First Monday

Consumers on the Web: Identification of usage patterns by Nina Koiso-Kanttila

Consumers on the Web: Identification of usage patterns by Nina Koiso-Kanttila
This article analyzes consumer behavior on the Web. The purpose is to research patterns that characterize consumer actions in this environment. The study employs Nielsen//NetRatings Internet panel data in Finland. The four-month data for 65 panelists suggest three interrelated Web usage patterns that are highlighted here. The text will outline how these conclusions were reached and present other observations.

One pattern relates to the mosaic of the Web. Most consumers visit popular sites. However, many of their own favorite sites are more specialized. Approximately four in ten of the individuals' top three sites were coded in less frequently appearing categories, and did not appear among popular site measurement records. Web usage appears to have maintained individual preference and taste variety.

The other pattern involves simultaneous presence of concentration and exploration. The familiar notion of 20-80 is employed, yet from the consumer point of view. The results point out that particularly in the case of high frequency users, a small number of sites accounts for a significant percentage of the pages viewed by an individual. At the same time, consumers can visit a large number of different sites. The data on popular Web sites are in line, showing how page view records display stronger concentration than audience measurement records.

The third pattern concerns navigation patterns from site to site. Consumers appear to use various means of navigation in a rather balanced way: links from other sites, search queries, and bookmarks. However, the percentages of these various means vary by usage intensity and age.


Research material and transferability of findings
The mosaic of the Web
Web site usage: Concentration and exploration
Navigation patterns: Opening a session
Pages in one Web session
Means of navigation from site to site





Consumers are increasingly integrating the World Wide Web into their daily life. Related research has addressed, for example, customer preferences, marketing implications of digitization, and human interaction with electronic environments. Examples can be found in Nielsen (1997), Dabholkar (2000), Windham and Orton (2000), IconMedialab (2001), Rangan and Adner (2001), Barwise and Strong (2002).

Nevertheless, research on consumer behavior directly remains less accessible. Notable exceptions include contributions by Kraut et al. (1998), Choo, Detlor, and Turnbull (2000), Rozanski, Bollman, and Lipman (2001), and also the studies reviewed in Bucklin et al. (2002). Despite these recent contributions, published longitudinal research on how consumers use the Internet in their daily life is relatively scarce. It is in this area that the present study positions itself. The purpose is to explore what characterizes consumer actions on the Web and what kind of usage patterns emerge.

The primary empirical data are four-month Web activity records of 65 individual panelists from the beginning of November 2001 to the end of February 2002. The research method and resulting material are presented in the following section. The next five sections review the findings and implications. A discussion closes the text.



Research material and transferability of findings

Methodological questions are addressed in this section. The method is observational. Observation entails the systematic noting and recording of events, behaviors and objects in the social setting chosen for the study [1]. The software-based technique provides accurate data, unobstrusively. The advantage is that insights are gained into the tasks and contexts where consumers choose their activities. Such emphatic design through observation in consumers' own environments holds potential for yielding obvious but still tremendously valuable basic information (Leonard and Rayport, 1997).

The research material is from Nielsen//NetRatings Internet panel in Finland. Nielsen//NetRatings have become the global standard in Internet audience measurement and analysis. The author is not affiliated with Nielsen//NetRatings. The panel formed a representative sample of the Internet user population in Finland. The Finnish panel was discontinued at the end of April 2002. Surveys were used in forming the panel. Consumer participation in the panel is voluntary and based on an explicit agreement that a real-time meter installed on the individuals' computers record Internet user behavior click-by-click. The Finnish panel was home computer-based.

The main Web activity data studied concern 65 panelists. They were selected by systematic sampling from an ordered list where all Finnish panelists active in the period were ranked according to their time spent online. The starting point was chosen with a random number generated within the sampling interval. As sampling is based on usage intensity, the sample was not required to be demographically fully representative. Later in the results presented Figure 4 will show how the sample covers the mass of Web behavior. The selected panelists are anonymous and non-identifiable to the author, having only an ad hoc number distinction. Figure 1 depicts research environment and analysis measures.

Figure 1: Research environment and measures.

