Keywords

1 Introduction

User behavior and satisfaction in web 1.0 is crucially linked to the efficacy of website navigation. Research on website menus constitutes a substantial part of online navigation studies. Despite the fact that today most of the studies use data-driven approaches to user reactions to website navigation elements, the growing variability of the latter demands more behavior-oriented research methods.

But despite the seemingly universal necessity of knowledge on menu structure in the changing design environment, the research that would address these rapid changes remains rare. Earlier works compare the three basic types of menus (static, sequential, and expandable) for two dimensions of reduction of mental load and growth of user satisfaction. The first one links task complexity to user experience, where task complexity is objectivized via measuring structural complexity (path length; [1]); and semantic relevance of menu indices to the search task [2]. The second approach looks for a perfect model of navigation representation, or menu type. Here, the term ‘menu’ often includes many other navigational elements, e.g. breadcrumbs.

But so far, two issues have not been raised substantially in the behavior-oriented navigation research. First, static and sequential menus have become virtually incompatible with the newest web design trends, especially with adaptive forms of web layouts. At the same time, new forms of navigation, like tagging-based ones, have taken dominant positions in information architecture. Second, these navigational elements tend to more and more co-exist on webpages, which is expected to highly complicate information search and, thus, affect user behavioral strategies and, ultimately, user satisfaction. Menu type diversity (or menu complexity) needs to be re-addressed today at user’s end.

In our research, we would like to address both aforementioned issues. Thus, we ask: Does the level of menu complexity influence the selection of search strategy for fulfilling navigation tasks? And does it influence user satisfaction in realization of information search task?

The remainder of the paper is organized as follows. Section 2 reviews the literature on menu complexity and its impact upon task performance. Section 3 presents our methodology and the conduct of the experiment. Section 4 presents our results and discusses them.

2 Menu Types and Task Complexity in User Navigational Behavior

2.1 Current Research on Task Complexity, Navigational Models, and User Performance

In today’s research on user navigation on web 1.0 websites, many attempts to establish the factors influencing efficiency of user navigational behavior have been made. The majority of these studies have been using two approaches.

The first approach relates user behavior and the respective user satisfaction to the nature of the task performed by a user. Its main focus is task complexity. In most cases, two dimensions of task complexity are discussed. Starting from the earliest research, the first one is objective and is linked to (and even defined by) the complexity of website structure; the second is subjective and, thus, user-dependent [3].

In more recent works, like the one by authors [1], scholars research upon the inter-relations between operational parameters of user behavior in the process of information search, the objective task complexity, and the subjectively perceived complexity of the task performance.

The authors claim that objective complexity is first and foremost defined by the length of the path that leads to the target information bit. This parameter is also studied as the level of location of target information within the website hierarchy, and the path length can be measured by the number of the levels in the website structure. The authors have also demonstrated that the objective complexity correlates with subjective perception of its complexity.

The study [2] adds another parameter, namely path relevance, to the factors of objective complexity. Path relevance is the extent to which the very description of the task hints to the way to target information. Path optimization is possible via using semantic linkages between the task definition and the menu indices (the names of menu rubrics).

Thus, the task complexity may be linked to a combination of structural (path length) and semantic (path relevancy) parameters of the path. In this case, low task complexity is characterized by a short path length and its high relevance, while high task complexity is, vice versa, characterized by a long path of low relevance. As the study [2] has shown, with the reduction of task complexity, precision and speed of task fulfilment grow, and with the growth of task complexity they both fall. These results correspond to earlier tests of complexity variance [4].

Earlier types of menus have understandably received more scholarly attention than those that appeared later; the former were compared for their efficiency. Thus, the authors [2] have compared the impact of expandable and sequential menus upon user productivity in information search. Expandable menus represent the whole context of possible user selections inside the information architecture, as well as the tree of reviewed pages; sequential menus feature only one level of website architecture and the current position of the user within it. The study has supported the idea of impact of task complexity upon the efficiency of a certain menu type: thus, for high complexity tasks, expandable menus were shown as more efficient. Also, expandable menus were found to be reducing subjective user disorientation (as evaluated by the assessors after the session). In a whole range of other works [5,6,7] the preference over sequential menus is also given to simultaneous menus that overview the content of a web portal and show all the levels of content location at the same time, even proving that task complexity is the determinant for the efficacy of sequential vs. simultaneous menus. According to another study [8], sequential menus work faster for simple tasks where comparing several sets of search results is not necessary. But with the growth of task complexity, e.g. when the users need to compare data on the screen, the efficiency of simultaneous (or expandable) menus grows. Thus, a linkage between cognitive load, the type of information search task, and the efficiency of a certain type of a navigational model was established.

