electron Commer 1997;I(2):59 88.

10 S. Min, M. Wolfinbarger / Journal of Business Research 58 (2005) Jarvenpaa SL, Tractinsky N. Consumer trust in an Internet store: a crosscultural validation. J Comput-Mediat Commun 1999;5(2) [Available at: Kalyanaram G, Urban GL. Dynamic effects of the order of entry on market share, trial penetration, and repeat purchases for frequently purchased consumer goods. Mark Sci 1992;11(Summer): Kenny DA, Bolger N, Kashy DA. Traditional methods for estimating multilevel models. In: Moskowitz DS, Hershberger SL, editors. Modeling intraindividual variability with repeated measures data: methods and applications. Mahwah, New Jersey: Lawrence Erlbaum Associates; p Lieberman MB, Montgomery DB. First-mover advantages. Strateg Manage J 1988;9: Lohse G, Bellman S, Johnson EJ. Consumer buying behavior on the Internet: findings from panel data. J Interact Market 2000;14(1): Nayyar PR. Information asymmetries: a source of competitive advantage for diversified service firms. Strateg Manage J 1990;11: Nielsen J. Did poor usability kill e-commerce? Available at: Nielsen J, Tahir M. Homepage usability: 50 websites deconstructed. Indianapolis (IN): New Riders; Pastore M. Q4 still the big payoff for e-commerce. Available at: cyberatlas.internet.com/markets/retailing/print/0,6061_478001,00.html. October 5, 2000a. Pastore M. Maybe the end can wait. Available at: com/markets/retailing/print/0,6061_343161,00.html. April 18, 2000b. Porter ME. Strategy and the Internet. Harvard Bus Rev. 2001;63 78 [March]. Raudenbush SW. Alternative covariance structures for polynomial models of individual growth and change. In: Moskowitz DS, Hershberger SL, editors. Modeling intraindividual variability with repeated measures data: methods and applications. Mahwah, New Jersey: Lawrence Erlbaum Associates; p Reichheld FF, Schefter P. E-loyalty: your secret weapon on the web. Harvard Bus Rev 2000; [July August]. Robinson WT. Sources of market pioneer advantages: the case of industrial goods industries. J Mark Res 1988;25(February): Robinson WT, Fornell C. Sources of market pioneer advantages in consumer goods industries. J Mark Res 1985;22(August): Silverstein M, Stanger P, Abdelmessih N. The next chapter in business to consumer e-commerce: advantage incumbent. Boston: The Boston Consulting Group; Suarez FF, Utterback JM. Dominant designs and the survival of firms. Strateg Manage J 1995;16: Urban GL, Sultan F, Qualls WJ. Placing trust at the center of your Internet strategy. Sloan Manage Rev 2000;42(1): Urban GL, Carter T, Gaskin S, Mucha Z. Market share rewards to pioneering brands: an empirical analysis and strategic implications. Manage Sci 1986;32(June): Wolfinbarger M, Gilly M. Shopping online for freedom, control and fun. Calif Manage Rev 2001;43(2): Wolfinbarger M, Gilly M. etailq: dimensionalizing, measuring and predicting etail quality. J Retail 2003;79: Zeithaml V, Parasuraman A, Malhotra A. Service delivery through web sites: a critical review of extant knowledge. J Acad Mark Sci 2002; 30(4): Zhu K, Kraemer K. E-commerce metrics for net-enhanced organizations: assessing the value of e-commerce to firm performance in the manufacturing sector. Inf Syst Res 2002;13(3):