A session refers to an online visit, starting when the user logs on and ending when the user either logs out or the activity ceases for more than 30 minutes. In the data one or two panelists used the Web continuously. Otherwise, sessions were rather clear. The implication is that while at home, consumers appear to 'go' to the Internet, in many cases several times a day. Time spending is measured with two practically perfectly correlated constructs, viewing time and online time. Online time is longer, including actual content viewing time and page load times. Page views record the specific Web content accessed. The correlation between viewing time and the number of pages is .89 in these data.

The 65 panelists accessed a total of over 80,000 Web pages during four months. The 82,674 records of these pages are at the heart of the analysis. Other key statistics are 2,717 sessions and 4,934 Web sites visited. Additionally, entire panel-based Web site data are employed to refine and confirm findings.

The external validity of the data is evaluated as follows. The probability sampling procedure is solid, the amount of data analyzed vast, but the number of persons studied is relatively small. Combining these factors, the author is confident that the results are indicative of Internet user behavior beyond the individuals studied.

As concerns the transferability of the results to other countries, country-specific aspects are to be assessed. This includes Internet penetration, Web content types used, and also cellular penetration. These aspects are reported below for Finland and to some extent the United States as well. On the positive side, the findings obtained with this Finnish material can cohere with prior comparable research in other markets, as is referenced in the text wherever applicable.

The data are from the beginning of November 2001 to the end of February 2002. Throughout the text, all figures and numbers are from this four-month span. The panelist usage of the Web is generally stable over this period. At that time, 54 percent of the Finnish population aged 15 to 74 used the Internet weekly (Taloustutkimus, 2002). For comparison, figures from April 2002 indicate that 59 percent of the United States population is online (Nua, 2002). In the period studied, the average time spent online is relatively small in Finland, approximately five hours per month compared with over ten hours per month in the United States (Nielsen//NetRatings, 2002a).

Further, 80 percent of the Finns had a cellular subscription in the period studied (Ministry of Transport and Communications Finland, 2002). Wired and wireless network usage can be interrelated in at least two ways: through the prevalence of cellular text messaging and through wireless content display on regular Web pages. The first of these can reduce wired Web usage, while the latter can increase it. A further content-related aspect is that Finns have been slower to adopt online purchasing of goods. In contrast, e-banking became common relatively early in Finland. In this study only secured https:// sites are regarded as instances of e-banking. Other bank pages are classified as financial information.



The mosaic of the Web

The first aspect that became apparent in the four-month data of the 65 panelists selected for a close-up is that simple generalizations were elusive. An example of such a generalization is that certain types of content or ages are clearly associated with the longest viewing times. The data are richer. For example, the ten most frequent users are of quite different ages. The age groups among the ten heaviest users are 16 and 17 (four panelists); 29 to 35 (four panelists); over 50 (two panelists). Similarly, no one content type distinguishes their Web activity: town pages, online auctions, music and sports, general interest portals, e-banking and Web based e-mail are all present. This certainly indicates a mosaic of Web usage.

Published material identified portals, banks and search engines as Finland's most popular Web properties (Nielsen//NetRatings, 2002b). In the panelist data, these sites would also appear to be the most popular aggregately. At the same time, the number of different sites visited by individuals was often staggering. To structure content, three Web sites with the largest number of page views were determined for each panelist. These sites were visited on the Web and coded by site type. Figure 2 summarizes the site types most frequently appearing in individual panelists' top three and also the average age of the persons.

Figure 2: The site types most frequently appearing in the individual panelists' top three sites:
number of mentions and average age of persons.

Sample n=65 panelists; average age 30 years.

An interesting aspect in Figure 2 is the relatively low sum for these content types; 55 percent of a total of 195 sites. In other words, 45 percent of the individuals' favorites were coded as something other than the types displayed. Other site types are, for example, computer games and betting; literature; professionally inclined content; news and weather; adult sites; and, communities. As the assignment of sites into categories involves judgment, it is important to compare the sites objectively with generally popular Web sites.

In such comparisons, Web sites are still counted as each individual panelist's favorites. Consequently, a search engine receives many mentions, all included in the comparison calculation. Also, to make the testing tougher, one appearance in the monthly popular site records for the period gets the site placed in that category. As a result, 44 percent of the individual favorites are among the top 50 sites. A further 18 percent appears in the subsequent 51 to 240 listing of sites. Consequently a full 38 percent of the individuals' favorites in the data appear only in this panelist data. When combined, these aspects support the mosaic view of the Internet. Web usage appears to have maintained individual preference and taste variety.

