2.1. Study of an Individual Behavior in Virtual Space

Currently, scientists from psychological, computational, and engineering research areas have developed human-oriented behavioral informatics, characterized by methods for analyzing and coding human behavior. The publication of Belarusian scholars differentiates the concepts of “informational behavior,” “media behavior,” and “internet behavior,” discusses ways of studying the Internet behavior of individuals, and also presents the classification of Internet users (Esposito, Faundez-Zanuy, Morabito, & Pasero, 2019; Fablinova, 2015; Narayanan & Georgiou, 2013).

2.2. Behavior Patterns and Models

User behavior on websites is described by several models and patterns. The word pattern is used in various scientific disciplines and fields of activities. In psychology, a pattern refers to a specific set, pattern of behavioral reactions, or sequence of stereotypical actions; therefore, we can talk about patterns regarding any area where a person applies patterns (almost all of them). When users browse websites in searching for necessary information, they automatically apply a certain procedure. User behavior while browsing is described by the three most typical models: the F-pattern, the Gutenberg diagram, and the Z-pattern (Bradley, 2011; Gonçalves et al., 2013; Kotyak & Narizhnyi, 2016; Nielsen, 2006; Pernice, 2017; Pernice, Whitenton, & Nielsen, 2014).

According to web design experts, the Gutenberg diagram is the pattern which best describes user behavior when browsing website pages. This pattern describes the model of user behavior when viewing information which they see on the screen. In further scrolling and going to the next pages, the model works according to the same principle (Fig. 1) (Kotyak & Narizhnyi, 2016).

Fig. 1.

Gutenberg diagram.

Conditionally, the page is divided into 4 zones:

It is considered that user behavior is influenced by the habit of reading and studying text content from left to right (the user attention vector actually shifts diagonally from the priority viewing area to the exit area), as well as the natural gravity of reading (the glance instinctively shifts from top to bottom). At the same time, a study by Portuguese scientists found that visual patterns differ depending on the task to be performed (using search engines, online purchases, reviewing the website and information search), and not on the website where it is resolved (Pernice et al., 2014).

The vast majority of web users solve their problems as quickly as possible with minimal effort; they visit the website because they want to find a quick answer, and not to read a large text on the topic. If the website creators did not take any steps to direct the user to the most relevant, interesting, or useful information, users will find their own way (Rosenfeld, Morville, & Arango, 2015). In the absence of any signals that direct the viewpoint, they will choose the path of least resistance and most of the fixings will be near the place where they begin to read (which is usually the leftmost upper word on a text page).

A well-known and frequently cited study of user behavior patterns is the work of analysts from Booz-Allen & Hamilton and Nielsen//NetRatings (Rozanski Bollman, & Lipman, 2001). This study groups users by individual behavior at a certain point of time, and not by demography or psychography. This form of segmentation can be called an accident because it is based on what people do on the Internet in different cases. They found that four parameters were most significant in determining discrete behavior clusters:

The study of click flow data, characteristics of the session of Internet users, revealed seven discrete segments based on the use of the Internet in various cases. A number of other studies show similar results (Choo, Detlor, & Turnbull, 2000; Kellar, Watters, & Shepherd, 2007; Morrison, Pirolli, & Card, 2001; Sellen, Murphy, & Shaw, 2002).

As a result of the study, seven types of behavior were identified:

2.3. Applying of User Behavior Information to Improve Service

In publications on librarianship, modeling user behavior is viewed as an element of a personalized environment. Personalization is defined as the way information and services can be adapted to meet the unique and specific needs of an individual or community(Callan et al., 2003). A user model is a data structure that represents user behavior and captures human factors (Frias?Martinez, Chen, & Liu, 2007).

Researchers are studying user interaction with electronic content to create a system of recommendations. User behavior is monitored by several indicators, which are subsequently used to submit recommendations (He & Zhang, 2016; Núñez-Valdez, Quintana, Crespo, Isasi, & Herrera-Viedma, 2018; Xu , 2017; Yi, Chen, & Cong, 2018).

