1424 Testing, Measuring, and Diagnosing Web Sites from the User’s Perspective APIEMS2004 Web page breakdown Item URL 1 http://www.math au/apiems2004/ 2 www.maths.qut.e ages/spacer.gif 3 www.maths.qut.e ges/geninfo.GIF 4 www.maths.qut.e kground_new.png 5 www.maths.qut.e 2004/mm_menu.js 6 www.maths.qut.e ges/sciprog.GIF 7 www.maths.qut.e ges/accprog.GIF 8 www.maths.qut.e s/Abstracts.GIF 9 www.maths.qut.e es/regaccom.GIF 10 www.maths.qut.e s/sponexhib.GIF 11 www.maths.qut.e images/comm.GIF 12 www.maths.qut.e ages/title2.png 13 www.maths.qut.e s/asor_logo.png 14 www.maths.qut.e es/qut_logo.png 15 www.maths.qut.e ges/address.png 16 www.maths.qut.e s_rightside.png 17 www.maths.qut.e r2_pictures.png Measured Components Item Type Start Offset DNS Connect Redirect First Byte Content Bytes Sec. Sec. Sec. Sec. DL Sec. Downloaded 1 Base Page - 1.05 0.35 * 0.35 0.77 12,076 2 image/gif 2.59 * 0.34 * 0.37 * 43 3 image/gif 3.31 * 0.4 * 0.35 * 1584 4 ima g e/ p n g 2.59 * 0.34 * 0.36 1.28 27400 5 application/x- javascript 2.58 * 0.34 * 0.35 1.42 30737 6 image/gif 4.09 * 0.34 * 0.35 * 1281 7 image/gif 4.57 * 0.36 * 0.36 * 1870 8 image/gif 4.7 * 0.34 * 0.37 * 1337 9 ima g e/ g i f 4.8 * 0.34 * 0.38 * 1919 10 ima g e/ g i f 5.3 * 0.35 * 0.39 * 1638 11 ima g e/ g i f 5.41 * 0.34 * 0.37 * 1420 12 2.59 3004 B y te Limit Exceeded 13 6.13 3004 B y te Limit Exceeded 14 7.64 3004 B y te Limit Exceeded 15 8.45 3004 B y te Limit Exceeded 16 6.06 3004 Byte Limit Exceeded 17 5.55 3004 Byte Limit Exceeded Average 0.06 0.36 - 0.37 2.04 78038 Total 18.59 1,326,644 * Either the item was not necessary (such as redirection, where there are no redirects) or the operation took < 0.01 seconds. Figure 3. Instant measurement of www.maths.qut.edu.au/apiems2004/ from Frankfurt DTAG performed by MyKeynote on 23-Jul-04 2:26:14 PM EDT (User: Demo) 1425 Testing, Measuring, and Diagnosing Web Sites from the User’s Perspective APIEMS2004 Web page breakdown Item URL IP address Port Bytes 0 http://www.maths.qut.edu.au/apiems2004/ 131.181.53.23 1512 12076 1 http://www.maths.qut.edu.au/apiems2004/mm_menu.js 131.181.53.23 1513 30737 2 http://www.maths.qut.edu.au/apiems2004/images/background_new.png 131.181.53.23 1512 27400 3 http://www.maths.qut.edu.au/apiems2004/images/title2.png 131.181.53.23 1513 164625 4 http://www.maths.qut.edu.au/apiems2004/images/mmmenu6_159x18_up.gif 131.181.53.23 1512 261 5 http://www.maths.qut.edu.au/apiems2004/images/mmmenu5_245x18_up.gif 131.181.53.23 1512 287 6 http://www.maths.qut.edu.au/apiems2004/images/mmmenu4_248x18_up.gif 131.181.53.23 1512 288 7 http://www.maths.qut.edu.au/apiems2004/images/mmmenu3_172x18_up.gif 131.181.53.23 1512 266 8 http://www.maths.qut.edu.au/apiems2004/images/mmmenu2_88x18_up.gif 131.181.53.23 1512 238 9 http://www.maths.qut.edu.au/apiems2004/images/mmmenu1_221x18_up.gif 131.181.53.23 1512 280 10 http://www.maths.qut.edu.au/apiems2004/images/spacer.gif 131.181.53.23 1512 43 11 http://www.maths.qut.edu.au/apiems2004/images/geninfo.GIF 131.181.53.23 1512 1584 12 http://www.maths.qut.edu.au/apiems2004/images/sciprog.GIF 131.181.53.23 1512 1281 13 http://www.maths.qut.edu.au/apiems2004/images/accprog.GIF 131.181.53.23 1512 1870 14 http://www.maths.qut.edu.au/apiems2004/images/Abstracts.GIF 131.181.53.23 1512 1337 15 http://www.maths.qut.edu.au/apiems2004/images/regaccom.GIF 131.181.53.23 1512 1919 16 http://www.maths.qut.edu.au/apiems2004/images/sponexhib.GIF 131.181.53.23 1512 1638 17 http://www.maths.qut.edu.au/apiems2004/images/comm.GIF 131.181.53.23 1512 1420 18 http://www.maths.qut.edu.au/apiems2004/images/ver2_pictures.png 131.181.53.23 1512 537055 19 http://www.maths.qut.edu.au/apiems2004/images/ver2_pictures_rightside.png 131.181.53.23 1513 465578 20 http://www.maths.qut.edu.au/apiems2004/images/asor_logo.png 131.181.53.23 1512 32504 21 http://www.maths.qut.edu.au/apiems2004/images/qut_logo.png 131.181.53.23 1513 4167 22 http://www.maths.qut.edu.au/apiems2004/images/address.png 131.181.53.23 1513 41410 Measured