There are also comments on the age averages in Figure 2. The two most deviant content types, music related and e-banking, differ from the sample age average by practically the same number of years. Slightly older age groups placed more emphasis on e-banking. This reflects the need for bill paying and other e-banking activities and therefore is an interesting aspect. It suggests that the e-environment may have matured; the choice of content seems to be guided by personally perceived value. It should, however, be born in mind that the data only include Internet users. Also, it is important that only the individual's favorite content types are displayed.

Additionally, it is interesting how pure entertainment value varied across age groups. Pre-teen users may spend time with various cartoon characters, while persons around 30 may manage a virtual football team. Thus the data suggest that specific content changes as people age, rather than necessarily the satisfaction derived from the content.



Web site usage: Concentration and exploration

This section addresses concentration and exploration in Web site usage. The analysis builds on the favorite sites identified. Two measures seemed to evaluate the relative position of the panelists' top three sites. The first measure compares the page views and viewing time of the top three sites with a person's total pages and viewtime. The other compares the three sites with the number of different sites visited by a panelist. Figure 3 shows the percentages represented by the top three sites at an individual level.

Figure 3: The percentages of total sites visited and total pages viewed for the individual panelists' top three sites.
Sample n=65 panelists. The panelists are arranged by viewing time, with the panelists recording the longest viewing times on the left.

Figure 3 displays individual variance in this sample. For comparison, a study based on a computer science department sample found that the top three pages accounted for 24 percent of the total pages visited (Cockburn and MacKenzie, in Brown and Sellen, 2001). These two results may be consistent, because throughout this data the lowest top three site page view percentages are associated with professionally inclined content, search engines, and music-related content. The highest shares are in turn associated with online auctions, real estate, and communities.

These panelists can also be placed in three groups. The first group comprises the 25 panelists with the longest viewing time. The second group is in the middle, also with 25 panelists. The third group comprises the 15 panelists with the least viewing time. Table 1 presents the data by group.


Table 1: The individual panelists' top three sites as percentages of total sites visited and total pages and viewtime, and the percentages of sites needed to arrive at 80 percent of the total pages viewed by an individual.
Sample n=65 panelists. Note that the first number in each cell displays the arithmetic mean, the average, while the second number in parenthesis depicts the amount of variation in the measure, the standard deviation.

Top three sites;
percentages of total sites visited
Top three sites;
percentages of all pages viewed
Top three sites;
percentages of total viewing time
Percentages of sites
needed to reach 80 percent
of the pages viewed
Longest viewing time
Middle viewing time
Shortest viewing time


Table 1 shows the pronounced position of the top three sites for high frequency users. In the first group, the top three sites account on average for a mere three percent of all sites visited by these persons. At the same time the three top sites account for 46 percent of the total pages viewed. Thus the weight is 15. In the middle group, this factor is a more moderate five. In the third group, the smaller amount of online activity affects the results, though indication is smaller relative weight for the top sites. If the first two groups are now combined (n=50), the individuals' top three sites represent 7.5 percent of all sites visited by one person and 50 percent of the total pages viewed by one person.

The favorite site emphasis of the most active users' represents a slightly surprising and important pattern. From the business viewpoint, this pattern makes gaining the loyalty of high frequency users likely to be rewarding, and relates to share of customer building. However, observational data do not reveal the extent concentration reflects brand loyalty, learning, or mere habitual choice (compare Jacoby and Kyner, 1973; Hoyer, 1984).

The preceding observations prompted us to ask whether the familiar notion of 20-80 would apply. This notion maintains that 20 percent of the subjects tend to account for 80 percent of a measure's volume. This widely applied notion has also been considered in the Web context (Pitt et al., 2001). Although the 20-80 ratio represents an idea, for clarity it is treated here as a rigid number combination.

To explore the applicability of the 20-80 notion, the number of sites needed to arrive at 80 percent of the total pages viewed by a panelist was checked. Now the testing was done in reverse, keeping the percentage of page views constant. The last column of Table 1 shows the results. Again, it is significant how the "gravitational" effect of their popular sites is strongest in the most active user group.