Specialists of Yandex (a Russian transnational company that owns the search engine of the same name, Internet portals, and web services in several countries) conducted a study of behavioral factors and concluded that an in-depth study of user experience helps significantly improve the current search system and the existing ranking model (Tolstikov, Shakhray, Gusev, & Serdyukov, 2014).

Researchers from Korea University, Seoul say that the analysis of cognitive and behavioral user information in a distributed environment is necessary to create an adaptive user interface. In their work, they developed a customizable UI/UX interface based on user cognitive response and behavioral information in a distributed environment (Ji, Yun, Lee, Kim, & Lim, 2017).

2.4. Web Analytics Use

Librarians use web analytics methods to evaluate the effectiveness of remote library services. They monitor user behavior using web analytics and transaction analysis. It is noted that these methods are promising to evaluate the effectiveness of various servicing points and planning a number of measures to enhance the user experience in remote library services (Tobias & Blair, 2015).

Libraries apply Google Analytics to help redesign their website. Fang (2007), the librarian from the Rutgers-Newark Law Library, used Google Analytics for this purpose. His goal was to analyze user behavior on the website. Fang (2007) analyzed usage statistics before and after library redesigns to determine the redesign effectiveness.

Many articles on library website evaluation consist of surveys and using Google Analytics to determine the top pages of content. Researchers at the Technical University of Texas point out that understanding user needs and behavior on library websites is crucial to make a successful library website. According to them, a proper assessment of the needs and behavior of regular visitors to an academic library is significant for the process of library website design (Barba, Cassidy, De Leon, & Williams, 2013).

Scientists point out that researchers use Google Analytics to study user behavior, website performance, and web traffic, allowing website administrators to study the online experience of their users and improve it (Clark, Nicholas, & Jamali, 2014; Clifton, 2012).

The ability of Google Analytics to track website use makes it a powerful tool for monitoring online behavior. The information provided by Google Analytics is used to answer questions regarding the website evaluation and library services use (Yang & Perrin, 2014).

In Russia, Yandex.Metrica analytical system is also used. This is a system similar to Google Analytics, but with additional functionality (WebVisor, link map, click map, scrolling map, form analytics). It is used for web analytics of any site, both commercial and non-profit organizations. The link map shows the statistics of clicks on links on the site, the click map shows statistics on clicks on the site, the scrolling map shows how the attention of site visitors is distributed in certain areas of the page, and form analytics shows how visitors to the site interact with the forms. Yandex.Metrica allows understanding of how users consume content on the library website: which words or areas attract their attention, what resources are in great demand, where the portions with the most click density are, and drawing of conclusions about how popular are the sections which the library itself considers important. At the same time, researchers note what web analytics metrics can show, what users do, but they usually do not show why users do it—and this is the task for analysts themselves to explain their behavior (Udartseva & Rykhtorova, 2018).

Researchers use Yandex.Metrica and Google Analytics web analytical systems to build custom models (Rykhtorova & Udartseva, 2018). Google Analytics event tracking can help determine the bounce rate, which is a metric often used to make decisions about library websites. Using the event function in Google Analytics is useful because knowledge of events creates an analytics environment where “iterative changes can be tracked in real time with minimal effort” (Loftus, 2012; Turner, 2010). Iterative changes are needed in an ever-changing library system, where services change regularly, and therefore websites are to be reorganized or redesigned.

The literature review in the work of researchers from the Boise State University showed that event tracking was not carefully studied in libraries. In the work, they showed the possibility of events tracking to increase retention, reduce the failure rate, and measure the actual drop-out from the library website (Vecchione et al., 2016).

Analysis of visitor actions allows determination of which pages are visited most often, where users go further, and so on. Understanding user behavior on the website provides an opportunity to improve library visitor servicing through the website and increase satisfaction.

This literature review has shown that quite a lot of research has been devoted to tracking user behavior. But the behavior of Russian library website users has not been carefully studied. This article presents a case study in which the application of browsing patterns and the implementation of event tracking led to improvements achieved through this technology and data analysis. Through identifying and researching these methods, the researcher has found a useful and successful methodology that any library can use on its website to effectively track user flow and improve virtual information services.