Components Item Pkt OK Pkt NOK rtype ctype DL time sec DNS DNS2 SYN CONNECT ACK2 GET FIRST INDEX CONTENT 0 9 0 S text/html 34212368 974006 6294 394020 1220 384273 944950 31507605 1 22 0 C application/ x-javascript 2427324 0 0 379175 939 398459 1648751 0 2 19 0 C image/png 1207986 0 0 0 0 397849 810137 0 3 113 0 C image/png 8030071 0 0 0 0 404991 7625080 0 4 1 0 C image/gif 390086 0 0 0 0 390086 0 0 5 1 0 C image/gif 381964 0 0 0 0 381964 0 0 6 1 0 C image/gif 382435 0 0 0 0 382435 0 0 7 1 0 C image/gif 382188 0 0 0 0 382188 0 0 8 1 0 C image/gif 381742 0 0 0 0 381742 0 0 9 1 0 C image/gif 381800 0 0 0 0 381800 0 0 10 1 0 C image/gif 381365 0 0 0 0 381365 0 0 11 2 0 C image/gif 386307 0 0 0 0 386307 0 0 12 2 0 C image/gif 1112084 0 0 0 0 1112084 0 0 13 2 0 C image/gif 384620 0 0 0 0 384620 0 0 14 2 0 C image/gif 384107 0 0 0 0 384107 0 0 15 2 0 C image/gif 384635 0 0 0 0 384635 0 0 16 2 0 C image/gif 383953 0 0 0 0 383953 0 0 17 2 0 C image/gif 383837 0 0 0 0 383837 0 0 18 369 0 C image/png 19868386 0 0 0 0 394812 19473574 0 19 320 0 C image/png 19543741 0 0 0 0 399137 19144604 0 20 23 0 C image/png 1179073 0 0 0 0 393571 785502 0 21 4 0 C image/png 387146 0 0 0 0 384619 2527 0 22 29 0 C image/png 1572654 0 0 0 0 393183 1179471 0 Figure 4. Instant measurement of www.maths.qut.edu.au/apiems2004/ from Wroclaw WASK performed by Wing on 23-Jul-04 20:15:46 Europe/Warsaw 1426 Testing, Measuring, and Diagnosing Web Sites from the User’s Perspective Transport protocols like TCP compute an esti- mate of the current RTT on each connection, but unfortunately, the results of such an estimation are not available for the user (application program). TCP uses an RTT estimate to determine how long to wait for an acknowledgment before retransmit- ting. We developed exactly the same method as one used by TCP, and RTT was estimated using the technique based on the measurements of time spacing between the SYN packet sent by the client to the server and the ACK-SYN packet sent back in the reply by the server. This is a CONNECT time as shown in Figure 1. Thanks to Wing ser- vice, we can perform the same analysis as TCP when estimates RTT. In order to estimate the average transfer rate of the TCP connection, we measured time spac- LQJEHWZHHQWKH¿UVWE\WHSDFNHWDQGWKHODVWE\WH packet of the object received by the client using that connection. Transfer rate was calculated by dividing a number of bytes transferred by the amount of time taken to transfer them. The throughput measured is the amount of pure data WUDI¿FDYDLODEOHDWWKHDSSOLFDWLRQOHYHOWKDWLV IP, TCP and HTTP headers are not included into the calculations. We monitored and traced the HTTP transac- tions that had been sent periodically by the IE client localized within our campus network during 20 weeks. Each time, our Web client requested WKHVDPH¿OHIURPDIHZGR]HQ:HEVHUYHUVDOO over the world. We chose to download the rfc1945. W[W¿OHWKDWZDVIRXQGLQVHYHUDO:HEVLWHV2XU chosen resource was large enough — it has an original size of 137,582 bytes — to estimate average transfer rate, and still was not too large to overload Internet links and Web servers. The target servers were found by the Google search engine. Among a few hundred links found, we chose 209 direct links to rfc1945 document. After preliminary tests, we decided to use, for further measurements, only 83 servers which were fully active during preliminary tests. Wing was pro- grammed to download (and monitor the HTTP transaction and measure) the rfc1945 document from every server 10 times a day. After 20 weeks of measurements, we received the database with 65,428 Web transactions. We investigated the correlation between a connection’s RTT and transfer rate to examine whether shorter-RTT connections tend to transfer more data. Based on the measurements that we y = 46456x -0,88 05 R 2 = 0,881 10 100 1000 10000 10 100 1000 1000 0 RTT [ms] Average transfer rate [Kb/s] 2. 4. 5. 9. 14. 16. 19. 21 . 24. 27. 33. 37. 40. 41. 43. 44 . 47. 51. 52. 