When the first two groups are combined, it is apparent that 25 percent of sites account for 80 percent of the pages viewed by an individual panelist. The 95 percent confidence level range of this combination does not cover 20. The same applies to the middle viewing time group separately. On the other hand, 20 is included in the respective range for the most active user group. The exact ranges at the 95 percent confidence level are as follows: longest viewtime group 14.9-20.8; middle viewtime group 28.4-34.3; longest and middle viewtime groups combined 21.8-27.4. This means that in the case of the most active users, the data on their site usage distribution are statistically in accord with the 20-80 ratio. For active and moderate users combined, only the general idea of the notion is supported in this four-month data set. The finding is the same for the 90 and 99 percent confidence levels.

Next it was analyzed how the 20-80 ratio would apply at the level of Web sites; records of the most popular sites in the period were therefore studied. The number of Web sites is 354 when the four-month data is combined. For comparability, two measures were calculated. They are based on page views and unique audience. The resulting numbers are summarized in Table 2. Of the two measures in Table 2, the page view records consider visit frequency, and can be regarded to communicate usage intensivity better than the unique audience measure. This page view distribution among Web sites agrees with the 20-80 ratio. The ratio is less descriptive for audience distribution.


Table 2: Popular sites position in page view and unique audience measurement.
Sample n=354 Web sites.

The first three sites on pages views/audience ...
account for 22 percent of the page views listed. account for 9 percent of the accumulated audience.
7.5 percent of the first sites on page views/audience ...
account for 59 percent of the page views listed. account for 40 percent of the accumulated audience.
To reach 80 percent of ...
the page views, 21 percent of the listed sites are needed. the accumulated audience, 36 percent of the listed sites are needed.


In Table 2, 'pages viewed' displays a stronger concentration for popular Web sites than 'audience'. This means that consumers explore several sites by logging onto them, but that a smaller number of sites attract them to stay longer and check out more content. This indication is consistent with the data on the sample of the individual panelists; heavy site concentration and exploration can be simultaneously present.

From the human point of view, both concentration and exploration may be associated with the large amount of choice and Web page proximity. Of these aspects, electronic proximity has been prominently noted in the literature (Dertouzos, 1997; also Nielsen, 1997; Shapiro and Varian, 1999). Amount of choice has also been researched extensively. Such research appears to indicate that while people like choice, a more reasonable number of options can be more comfortable (Jacoby, Speller, and Kohn, 1974; Iyengar and Lepper, 2000; Schwartz, 2000). A related and often quoted insight by Simon [2] is that a wealth of information creates a poverty of attention and a need to allocate that attention efficiently among the overabundance of information sources that might consume it. Another reflection of simultaneously present concentration and exploration relates to the concept of paradox (Mick and Fournier, 1998). Through electronic proximity and a large amount of choice, the digital environment may both promote and discourage experimentation, giving rise to an eventual paradox.



Navigation patterns: Opening a session

In the three remaining sections the focus is on navigation patterns. This section addresses how panelists open their sessions.

General interest portals have been prime entrance gateways to the Internet. These portals usually sell network connections; the connections have provider front pages preset as home pages. If users want to enter the Web through another site, they will need to change their preferences or settings. Then it is of interest to see the extent home users employ portals as Web session opening sites and whether they stay at that site.

In the data, some panelists appear to have a broadband connection which they do not close. Other panelists may have started some of their sessions after use by another family member, while the network connection was already open. These aspects condition the data. Nevertheless, patterns of how a panelist usually chooses to start a Web session did emerge. The analysis proceeded so that the session opening frequencies of different sites were first counted for each panelist. This indicated the Web site most often serving as the session opening site. Each specific site was identified as either a portal or a specialized content provider. Then the number of pages per session requested at that site by the panelist, beyond the first page, was computed. The resulting data varied with reasonable consistency across viewing times. Consequently, panelists are grouped by session opening pattern instead of viewing time. The main insights are described below.

