Abstract
User engagement is a relation of emotion, cognitive, and behavior between users and resources at a specific time or range of time. Measuring and analyzing web user engagement has been used by web developers as a means to gather feedback information from web users in order to understand their behavior and find ways to improve the websites. Many websites have been successful in using analytics tools since the information acquired by the tools helps, for example, to increase sales and the rate of returning to the websites. Most web analytics tools in the market focus on measuring engagement with the whole webpages, whereas the insight information about user behavior with respect to particular contents or areas within webpages is missing. However, such knowledge of web user engagement based on contents of the webpages would provide a deeper perspective on user behavior, compared to that based on the whole webpages. To fill this gap, we propose a set of web-content-based user engagement metrics that are adapted from existing web-page-based engagement metrics. In addition, the proposed metrics are accompanied by an analytics tool which the web developers can install on their websites to acquire deeper user engagement information.
Access provided by Autonomous University of Puebla. Download chapter PDF
Similar content being viewed by others
Keywords
1 Introduction
User Engagement is a relation of emotion, cognitive, and behavior between users and resources at a specific time or range of time [1]. In the past years, user engagement analytics on websites has been widely used by web developers to gather feedback from web users. In a survey of datanyze.com, 97.6 % of Alexa top one million websites use web analytics tools to analyze their user behavior data to obtain the insight of web usage [2]. The example of the analytics tools with the highest market share are Google Analytics, Google Universal Analytics, Yandex Metrica, comScore, and Quantcast, where Google Analytics occupies 45.5 % among these tools.
The analytics tools can provide web developers with the insight and feedback of web usage including what the users do on the websites, when they visit, how they interact with the websites, and many more. Once equipped with such information, the web developers can find ways to improve their web sites. For example, using an analytics tool, an e-commerce company [3] could identify the lost revenue due to high shopping cart abandonment rate and improved its website features to finally obtain an increase in checkout to payment page. Another case of a company [4], using an analytics tool to analyze which versions of its landing page should be used, could obtain an increase in homepage engagement and a boost to e-commerce conversion rate.
To analyze user engagement, a number of metrics have been developed to measure web usage and calculate user engagement. Click-Through Rate, Time Spent on a Site, Page Views, Return Rates, and Number of Unique Users are examples of popular metrics [5]. However, available analytics tools and metrics focus on analytics of user engagement with the whole webpages, rather than the web contents. Figure 1 depicts web contents, namely c1, c2, and c3, on a webpage, where Fig. 1a shows page-based engagement data that can be collected from users, i.e. mean visit to a page, mean hover over a page, and mean click to a page. However, the usage information regarding a particular content on a page is not known. If we can enhance page-based engagement metrics with content-based ones, i.e. mean visit to a content, mean hover over a content, and mean click to a content as in Fig. 1b, web developers can obtain deeper understanding of how users interact with the contents and can improve their websites. This paper proposes a set of web-content-based user engagement metrics that are adapted from existing web-page-based engagement metrics. In addition, the proposed metrics are accompanied by an analytics tool which the web developers can install on their websites to acquire deeper user engagement information.
The rest of the paper is organized as follows. Section 2 discusses related work. Section 3 explains the web contents model and content-based user engagement measurement method. In Sect. 4, user engagement metrics based on web contents are presented together with a case study. Section 5 describes the content-based user engagement analytics tool, and the paper concludes in Sect. 6.
2 Related Work
We focus the discussion about related work on user engagement metrics and a popular web analytics tool.
2.1 Web User Engagement Metrics: A Web Analytics Approach
Lehmann et al. [6] study models of user engagement with web applications which can vary as the way users engage with different applications can be very different. A set of metrics, grouped into popularity, activity, or loyalty categories, is used in their study as listed in Table 1. These metrics are widely used to measure how users engage with webpages or websites. This paper adapts from these metrics to devise a set of web-content-based engagement metrics.
2.2 Web User Engagement Metrics: A Physiological Approach
A physiological approach to user engagement measures the users involuntary body responses, e.g. eye tracking, mouse tracking, and facial expression analysis [7]. The measurement may involve specific hardware device, e.g. eye tracking, or only software, e.g. mouse movement detection. In this paper, we consider only the metrics that can be measured by software, i.e. mouse gestures, in particular. Huang et al. [8] examine the users’ mouse cursor behavior on search engine results pages to help better design effective search systems. The mouse cursor behavior includes clicks, cursor movements, and hovers over different page regions. We adapt from the metrics in Table 2 which are used in their work.
2.3 Web Analytics Tool
We discuss Google Analytics [9] as an example of the tools which can be installed on a website to track how users engage with the website. Table 3 shows metrics that are used.