54. 57. 61. 62. 64 . 69. 71. 72. 74. 77. 78. 81. 87 . 88. 89. 90. 101. 103. 104. 107. 10 9. 110. 111. 114. 117. 123. 124. 125. 12 6. 128. 132. 135. 136. 141. 142. 146. 15 4. 158. 160. 161. 163. 164. 166. 167. 16 8. 170. 171. 173. 174. 177. 179. 181. 18 2. 184. 187. 188. 191. 192. 193. 199. 20 0. 202. 205. 209. Figure 5. Distribution of median values of average transfer rate vs. RTT 1427 Testing, Measuring, and Diagnosing Web Sites from the User’s Perspective had analyzed so far, it was inconclusive to say that we can show such tendency for individual con- QHFWLRQVEXWZHFDQVKRZVSHFL¿FSHUIRUPDQFH behavior in the sense of a global performance FKDUDFWHULVWLFVIRXQGIRUWKHVSHFL¿FORFDWLRQRI users in the Internet, such as in our case for the Wroclaw University of Technology location. The presence of such characteristics can be discovered when we plot a graph for the median values for the average transfer rate vs. RTT for all servers in question across the period of 20 weeks. Figure 5 shows such distribution of the median values for the average transfer rate vs. RTT in double logarithmic scale for all servers in question across the period of 20 weeks. We obtained a power-law behavior where the distribution of the average transfer rate vs. RTT can be described using power law of the form y = kx D with k = 46456 and D = -0.8805. PERIODIC FULL PAGE MEASURE- MENTS — URL “MORTALITY” STUDY We are now just learning that the Web content is very volatile. Wing has been used in the study on the reliability (availability) of Web servers and the ³PRUWDOLW\´RI85/VLQWKHVHQVHRIWKHLUGLVDS- pearance in time. In our experiment, we wanted to portray the statistics of the disappearance of measured URLs. We have continued the measure- ments which were introduced in Section 5 for the next 27 weeks, collecting almost 150,000 Web transactions together. These measurements have EHHQDQDO\]HGLQWKHFRQWH[WRIWKH³PRUWDOLW\´RI 8 5 / V 2 X U P H D V X U H PH Q W V G L V F RYH U H G W K H ³ PR U W D O - ity” process of URL links as presented in Figure 6. We determined the death-rate of measured URL links as -0.06. Only about 80% of URLs available in the beginning of measurements were still valid in the end of experiment. CONCLUSION In this paper, we have presented the Wing In- ternet service, which can be used by a user to visualize and analyze how the target Web page is downloaded. In instant measurement mode, Wing KHOSVWRLGHQWLI\LQHI¿FLHQWQHWZRUNXVDJHE\WKH browser and server and helps to tune applications DQG:HESDJHVWRXVHWKHQHWZRUNHI¿FLHQWO\7KLV is done by precise data collection and exact visual- y = -0,0642x + 10 0 R 2 = 0,9689 75 80 85 90 95 100 0 14 28 42 56 70 84 98 112 126 140 154 168 182 196 210 224 238 252 266 280 294 308 322 D ay of observatio n U RL mortality[ %] Figure 6. URL “mortality” vs. day of measurement 1428 Testing, Measuring, and Diagnosing Web Sites from the User’s Perspective L ] D W LRQRIL Q VW D QW PHD VX U H P H QWW U DI ¿F7 KH UHIRU H Wing can be a good analysis tool for Web page and network application developers. Wing can also be used in periodical mea- surements. We presented the results of two such H[SHULPHQWV ,Q WKH ¿UVW VWXG\ WKH FROOHFWHG data was employed in model building of Web characteristics of WUT location. We obtained a power-law behavior where the distribution of the transfer rate vs. the round-trip time can be described using power law of the form y = kx D with k = 46456 and D = -0.8805. In the second H[SHULPHQWZHVWXGLHG85/³PRUWDOLW\´LVVXH and determined URL death-rate. REFERENCES Borzemski, L. (2002). The impact of cluster- based Web systems design on user-perceived performance. In P. Isaias (Ed.), Proceedings of the IADIS International Conference WWW/Internet 2002 (pp. 642-647), Lisbon, Portugal. Lisbon, Portugal: IADIS Press. Borzemski, L., & Nowak, Z. (2004a). WING: A Web probing, visualization and performance analysis service. In Web Engineering: 4th Interna- tional Conference, ICWE 2004, Munich Germany, Lecture