Twenty-five percent of these panelists are likely to have changed their home page settings, as their start-up page is not a portal. In all cases, this site is among the individual's top three sites. Accordingly, the average number of additional pages requested is 6.7 per session. These panelists commonly use several sites to start their sessions; a mere 48 percent of the sessions were opened with the one site. The panelists with the very longest viewing times tend to be in this group. Additionally, the group includes more men than the sample gender distribution would suggest.

Fifteen percent of the sample also has a non-portal as their most frequent session start-up site. In contrast to the preceding group, significantly fewer home site pages were requested, only 0.5 pages per session. The panelists with the shortest viewing time tend to be here. The smaller amount of online activity makes this group less stable than the other three.

Twenty-three percent of the sample employ a portal as their session opening site. These panelists frequently return to their home page in the middle of a session for a transit-like visit, or to end their session. The average is 3.4 additional pages per session at the opening site. The group is most consistent in their session starting patterns, 85 percent of their sessions being initiated with their favorite portal. The group has moderately higher average age than the sample.

Thirty-seven percent of the panelists also start their sessions with a general interest portal but without requesting more than an average of 0.7 additional pages per session there. This is the most common way to open a session in the sample. Before moving on, these panelists may well spend some time viewing the contents of their home page. 75 percent of their sessions were started with the portal.



Pages in one Web session

The relationship between page views and sessions is highlighted in this section. Here the panelists' total page views are divided by their total sessions. In this pages-per-session ratio, the variance is interesting and important, and more specifically how the variance is rather consistent across online times. In other words, people with little overall online time may view the same number of pages once they log on as high frequency users. Figure 4 illustrates this phenomena. In this figure, there are data on the 65 panelists selected for the actual analysis and on all the panelists active in the period. This exception is made both because only on this occasion was there an opportunity, and because the figure gives additional information on the position of selected panelists in the larger group.

Figure 4: Page views per session ratio: panel and sample.
The panelists are arranged by online time, with those recording the longest online time on the left.

The background of the variance pictured in Figure 4 was examined with a sample of 65 panelists. Age correlates weakly though statistically significantly with the ratio in this data set: The younger the panelist, the more pages likely to be viewed in a Web session. Gender was not a factor. The kind of content at individual level typically associated with different ratios was also checked. The most distinctive content types were the following: music, town, and career-development-related site usage was scattered with relatively higher page-view-per-session ratios, while e-banking was concentrated with relatively lower ratios. However, Web site data indicated that e-banking contains a relatively high number of pages per visit. Hence, the connection would be more in the context and character of a session.

The connection to session character is more apparent when the individuals' data are further reflected with data on Web sites. The individuals' sessions and visits per site are compared. When individuals are concerned as in Figure 4, the pages viewed in a session average is 39 and the median 28. For Web sites, the corresponding statistics are 6.4 and 4.6. The differences between these numbers imply that consumers would visit several sites in one Web session. Then the intervening variable for the variance presented in Figure 4 can be how leisurely or evaluative, versus routine task-oriented the sessions of a person would likely be (see also Rozanski, Bollman, and Lipman, 2001).

The pattern of visiting several sites in one Web session is evident in the panelist data. Although there are exceptions, consumers typically navigate from site to site rather than making one-site-only Web visits. In qualitative terms, an aspect of learning may be more present for light users viewing many pages in their sessions (compare Bucklin and Sismeiro, in Bucklin et al., 2002).



Means of navigation from site to site

The final aspect verified from the panelist data concerns of means of navigation from site to site. The analysis focuses on how Web sites are entered in the middle of a session: By a link from the preceding site, via a search engine query, or by other means. Other means refer collectively to bookmarks, address typing, home page return, and other browser commands.

As session opening patterns have been addressed, attention is exclusively on mid-session site accesses. Hence, if a person starts a session with a site and stays there without making intermediate trips to other sites, the site is not counted. This can reduce the relative share of the individuals' favorite sites. In the analysis, interpretation and judgment were needed to decide the stage at which a site is counted again in a session. A position was taken that if a short visit with a couple of page views is made at another site, or if it appeared that a person keeps several windows open, hopping back and forth, the same site is counted only once. Panelists continue to be treated alike. Thus emerging differences among individuals are of interest.