4.1. Optimization of the Web-Pages Structure of the SPSTL SB RAS Website

In this paper, we studied the behavior of SPSTL SB RAS website users through web analytics. WebVisor, a tool of YandexMetrica, allows the following: seeing how website visitors behave, analyzing usability, and finding errors in structure and design. Using Webvisor player, it is possible to see the exact repetition of all visitor actions on the website: mouse movements, clicks, page scrolling, keystrokes and filling in forms, and highlighting and copying text. WebVisor records the visitor actions at the web resource and allows viewing them as live video. Viewing records from WebVisor showed that users usually drive the mouse in a zigzag (Gutenberg diagram).

Using browsing pattern data made it possible to optimize the page structure of the SPSTL SB RAS website. The pages of the SPSTL SB RAS website with services for remote users contain a call to action: fill in the form, pay, or ask a question. It was necessary to draw user attention to the possibility of application online services and to encourage service payment through the website. In accordance with the studied browsing patterns, it was decided to place the payment and order buttons at the bottom right, in the exit and decision area, according to the Gutenberg diagram (Figs. 2 and 3). This change, we hope, will allow the user to immediately see all the possibilities of getting the service and make a decision. The online assistant window was also placed in the bottom right of the screen (Fig. 2).

Fig. 2.

Services order and payment buttons before change.

Fig. 3.

Order, payment, and online assistant buttons moved to exit and decision area.

4.2. The Mechanism of Recommendations on the SPSTL SB RAS Website

Further, we used data from the research by analysts at Booz-Allen & Hamilton and Nielsen//NetRatings. We have analyzed the data of visiting the SPSTL SB RAS website in accordance with the types of behavior described in Rozanski et al. (2001). These data gave an idea of the purpose for which visitors come to the website. These common behaviors can have important implications for library marketing tasks. By examining how the four session variables (session duration, page time, category concentration, and website familiarity) define different segments, we can identify behavioral models that can help in creating and posting content.

Analysis of website visits using Yandex.Metrica provides an opportunity to obtain data on the duration of sessions and the number of visits to identify the intended purpose of user visits. According to Yandex.Metrica reporting, the majority of SPSTL SB RAS website visitors come one time and spend 1.5 minutes there. Visitors with 2-3 visits are in second place where the average time on the website is about 3 minutes, and visitors with 4-7 visits take third place with time spent—4.5 minutes. Applied to the above, the main types of user behavior on the SPSTL SB RAS websites are Single Missions, Do It Again, and Loitering. If we summarize these results with the types of behavior proposed in the study (Kellar et al., 2007), we can conclude that users visit the SPSTL SB RAS website for information gathering, monitoring, and viewing. And this mode, according to the study authors, means that users tend to stay on one page and may be affected by various messages or recommendations.

These findings allowed us to optimize the page structure of the SPSTL SB RAS website for specific types of user tasks. So, in 2018, a recommendation mechanism was introduced on the SPSTL SB RAS website. When browsing pages related to patent topics, users were shown links to similar pages and sections of the website also related to patents. The analysis of Google Analytics data showed that after the implementation of the recommendation technology, users rather actively use them. The average number of transitions from the pages of patent subjects to other recommended pages is 40%, which indicates the efficiency of this technology (Shevchenko, 2019).

4.3. Library Resource Access Improvement

To improve access to library resources, it was necessary to identify the most popular of them. The analysis (carried out from January 1 to August 30, 2018) of page traffic according to Google Analytics showed the most popular resources of the SPSTL SB RAS website. The data on the most visited resources of the SPSTL SB RAS website are presented in Table 1, but they may be diverse for different libraries, depending on the audience of library users. On the SPSTL SB RAS website, the most demanded resources are: the electronic catalog (23.3%), the main page (8.6%), Resources and Services section (5.4%), About the Library section (3%), For Librarians (2.6%), news (2.5%), page with examples of bibliographic descriptions of publications 1.7%, and Remote Resources section (1.1%).