Notes in Computer Science, Vol. 3140 (pp. 601-602). Berlin Heidelberg: Springer-Verlag. Borzemski, L., & Nowak, Z. (2004b). An empirical study of Web quality: Measuring the Web from Wroclaw University of Technology campus. In M. Matera & S. Comai (Eds.), Engineering Advanced Web applications (pp. 307-320). Princeton, NJ: Rinton Press. Borzemski, L., & Zatwarnicki, K. (2003). A fuzzy adaptive request distribution algorithm for cluster-based Web systems, In Proceedings of Eleventh EUROMICRO Conference on Paral- lel, Distributed and Network-based Processing (pp. 119-126), Genova, Italy. Los Alamitos: IEEE Computer Society Press. Jacobson, V. (1994). TCPdump, The protocol packet capture and dumper program. Retrieved from ftp://ftp.ee.lbl.gov/tcpdump.tar.Z Hall, J., Moore, A., Pratt, I. & Lesli, I, (2003). Multi-protocol visualization. In Proceedings of the ACM SIGCOMM 2003 Workshop on Models, Methods and Tools for Reproducible Network Research (pp. 13-22), Karlsruhe, Germany. New York: ACM Press. Luckie, M. J., McGregor, A. J., & Braun, H W. (2001). Towards improving packet probing tech- niques. ACM SIGCOMM Internet Measurement Workshop (pp. 145-150), San Francisco. New York: ACM Press. Mogul, J. (2002). Clarifying the fundamentals of HTTP. In Proceedings of the Eleventh Interna- tional Conference on World Wide Web (pp. 25-36), Honolulu, Hawaii. New York: ACM Press. Ostermann, S. (1996). TCPtrace. Retrieved from http://irg.cs.ohiou.edu/software/tcptrace/index. html Yoder, J. (2002, February 27). Better end user vis- ible Web browning performance. Presented at Intel Developer Forum, Spring 2002, San Francisco. Available at http://www.intel.com/cd/software/ products/asmona/eng/219852.htm Zhi, J. (2001). Web page design and download time. CMG Journal of Computer Resource Man- agement, 102, 40-55. ADDITIONAL INFORMATION http://www.empirix.com http://www.mykeynote.com http://www.paessler.com http://www.patrick.net http://www.softwareqatest.com 1429 Testing, Measuring, and Diagnosing Web Sites from the User’s Perspective http://www.w3schools.com http://www.websiteoptimization.com This work was previously published in International Journal of Enterprise Information Systems, Vol. 2, Issue 1, edited by E. Szczerbicki, pp. 54-66, copyright 2006 by IGI Publishing (an imprint of IGI Global). 1430 Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited. Chapter 5.3 The Effects of Digital Marketing on Customer Relationships Marko Merisavo Helsinki School of Economics, Finland ABSTRACT Viewing the use of digital channels in marketing from a customer relationship perspective offers VHYHUDOEHQH¿WVWRDPDUNHWHU%UDQGFRPPX- nication can be frequent and personalized, and GLIIHUHQWRSWLRQVIRUDGLDORJXHH[LVW7KHEHQH¿WV of dialogue include learning from and about cus- tomers, revealing their needs and interests, and being able to provide them with better and more personal service. From a customer perspective, the Internet has given power to get up-to-date information, to compare products and services more easily, and to get in touch with marketers. In a digital environment, customers increasingly use self-service and have more options to initiate action. A customer’s perceived value of using digital channels and interacting with a marketer can come in several ways; it could be money, time, information, convenience, and so forth. This chapter examines how marketers can use digital channels to develop and strengthen customer relationships. In our framework, brand communi- cation, service, personalization, and interactivity are suggested to be the key elements. INTRODUCTION: DIGITAL MARKETING AND CUSTOMER RELATIONSHIPS According to Schultz (2002), multi-channel mar- keting is just a new term with old challenges; it appears to be 21 st -century terminology for how a marketing organization makes its products and services available to customers and prospects and how the marketer determines the best choice of distribution systems and type of communication