Google Analytics visualizes the measurements as graphs and tables. Figure 2 depicts how Google Analytics, for example, shows pageviews of a single page in a graph and others in a table. This paper will use similar visualization techniques to report web-content-based user engagement information.
3 Web Contents Model and Content-Based Measurement
In this section, we give a definition of the web contents model, characteristics of web contents, and how user engagement is measured in this model.
3.1 Web Contents Model
The web contents model is depicted in Fig. 3. A website is a collection of webpages, where each webpage is composed of multiple containers. A container is a specific area on a particular webpage and can be nested inside another container. A web content is contained in a container, and can be seen on a web browser, e.g. text, image, sound, video etc. It can also be part of another content, e.g. text on an image.
3.2 Characteristics of Web Contents
Analyzing user engagement with web contents is different from that with webpages because of some unique characteristics of web contents as follows.
First, a web content can appear in different containers on different webpages at the same time. As shown in Fig. 4, a web content A can be in different containers on webpages 1 and 2 at the same time. We name a web content as cc or content in a container. That is, the same web content in different containers would be referred to differently, e.g. the web content A in Fig. 4 is referred to as c1 and c2 with regard to its container on a webpage 1 and another one on a webpage 2 respectively.
Second, a container on a page can contain different web contents at different time. As shown in Fig. 5, a web content B contained in a container on a webpage 3 at time \(= 1\) is referred to as c3. Later at time \(= 2\), that container contains a different content C which will be referred to differently as c4.
Given these characteristics of web contents, we require a different set of metrics and tool for web-contents-based engagement analytics.
3.3 Measuring Content and Container Visit
Measuring visit to web contents and containers is different from measuring visit to webpages since the latter can be done when the page is loaded. However, for web contents and containers, they might be at the bottom of the page and cannot be seen by the user when the page is loaded. As shown in Fig. 6a, only the Content 1 is shown in the viewport but the Content 2 is not because it is off the screen. In the case of measuring the visit to web contents and containers, the viewport of the screen will be used to determine whether the web contents and containers are visited. For example, Content 2 will be considered as visited when the page is scrolled down until shown in the viewport as in Fig. 6b.
4 User Engagement Metrics Based on Web Contents
The proposed user engagement metrics based on web contents are classified into three levels, i.e. low-level, high- level, and overall-level metrics. The low-level metrics are used to measure directly the behavior of the users on web contents. High-level metrics then are derived from the low-level ones. The overall-level metrics are further derived from the high- level metrics to obtain the overall engagement information. These three classes of metrics are adapted from several of those widely used engagement metrics for webpages discussed earlier in Sect. 2. Before we describe the proposed metrics, we first introduce a case study to which the metrics will be applied as an example.
4.1 Case Study
The example in Fig. 1 is used as a case study and revised with some information added as shown in Fig. 7. There are five CC or content in a container. c1, c2, and c3 are on page 1. At time \(= 1\), c4 on page 2 has the same content (i.e. a music video A) as c2 on page 1. When time \(= 2\), the container of c4 has its content changed to a music video B, and that location is then identified as c5. Table 4 summarizes the web content and container of each CC in the case study.
4.2 Low-Level User Engagement Metrics
Low-level user engagement metrics measure usage data directly from three sources, i.e. webpages, web contents, and containers. Since a content and a container share the same location on a webpage at the same time, we use the same low-level metrics for these two sources. Each metric is listed with its definition in Table 5.
Note that the metrics for webpage sources are taken from existing metrics but we enhance by proposing additional metrics for web contents and containers. Using the low-level metrics, we obtain the measurements of the case study in Tables 6 and 7.
4.3 High-Level User Engagement Metrics
High-level user engagement metrics are derived from low- level metrics and categorized by user behavior approaches, i.e. visit, click, and hover. Each metric is listed in Table 8 together with its definition. Table 9 shows the calculation results for the case study.
High-level metrics can be used to determine user engagement with a web content in a container. However, to determine the overall user engagement with a web content that may appear in several containers as well as the overall user engagement with a container that may contain several contents over time, we need additional overall-level metrics.
4.4 Overall-Level User Engagement Metrics for Web Content
As depicted in Fig. 4, a web content may be contained in several containers across different webpages. To determine user engagement with this particular web content, we calculate an average engagement values of that web content over all containers in which it is contained:
where Engagement \(_{content}\) = overall user engagement with a web content at a particular time
n = number of containers in which that web content is contained, and
Engagement \(_{content,i} =\) user engagement with that web content in the container i (obtained by using high-level metrics).
In the case study, c2 and c4 have the same content 2 (music video A). Using the engagement measurements for c2 and c4 in Table 9 as Engagement \(_{content,i}\), we calculate the overall user engagement with the music video A at a particular time. In Table 10, the column “Content 2 at time \(= 1\)” lists the overall user engagement measurements for the music video A at time \(= 1\). For example, VSR\(_{content2} = (0.90\,+\,1.0)/2 = 0.95\) at time \(= 1\).