In the results, the amount of online activity is again relevant. Additionally, panelist age is significant. Hence, previously used viewing-time-based groups are employed in reporting, complemented by age-based distinction. The implication of finer reporting is that group sizes become small, emphasizing the descriptive nature of the analysis. Figure 5 summarizes the results of this effort to quantify navigation path structures.

Figure 5: Mid-session site access patterns, measured at the level of individual panelists and presented as percentages.
Sample n=65 panelists.

A general observation is that with only a few exceptions, all panelists in the longest and middle viewing time groups use search-engine-query-based links, links from other referring sites, and other means such as bookmarks to access sites. The conditions for employing different means of navigations are discussed below. Another general observation relates to the same site access patterns over time. There were instances where a site first accessed with the help of a search engine link was later accessed through other means, presumably being bookmarked. On the other hand, there were instances where the same sites were repeatedly accessed through a search engine query and link. In those cases, search engines may have been used as an external memory.

At a summary level the most typical mid-session accessed site is in the category of other sites, reached through other means such as bookmarks or address typing. It is possible that these sites are familiar to a panelist. Also, panelists' favorite sites maintained their importance, even as only site access frequencies are counted. The industry implication is that adequate memory is needed for bookmarks and their user-considerate presentation, either in devices or in network applications. Given the number of different sites visited by individuals, the estimated bookmark needs are 65, 15 and 6 for the three viewing-time-based panelist groups in the period studied. So high frequency Web users can easily need space for over one hundred bookmarks.

Peer site referring links represent the second most common means of navigation when data are summarized. This was somewhat surprising. Such chaining often took place in ordinary navigation interaction. For instance community, building material, statistics, enterprise register, and television show sites were reached with site-to-site links. It seemed that links were employed when a panelist knew a suitable starting point (also Choo, Detlor, and Turnbull, 2000). Referring sites tended to be specialized content sites. From the business perspective, the pattern reminds of the value of cross-referencing links for sites that are not obviously known by consumers. The popularity of the means can also be seen as a tribute to the early hypertext system ideas (Nelson, 1987).

The third means tracked, search engine queries, are important but slightly less prevalent. By way of comparison another study, on high frequency users, found that six percent of the Web sites were accessed via search engines (Rozanski and Bollman, 2001). That is consistent with the present findings for at least 20-year-old panelists with the longest viewing time. Searches appeared to be used when the relevant terrain was larger, less familiar, or when panelists knew what they were looking for by specific name. Examples in the data include bands and lyrics, travel destinations, automobile clubs, medical services, and specific types of swords.

The longest and middle viewing time groups have reasonably similar profiles. Differences include that search-based site entries are relatively less common for most active than moderate users, in both age sub-groups. However, when panelists with the longest viewing time made searches, the queries tended to be very specific and well defined.

The age distinction in the two more active user groups also merits a comment. Interestingly, in both teenager sub-groups the share of sites accessed via searches is double that of the reference adult sub-group. Grown-ups may have found their favorites, being content with them. But the finding may also reflect the supply situation. Many of the teenager search queries are for music-related content, which appeared to be quite scattered on the Web.

As regards the shorter viewing time group, caution is in order when interpreting the results because more limited data affect its numbers. Two aspects are tentatively interesting in that group: the large share of sites accessed through bookmarks and similar other means in the adult sub-group (some 85 percent of mid-session site accesses) and the use of peer site referring links in the teenager sub-group (some 47 percent of mid-session site accesses). If teenage light users of the Web are considered to be novices, then the suggested tendency of relying on site-to-site links may be a form for seeking reassurance from others (compare Alba and Hutchinson, 1987).

A closing remark on the data concerns Web traffic flows. In this respect those panelists found to start their sessions with a non-portal site were specifically followed in their navigation records. The conclusion is that a clear majority of all panelists at some point visit both general interest portals and search engines. So in the mosaic of Web usage, portals and search engines served as common navigation nodes.




The analysis of the research material revealed a number of findings that have been reviewed and discussed. Three main conclusions were also highlighted in the abstract. These are the mosaic of the Web; the simultaneous presence of concentration and exploration; and rather balanced use of different navigation means. This section first briefly suggests the further implications of these conclusions, then evaluates the meanings of the study, and finally proposes potential avenues for research.