program to use. However, nowadays, the use of digital channels like the Internet, e-mail, mobile phones, and digital television in marketing has increased, and marketers need to understand how to use these channels for different purposes. For example, some marketers use them for acquiring 1431 The Effects of Digital Marketing on Customer Relationships new customers, while some focus on serving cus- tomers better. Digital marketing communication typically has been information about products, advertising on the Web, and direct marketing via e-mail and mobile. It seems like the channels and concepts of digital marketing have been used separately and operationally, while the strategic understanding and models of how and why to use different digital channels are still developing. At the same time, marketers have acknowledged the importance of cultivating customer relationships. Consequently, there is now increasing discussion about how to use digital marketing to develop customer relationships, an approach that has not been fully utilized yet. To look at the use of digital channels in mar- keting from a customer relationship perspective RIIHUVVHYHUDOEHQH¿WVWRDPDUNHWHU%UDQGFRP- munication can be frequent and personalized, and GLIIHUHQWRSWLRQVIRUDGLDORJXHH[LVW7KHEHQH¿WV of dialogue include learning from and about cus- tomers, revealing their needs and interests, and being able to provide them with better and more personal service. From a customer perspective, the Internet has given power to get up-to-date information, to more easily compare products and services, and to get in touch with marketers. Furthermore, in a digital environment, custom- ers increasingly use self-service and have more options to initiate action (e.g., give feedback, check account balances, request and personalize marketing communication). From the customer perspective, perceived value of using digital chan- nels and interacting with a marketer can come in several ways. It could be money, time, information, convenience, pleasure, entertainment, assistance, social interactivity, prestige, or something else the customer appreciates. This chapter examines how marketers can use digital channels to de- velop and strengthen customer relationships. In our framework, brand communication, service, personalization, and interactivity are suggested to be the key elements. BOOSTING CUSTOMER RELATIONSHIPS WITH DIGITAL MARKETING Central Elements of Digital Marketing :H GH¿QH digital marketing as marketing and interaction via digital channels— Internet, e-mail, mobile phones, digital TV, and so forth. Digital marketing includes communication (e.g., advertis- ing, newsletters), service, and tools for customer- managed communication and interaction (e.g., FAQs, games, etc). Both marketer and customer can initiate interaction. Digital marketing as a BRAND COMMUNICATION •Frequency •Content •Functionality SERVICE •Perceived value by a customer •Efficiency •Accessibility •Reliability CUSTOMER RELATIONSHIPS Effects on brand loyalty: •Purchases •Usage •Awareness •Attitudes •Images •Recommendation •Activity •Satisfaction and service perceptions PERSONALIZATION INTERACTIVITY o Figure 1. Central elements in boosting customer relationships with digital marketing 1432 The Effects of Digital Marketing on Customer Relationships term is more inclusive than e-marketing, because it encompasses marketing via digital channels independent of the medium. Interactive marketing, one-to-one marketing, customer relationship management (CRM), brand management, and service management literature emphasize the importance of building and manag- ing customer relationships, and each offers useful models to understand how customer relationships develop. The literature reveals certain concepts and trends that could be central elements in boosting customer relationships with digital mar- NHWLQJ)LUVWEUDQGFRPPXQLFDWLRQLVLGHQWL¿HG as an important way to create customer loyalty. Second, service is considered