Table 1.

Most visited resources of the State Public Scientific-Technological Library of the Siberian Branch of the Russian Academy of Sciences (SPSTL SB RAS) website

After analyzing the data and determining the most popular content, it was decided to make these resources more visible on the SPSTL SB RAS website. One of the most popular and sought-after resources is the SPSTL SB RAS electronic catalog, which as confirmed by YandexMetrica click map is a tool for measuring and displaying statistics on website clicks. The map displays clicks on all page elements, warm colors correspond to frequent clicks, and cold correspond—to rare clicks (Fig. 4).

Fig. 4.

Click map of the State Public Scientific-Technological Library of the Siberian Branch of the Russian Academy of Sciences (SPSTL SB RAS) website. Link to electronic catalog is highlighted.

Clicking on the link to the electronic catalog, the user enters the platform of the Integrated (developing) library information system (IRBIS) automated library information system with a large number of catalogs and databases (more than 60), which can be misleading. To avoid this, the Resources and Services section was reorganized and divided into two sections: Resources and Services. The Resources section included links to all catalogs of the SPSTL SB RAS database and remote access resources. Links are grouped by type composition: books, journals, abstracts, etc. Links from these sections lead directly to the books, journals, and author’s abstracts catalogs, and also to databases, which increased the usability of the electronic catalog and databases. Also, in the process of analyzing the most visited pages, it turned out that the website page containing examples of bibliographic descriptions of various publications is quite popular—1.7% of all visits, but it is very difficult to find from the main page. As a result, the page was placed at the second level of the website in the section Researcher Support.

4.4. Event Tracking to Analyze User Paths

Identification of popular content helps establish the preferred paths used by the largest number of users. However, the user path report does not show where users move when they leave our website. If the user follows the link to a remote resource, in the report it looks like a session interruption, a user exit. In the “Remote Resources” section, links to subscription remote resources are grouped. As shown in Fig. 5, Google Analytics does not take into account how users accessed library resources in the Remote Resources section; we only see that they left the SPSTL SB RAS website. The report on the behavior map on the Remote Resources page shows that 49.7% of users left the page—374 interruptions out of 753 sessions (Fig. 5).

Fig. 5.

Map of user behavior from the Google Analytics report.

Google Analytics event tracking can help determine bounce rates. After implementing event tracking using Google Analytics, it became clear that users are not just leaving our website, but are switching over to subscription remote resources (Table 2).

Table 2.

Event map data on remote resources website page according to Google Analytics

SPSTL SB RAS, State Public Scientific-Technological Library of the Siberian Branch of the Russian Academy of Sciences.

The report on the events Switching to a Remote Resource showed that the most transitions are made to the SPSTL SB RAS electronic catalog, with the same percentage of transitions to Scopus database, journals of the American Physical Society, Lan (Fallow-Deer) electronic library system, and the Information System for the Maintenance of Licensed Access to Remote Scientific Resources of the Ministry of Education and Science of the Russian Federation.

Yandex.Metrica also allows tracking of events. To understand the demand for a new integrated service for authors helping to prepare texts for publication with the help of Yandex.Metrica, the Composite Purpose was created. When setting it up, several conditions were set that users shall meet in order for the goal to be considered achieved—page viewing with a service description, filling out the order form, and payment. This will give the opportunity to track all the way from the website visitor to the user who ordered this service. At the same time, we will receive detailed statistics on each step and see at which stage users most often leave the page, do not fill out the service order form, and do not make payment. This algorithm is useful to determine the ordering step where users may have problems.

The preliminary results of tracking a composite goal show that website users cover the entire chain before filling out the order form. Yandex.Metrica reporting shows that there were 240 page views of the Services for Authors page for the period from October 1 to December 31, 2018; 98 goals were achieved when choosing a particular service, and users completed an order form 10 times. There are not many service orders yet, which indicates the necessity to promote this service. A service promotion system is currently being developed: an advertising booklet is being prepared, a so-called “selling article” about this service is being written, and advertising information is being placed in the news feed of the website.