Similar results have been received in a later study [9] that involved design aspects – namely, on vertical and dynamic menus. Here, the links between task complexity, menu design, and user performance and subjective experience evaluation.

The second approach focuses on finding an optimal model of representation of the website structure in the navigational instruments – that is, in most cases, on the menu type. A range of early studies of 2000s [10,11,12] have demonstrated that consecutive website navigation based on use of the so-called ‘breadcrumbs’ was preferential to the traditional menus of the early stages of web 1.0 development. This was especially true for the large web spaces with diversified architecture and strictly hierarchical content distribution between the levels of hierarchy.

This approach has been expanded by the works that take into account the cultural aspects of user perceptions of different menu types. Thus, the study [13] arguments the selection of an optimal menu model via the culturally defined peculiarities of user behavior in the process of information search; the authors call them cognitive styles. Depending on user belonging to either Western or Eastern cultures, the authors differentiate two navigational models – the ‘wide’ one (oriented to ‘overview’ perception of the website content) and the ‘deep’ one (oriented to hierarchical connections within the narrow search task).

Summarizing the main research approaches on the parameters of efficient navigation within a web interface, we may state the following. First, we need to take into account the task complexity – or, in other words, the cognitive load embedded into the structure of the task, and the ways to minimize it. Second, we need to remember that the axes along which the search of an optimal menu model was going were overview vs. linearity and context vs. hierarchy.

2.2 Today’s Trends in Web Design and Lack of Research on Menu Complexity

But today’s research still does not take into consideration three crucial changes that have been around on the Web in the recent years. This happens due both to general lack of navigation studies and also the rapid development of media design for portable, mobile, and wearable devices.

The first trend is proliferation of adaptive strategies of web design instead of those of responsive design preferred by web developers and designers in the 2000s. Due to adaptation of designed interfaced to portable and mobile devices, webpage design has re-oriented to vertical scrolling, representation via visual blocks, and growth of identical (and often dynamic) webpage elements instead of imitating pages of print media out of which the ‘classic’ landing pages grew. New web projects have also re-oriented to user-generated content and have integrated user-oriented popularity tracking mechanisms.

These trends, in its turn, have called for appearance and rapid proliferation of new types of menus, such as tag clouds, hub pages, and selective menus. They are dominating mobile and user-generated design of today, but the research on their efficiency is scarce enough.

Third, in the current situation, earlier types of menus co-exist with the newest ones, substantially raising the menu complexity and, allegedly, user frustration in the process of information search – despite the fact that the menus are developed to raise the speed of task fulfilment and user satisfaction. But, till today, we lack studies that would answer to the following questions: What happens with user search behavior if several types of menus are co-functioning on one webpage? How the diversity of the navigational system affects both information search efficiency and subjective user satisfaction?

We try to address these gaps in the current research on user navigation. In it, we employ the findings in the area of cognitive strategies, especially in that of error correction. Thus, a study by the authors [14] shows that error correction strategies are applicable to possible scenarios of information search behavior. This research area is especially important for the web spaces with big amount of textual data and low navigation relevance – e.g. for university web spaces.

As a rule, to reduce the content complexity, such web spaces offer to their users at least several alternative navigation instruments. But it is still unclear whether the efforts of web designers bring more trouble or more satisfaction to the users.

3 Research Design, Method, and Conduct of the Experiment on User Behavior

3.1 The Research Questions and Hypotheses

Following the research on error correction [14], we have divided the possible strategies of information search into productive and non-productive. The former strategy may be described as systemic information search based on use of one menu only or a certain evident principle in combined use of several menu types. The latter strategy may be described as ‘the method of trials and errors’ when search demonstrates a chaotic, sporadic character revealing the user’s disorientation.

We have chosen regressions – repetitions and recursions in referring to menu rubrics of the same menu, as well as returns to previously used menus – as the main parameter revealing a non-productive strategy of information search. Special attention should be given to the cases of repeated search in the menu rubrics where the target information is patently absent. According to the authors, such ‘rigid exploration’ [14: 427] signifies loss of user control upon the search process.

For this study, we have posed two formal research questions:

  • RQ1: With the growth of task complexity, does navigation (menu) complexity enhance or diminish efficiency of user search behavior?

  • RQ2: With the growth of task complexity, does navigation (meny) complexity enhance or diminish subjective user satisfaction?

Accordingly, our research hypotheses look as follows:

  • H1. With the growth of task complexity, navigation (menu) complexity grows the users’ inclination towards non-productive search strategies.