a central element of customer relationships, and many examples show that digital channels (i.e., the Internet) can be good tools for customer service. Third, personalization is expected to create value for customers. Fourth, interactivity is presented as one of the major ad- vantages of the Internet and other new marketing channels. We think that brand communication and service are the key elements when boosting customer relationships via digital marketing, and elements of personalization and interactivity can be mixed into them to reinforce the effects (see Figure 1). We suggest that digital marketing can have a positive effect on customers’ brand loyalty (i.e., purchases, attitudes, etc.). Three main factors in brand communication are expected to affect customer relationships: frequency (e.g., how many brand messages per week), content (e.g., promotional offers or informa- tion), and functionality (e.g., how easily customers ¿QGLQIRUPDWLRQRQD:HEVLWH%UDQGFRPPX- nication also can be personalized. For example, VHJPHQWDWLRQRUFXVWRPHUSUR¿OHVFDQEHXVHGWR create different messages for different customers, sent via their preferred channels. In addition, brand contacts can be interactive, which means that customers can search for information, make inquiries, give feedback, and engage in various other activities with marketers or each other by playing games on a brand’s Web site, chatting with other brand users, and so forth. Service that is perceived as valuable by custom- ers is expected to affect customer relationships positively. This value could be money, time, information, convenience, pleasure, entertain- PHQW DVVLVWDQFH DQG VR IRUWK (I¿FLHQF\ DQG accessibility of the service are also important, which could mean easy access to a Web site and WKHDELOLW\WR¿QGGHVLUHGVHUYLFHRSWLRQVDQG information associated with them with minimal effort. Reliability of the service is essential, as well (e.g., the technical functioning and safety of an online bank). Service also can be person- alized (i.e., by offering different service options for different customers and remembering their individual preferences of service situations). Various elements of interactivity (e.g., online chat with service personnel) can be combined with service. In the model, brand communication and ser- vice are treated as separate elements, acknowl- edging that sometimes digital communication can be considered as a service, as Brondmo (2000) suggests for permission-based e-mail, while in other cases, it may be seen as a promotion. Brand communication and service do not always have to be personalized or interactive in order to have effects on customer relationships. Adding person- alization and interactivity, however, can increase the positive effects of digital brand communication and service on customer relationships. Brand Communication Digital channels offer opportunities to keep up frequent brand communication with customers. Brand communication refers generally to all com- munication between the brand and customers. This complies with the view that the process of build- ing brands and customer relationships is much more than traditional media advertising (Aaker & Joachimsthaler, 2000). Digital marketing can be used to create brand communication that custom- ers perceive to be informative, entertaining, and interesting, and that building interactive brand 1433 The Effects of Digital Marketing on Customer Relationships contacts. Frequent communication has positive effects on customers’ brand loyalty, and customer- brand relationships are deepened (Aaker, 1996; Kapferer, 1998). As in the off-line environment, there is word- of-mouth communication in the digital environ- ment, one visible form of that being a brand community. Brand communities are important for a marketer, because brand knowledge and experi- ences are quickly spread via those communities, and new opportunities for brand communication are opened. In brand communities, customers interact with