The mosaic-like preferences and exploration patterns fit today's wired Internet. But these usage patterns can be equally relevant for the development of the wireless Internet. Providing variety is likely to be central for attracting users' acceptance for the mobile environment. The usage concentration on individuals' favorite Web sites may connect with direct advertising; users could welcome discreet commercial messages in their key interest areas. In navigation patterns, bookmark and cross-referencing link use can have implications for application usability and device hardware design.

The basic research nature of this work can contribute to collective knowledge. Bringing state-of-the-art panel data into the sphere of academic analysis can facilitate other research. Hence results were presented in a detailed and comprehensive way. Other researchers, and practitioners, may then apply the specifics reported for their own tasks. It is, however, central to bear in mind that although the amount of data analyzed here is vast, the number of persons studied is relatively small.

This probability-sampling-based study covered the mass of consumer Web behavior. Exceptional users with over one hundred pages per Web session were visible in the panel records but not included by chance. Future research, perhaps in the behavioral sciences, may address such users and their patterns. Of background variables, only age and gender data were analyzed with the usage records here. Connections to other socioeconomic factors were not drawn due to the small number of persons studied. Research using larger samples may analyze these background factors.

When further reflecting on the results presented in this article, it is important to remember that the observational data do not communicate consumer thoughts and meanings. Key future research opportunities can therefore arise in studying why consumers behave the way they appear to do. An example is analyzing the reasons behind the observed and familiar 20-80 ratio. For that work, the present study provides a basis by describing consumer usage patterns on the World Wide Web. In essence, the strength and limitation of this analysis in complementing other research is concentrating on how consumers use the Web — day in, day out. End of article


About the Author

Nina Koiso-Kanttila is a doctoral candidate in marketing at the Helsinki School of Economics, Finland. She has been a full-time researcher since November 1999. Prior to this academic project, Ms. Koiso-Kanttila worked in high-tech industry for six years in Europe and in the Asia Pacific region.



The author thanks Nielsen//NetRatings for their contribution, and Professor Olli Ahtola and Dr Juhani Strömberg for their comments, which helped shape the article. The author is grateful to the persons who participated in the panel in Finland. This research is financed by grants from Foundation for Economic Education, Alfred Kordelin Foundation and Yrjö Uitto Foundation.



1. Marshall and Rossman, 1995, p. 79.

2. Simon, 1971, pp. 40-41.



Joseph W. Alba and J. Wesley Hutchinson, 1987. "Dimensions of Consumer Expertise," Journal of Consumer Research, volume 13, number 4 (March), pp. 411-454.

Patrick Barwise and Colin Strong, 2002. "Permission-Based Mobile Advertising," Journal of Interactive Marketing, volume 16, number 1 (Winter), pp. 14-24.

Barry Brown and Abigail Sellen, 2001. "Exploring Users' Experiences of the Web," First Monday, volume 6, number 9 (September), at, accessed 29 January 2003.

Randolph E. Bucklin, James M. Lattin, Asim Ansari, Sunil Gupta, David Bell, Eloise Coupey, John D.C. Little, Carl Mela, Alan Montgomery, and Joel Steckel, 2002. "Choice and the Internet: From Clickstream to Research Stream," Marketing Letters, volume 13, number 3 (August), pp. 245-258.

Chun Wei Choo, Brian Detlor, and Don Turnbull, 2000. "Information Seeking on the Web: An Integrated Model of Browsing and Searching," First Monday, volume 5, number 2 (February), at, accessed 29 January 2003.

Pratibha A. Dabholkar, 2000. "Technology in Service Delivery: Implications for Self-Service and Service Support," In: Teresa A. Swartz and Dawn Iacobucci (editors). Handbook of Services Marketing and Management. Thousand Oaks, Calif.: Sage, pp. 103-110.

Michael L. Dertouzos, 1997. What Will Be: How the New World of Information Will Change Our Lives. New York: HarperCollins.

Wayne D. Hoyer, 1984. "An Examination of Consumer Decision Making for a Common Repeat Purchase Product," Journal of Consumer Research, volume 11, number 3 (December), pp. 822-829.

IconMedialab, 2001. "Christmas Shopping 2001: Measuring the online Customer Experience of Northern European online retailers," IconMedialab and IBM usability study (December), at, accessed 29 January 2003.