  • H2. With the growth of task complexity, perceived interface complexity also grows.

3.2 The Research Design

Based on our research premises, we have founded our experiment on two parameters: navigational (menu) complexity/variability and structural task complexity. For the clarity of the experiment, we have omitted the parameters of cultural proximity and semantic relevance from our analysis.

We have used one of the old layouts of the website of St. Petersburg State University, Russia (SPbU). Earlier, we have shown that, of three universities in Russia and the USA, this website demonstrated mid-range results [15, 16]; but, according to our preliminary research, it was the menu complexity that raised the biggest number of claims among the scholarship and studentship of this university who dealt with the university website. Thus, we have elaborated the experimental prototype based on a real-world design of the main hub of a large university web space.

Measuring Navigational (Menu) Complexity.

Menu complexity was measured by the number and nature of the navigational instruments (types of menus) offered to a user simultaneously within the search task fulfilment procedure. Thus, on one page, combinations of up to four different types of menus could exist:

  1. (1)

    an expandable hierarchic menu (see Fig. 1);

    Fig. 1.
    figure 1

    The real-world combination of case (1): hierarchical menu (‘ГЛABHOE MEHЮ’, main menu) – and case (3) – content-based hub (‘ДOКУMEHTЫ’, documents)

    Full size image
  2. (2)

    an expandable bar menu with dropdown second-level menu (all-encompassing, with low path depth) (see Fig. 2);

    Fig. 2.
    figure 2

    Case (2): expandable bar menu

    Full size image
  3. (3)

    a hub page with content-based navigation (see Fig. 1);

  4. (4)

    a popularity-based selective tag menu (see Fig. 3).

    Fig. 3.
    figure 3

    A combination of case (2), case (3), and case (4) – selective popularity-based tag menu

    Full size image

On the real-world SPbU web portal, several varying categorization principles were used. Thus, in cases (1) and (3), classification was based on activity types, in case (2) – on document types, in case (4) – on popularity of certain topics and individual pages among the website users. We kept this scheme in our prototype.

We have defined three combinations of menu variability – from earlier to earlier + newer menu types. In this way, we could assign the level of menu complexity to each of these combinations (see Table 1). For the real-world SPbU web portal, a three-menu style was typical for most pages, but many sub-nodes had the four-component one.

Table 1. The menu variability levels for the SPbU web portal.
Full size table

We have ensured the possibility of alternative search to a full extent via selecting the target information the way that it could be found by any navigation scheme independently of each other.

Measuring Structural Task Complexity.

The second parameter in our experiment was structural task complexity. Structural complexity was assessed by two dimensions. The first one, relevant for cases (1) and (3), was path depth (choice between 2 and 3 immersion levels). The second one, relevant for case (2), was menu options diversity: low-complexity bar menu had five or fewer indices, high-complexity bar menu had over five indices. Thus, in sum, high-complexity tasks were based on three immersion levels and 5 + bar menu indices, while low-complexity tasks were based on two immersion levels and under five or fewer bar menu indices.

Thanks to this, we could define the level of task complexity (see Table 2).

Table 2. Levels of task complexity
Full size table

In accordance with the defined navigational schemes and task complexity levels, six local html-versions of a university website prototype were created.

3.3 Conduct of the Experiment

Assessor Groups and Data Collection.

Two groups of assessors, 10 persons in each, were part of the experiment, all being students of another university in St. Petersburg, Russia; this was done to ensure that the assessors have no cause for regular use of the SPbU web portal and are ignorant of its inner features.

During the experiment, the two groups fulfilled the search tasks of varying complexity upon the six html-versions of the SPbU web portal. Each of the tasks (of low and high task complexity) was performed on the three versions with varying navigation schemes.

We used a JavaScript scenario for registration of the website sections reached by a given assessor. The scenario created a descriptor note on each menu index: with each click of a user on a menu section, the respective note would appear in the log journal of the given user. These notes were later used as the research sample.

Description of the tasks.

As stated above, the tasks were created the way that they could be fulfilled disregarding the chosen navigational scheme.

Additionally, we loaded the assessors with at least two categorization regimes. The first task was about finding the target information on types of university activities. This categorization scheme corresponded with the tree of the expanded hierarchical menu of the case (1) obviously linked on the real-world web portal to the administrative, teaching, scientific, and expert activities in the university. The second categorization scheme was linked to the case (2) of the upper one-level expandable menu and is oriented to the target document search, as the menu indices mention normative documents, lists, announcements, archives etc. The case (3) of content-oriented navigation actually combines these two categorization schemes, while the case (4) of the tag-based menu breaks them. This is why the task description had to contain the indications of both the document type and the type of university activity (see Table 3).