each other and brands by commu- nicating via different channels or face-to-face. In many cases, consumers and marketers jointly build brand communities. These communities may form around any brand, but it is more likely to form around brands with a strong image and a rich and lengthy history that threaten competition and are publicly consumed (Muniz & O’Guinn, 2001). Marketers are encouraged to create brand communities, because customers belonging to them are more loyal than others; they are less apt to switch brands and more motivated to provide feedback (McAlexander et al., 2002). The digital environment offers great tools for community building. Virtual communities can have greater geographical scope and narrower focus than most off-line communities (Wind et al., 2002). Nevertheless, brand communities often mix the off-line and online environments, as community members often meet and discuss both on the Web and in the real world. The im- portance of community Web pages to consumers’ day-to-day lives is demonstrated by their rich F R QW H Q W FRQW L QXRX V X S G DW H D QGPR G L ¿FDW L R Q D QG active discussion. For example, Dynamiitti.com is a Finnish brand community site that connects WKHGULYHUVDQGHQWKXVLDVWVRI¿YH-DSDQHVHFDU brands (i.e., Honda, Mazda, Mitsubishi, Toyota, and Nissan). The site offers a discussion board with various topics for each brand, as well as areas for general discussion of driving and rac- ing. Members share their views of new models, accessories, repair services, tests, and so forth. Generally, news spreads very quickly within this community. There also is a market forum, where community members can buy and sell cars and other items. This community converges virtual and physical communities, as active members regularly meet each other and join events in the bricks-and-mortar world. Brand communities represent both opportuni- ties and challenges for companies. There is an op- portunity to build an online infrastructure where brand users freely share their brand knowledge, experiences, and feelings with each other. This information can be collected easily by the mar- keter, which gives an important inside view for the customers. Active community members can be considered opinion leaders who can be used to test new ideas and concepts and, hopefully, to spread positive word-of-mouth comments. There is also an opportunity to communicate directly to the brand users. For example, a representative of the marketer can be an active member of the brand community who joins everyday discus- sions among the other members. Also, ads can be placed for a sponsored brand community site. On the other hand, a company’s ability to control the discussion of brand communities can be relatively limited or non-existing. Even on sponsored brand community sites, advertising must be kept at a minimum to keep members from being bombarded by too many commercial messages. Service The development of the Internet and other new FKDQQHOVKDVPDGHLWSRVVLEOHIRU¿UPVWRFUHDWH new services and develop interactive and rela- tionship-building contacts with their customers. To further emphasize the importance of online service, extensive survey data from more than 2,000 e-tail sites shows that price-sensitive cus- tomers may be the least loyal, whereas customer service support is the main factor that attracts repeat buying (Reibstein, 2002). Indeed, the . Web, and direct marketing via e-mail and mobile. It seems like the channels and concepts of digital marketing have been used separately and operationally, while the strategic understanding and. being a brand community. Brand communities are important for a marketer, because brand knowledge and experi- ences are quickly spread via those communities, and new opportunities for brand communication. buy and sell cars and other items. This community converges virtual and physical communities, as active members regularly meet each other and join events in the bricks -and- mortar world. Brand