Sheena S. Iyengar and Mark R. Lepper, 2000. "When Choice is Demotivating: Can One Desire Too Much of a Good Thing?," Journal of Personality and Social Psychology, volume 79, number 6 (December), pp. 995-1006.

Jacob Jacoby and David B. Kyner, 1973. "Brand Loyalty Vs. Repeat Purchasing Behavior," Journal of Marketing Research, volume 10 (February), pp. 1-9.

Jacob Jacoby, Donald E. Speller, and Carol A. Kohn, 1974. "Brand Choice Behavior as a Function of Information Load," Journal of Marketing Research, volume 11 (February), pp. 63-69.

Robert Kraut, Michael Patterson, Vicki Lundmark, Sara Kiesler, Tridas Mukopadhyay, and William Scherlis, 1998. "Internet Paradox: A Social Technology That Reduces Social Involvement and Psychological Well-Being?," American Psychologist, volume 53, number 9 (September), pp. 1017-1031.

Dorothy Leonard and Jeffrey F. Rayport, 1997. "Spark Innovation Through Empathic Design," Harvard Business Review, volume 75, number 6 (November-December), pp. 102-113.

Catherine Marshall and Gretchen B. Rossman, 1995. Designing Qualitative Research. Thousand Oaks, Calif.: Sage.

David Glen Mick and Susan Fournier, 1998. "Paradoxes of Technology: Consumer Cognizance, Emotions, and Coping Strategies," Journal of Consumer Research, volume 25, number 2 (September), pp. 123-143.

Ministry of Transport and Communications Finland, 2002. Televiestintätilasto 2002. Forssa: Forssan Kirjapaino.

Ted Nelson, 1987. Computer Lib; Dream Machines. Redmond, Wash.: Tempus Books of Microsoft Press.

Jacob Nielsen, 1997. "Why Web Users Scan Instead of Read," at, accessed 29 January 2003.

Nielsen//NetRatings, 2002a. at, accessed 29 January 2003.

Nielsen//NetRatings, 2002b. "Top 10 Web Properties for the Month of ... 2002," at, accessed 29 January 2003.

Nua, 2002. at, accessed 29 January 2003.

Leyland F. Pitt, Pierre Berthon, Richard T. Watson, and Michael Ewing, 2001. "Pricing Strategy and the Net," Business Horizons, volume 44, number 2 (March-April), pp. 45-54.

Subramanian Rangan and Ron Adner, 2001. "Profits and the Internet: Seven Misconceptions," MIT Sloan Management Review, volume 42, number 4 (Summer), pp. 44-53.

Horacio D. Rozanski and Gerry Bollman, 2001. "The Great Portal Payoff: Matching Internet Marketing to Consumer Behavior," Booz Allen Hamilton study (February), may enquire at

Horacio D. Rozanski, Gerry Bollman, and Martin Lipman, 2001. "Seize the Occasion: Usage-Based Segmentation for Internet Marketers," Booz Allen Hamilton and Nielsen//NetRatings study (April), at, accessed 29 January 2003.

Barry Schwartz, 2000. "Self-Determination: The Tyranny of Freedom," American Psychologist, volume 55, number 1 (January), pp. 79-88.

Carl Shapiro and Hal R. Varian, 1999. Information Rules: A Strategic Guide to the Network Economy. Boston: Harvard Business School Press.

Herbert A. Simon, 1971. "Designing Organizations for an Information-Rich World," In: Martin Greenberger (editor). Computers, Communications, and the Public Interest. Baltimore: Johns Hopkins Press, pp. 37-52.

Taloustutkimus, 2002. at, accessed 29 January 2003.

Laurie Windham and Ken Orton, 2000. The Soul of the New Consumer: The Attitudes, Behaviors, and Preferences of e-Customers. New York: Allworth Press.

Editorial history

Paper received 20 February 2003; accepted 24 March 2003.

Contents Index

Copyright ©2003, First Monday

Copyright ©2003, Nina Koiso-Kanttila

Consumers on the Web: Identification of usage patterns by Nina Koiso-Kanttila
First Monday, volume 8, number 4 (April 2003),