Table 3. Examples of the task descriptions in the experiment.
Full size table

Additionally, two crucial conditions for task description had to be fulfilled:

  1. 1.

    The task description may refer to several indices of different menus on one page at the same time (see Table 3, comments on Task 1 description), which rises the user’s cognitive load.

  2. 2.

    The task must not coincide literally with the menu indices but may contain direct referrals to the name of the target document.

The Conduct of the Pre-test Experiment.

The experiment that we have designed consisted of two sessions. Throughout the test, both groups were taking part in it; this is why, for each new session, the task content changed. The changes were related to the menu indices /referrals under scrutiny while the menu structure was preserved. The tasks of high complexity the referrals were related to scientific activities; those of low complexity were linked to studying (see Table 4). Also, to make the process traceable for us, each assessor place had a pen-and-paper set, where the assessors could mark their trajectory if they occurred to have regressions.

Table 4. Examples of the task descriptions in the experiment.
Full size table

In the immediate aftermath of the sessions, the assessors were asked to evaluate their user experience in terms of perceived navigation complexity. They answered the question ‘Please evaluate how easy it was for you to navigate this page’; the available answer options were structured according to a modified Likert scale and included the following answers:

  1. 1 –

    ‘extremely uneasy and complicated, to the extent that I find it annoying’;

  2. 2 –

    ‘not that easy, to the extent that I felt discomfort’;

  3. 3 –

    ‘usual level of complexity, typical for the majority of websites’;

  4. 4 –

    ‘easy enough’;

  5. 5 –

    ‘way too easy, I did it intuitively and with no effort from my side’.

The results of the experiment and presented and discussed below.

4 Results and Discussion

4.1 The Research Results

The results of our experiment are presented in Table 5.

Table 5. Examples of the task descriptions in the experiment.
Full size table

We did not calculate any correlations between the menu complexity levels and the user strategy, as our data were too scarce to be statistically significant, but the mean figures that we have calculated are also telling.

Thus, Table 5 shows that, for the low complexity tasks, menu complexity has virtually no impact upon navigational behavior, as well as to the subjective evaluations of easiness at information search. In this case, all the three navigational schemes saw the assessors search for the target information confidently and consistently; the absence of regressions and productive (mostly linear) patterns of finding the target information clearly tell of this. For the low-complexity tasks, the majority of assessors used the upper expandable menu and viewed the menu indices from left to right.

The impact of menu variability upon user search behavior is found only for the high-complexity tasks, and only in the cases of three and four menus represented on one page. Here, the level of subjective easiness of search drops almost twice and the percentage of regressions rises to 5% for three-menu pages and 8% for four-menu pages. Also, user preferences change radically: if two-menu pages invoked clicks on the upper one-level menu (84% to 90%), the appearance of additional navigation instruments fosters the growth to hierarchical left-side menu (from 9–15% to 37% of moves). Surprisingly, the assessors never used the contextual and tag-based menus, which suggests that, in case of growing complexity, the older, perhaps more reliable and usual navigational instruments are perceived as all-encompassing and become a natural starting point for information seeking. It is on three- and four-menu pages where the assessors’ activities lost its consistency: the users allow repetitive study of the rubrics they had assessed earlier and are switching from one menu type to another.

Thus, our H1 and H2 on the growth of non-productive search strategies with the growth of task complexity have proved partially right – they are true for the cases of middle and high menu complexity. Our pre-test results show that, unlike common-vesical expectations, user satisfaction does not grow linearly with the growth of menu complexity: the three-menu pages invoked the results even worse than the highly complicated four-menu ones. Thus, saturating web pages with the maximal amount of navigational instruments shows up as clearly counterproductive, and finding an optimal scheme is trickier than one could expect. Our results suggest that we can recommend a two-menu navigational scheme; we consider all other navigational instruments on the same page excessive, disorienting, and annoying.

Our results are, of course, subject to substantial limitations. Thus, we did not involve such parameters as time of task performance and user metadata into the current report on our data; this was due to the pre-test nature of our experiment. Also, the questionnaire has to be expanded in future to more precisely track the users’ attitudes towards the navigation diversity. Also, the assessors’ absolute disregard of the newer navigation instruments seems odd and definitely deserves a further investigation. But we can show that, in the online world where the web space complexity grows as well as the search tasks do, excessive complexity seems to be counter-productive.