4.5 Overall-Level User Engagement Metrics for Web Container
As depicted in Fig. 5, a container may contain several web contents at different time. It is useful to get an insight into how a particular container is engaged in order to design a container better or select suitable content for a container. To determine user engagement with a particular container over time, we calculate an average engagement values of that container over all contents contained in it:
where Engagement \(_{container}\) = overall user engagement with a web container over a range of time
t = number of times that user engagement with that container is determined (at a regular interval) over the time range, and
Engagement \(_{container,i}\) = user engagement with that container measured at time i (obtained by using high-level metrics).
In the case study, c4 and c5 refer to the same container 4 (on webpage 2) with different contents (music video A and B) at different time. Using the engagement measurements for c4 and c5 in Table 9 as Engagement \(_{container,i}\), we calculate the overall user engagement with this container 4 over a range of time. In Table 10, the column Container 4 from time = 1 to time = 2 lists the overall user engagement measurements for this container 4. For example, VTR\(_{container4} = (1.0+0.4)/2= 0.7\) over that time range.
5 Web-Contents-Based Analytics Tool
We develop a web-contents-based analytics tool that collects user interaction data, determine user engagements using the three classes of metrics, and visualize engagement information. The tool is implemented in PHP, with MySQL database. To install the tool on a website, a web developer has to register the website and generates jQuery code to put on the website. Once the code is installed, the tool will automatically collect necessary data, send to the server to process, and visualize the analytical results. The tool provides a dashboard by which the web developer can select particular web contents or containers (based on how the developers design the pages) and the engagement metrics that are of interest. Figure 8 shows an example of a graph displaying the UCR\(_{C}\) (User Clickthrough Rate) for the content 4 (music video B) at different time.
6 Conclusion
The paper proposes an enhancement to existing web user engagement metrics by introducing additional engagement metrics that can take into account engagement with particular contents and areas on webpages. The analytical result is aimed to give an insight into user behavior and to help determine a way to improve website design by placing the right web contents at the right location. We are conducting a test on a commercial website that has installed the web-contents-based analytics tool in order to see if the site can benefit from the analytical results and improve their website design. Other future work includes improving the visualization of the tool and extending the set of engagement metrics.
References
Attfield, S., Kazai, G., Lalmas, M. et al.: Towards a science of user engagement (Position Paper). In: WSDM Workshop on User Modelling for Web Applications. http://www.dcs.gla.ac.uk/~mounia/Papers/engagement.pdf (2011). Accessed 9 April 2016
Datanyze: Analytics market share in the Alexa top 1M. https://www.datanyze.com/market-share/Analytics/ (2016). Accessed 9 April 2016
Google Analytics: Brian Gavin Diamonds sees 60 % increase in customer checkout with Google enhanced ecommerce. https://static.googleuserconent.com/media/www.google.com/en//intl/es%20_ALL/analytics/customers/pdfs/brian-gavin.pdf (2016). Accessed 9 April 2016
Google Analytics: With help from Periscopix, WBC employs advanced segments in Google Analytics to boost its e-commerce conversion rate by more than 12 %. https://static.googleusercontent.com/media/www.google.com/en//intl/es%20_ALL/analytics/customers/pdfs/wbc.pdf (2016). Accessed 9 April 2016
Lehmann, J., Lalmas, M., Baeza-Yates, R. et al.: Networked user engagement. In: Proceedings of the 1st Workshop on User Engagement Optimization (UEO’13) (2013)
Lehmann, J., Lalmas, M., Yom-Tov, E. et al.: Models of user engagement. In: Proceedings of the 20th International Conference on User Modeling, Adaptation, and Personalization (UMAP’12) (2012)
Lalmas, M., O’Brien, H., Yom-Tov, E.: Measuring User Engagement. In: Marchionini, G. (ed.) Synthesis Lectures on Information Concepts, Retrieval and Services. Morgan & Claypool Publishers (2015)
Huang, J., White, R., Dumais, S.: No clicks, no problem: using cursor movements to understand and improve search. In: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI’11) (2011)
Google Analytics: Google Analytics Solutions. https://www.google.com/intl/en/analytics/#?modal_active=none (2016). Accessed 9 April 2016
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2016 Springer International Publishing Switzerland
About this chapter
Cite this chapter
Chokrasamesiri, P., Senivongse, T. (2016). User Engagement Analytics Based on Web Contents. In: Lee, R. (eds) Computer and Information Science. Studies in Computational Intelligence, vol 656. Springer, Cham. https://doi.org/10.1007/978-3-319-40171-3_6
Download citation
DOI: https://doi.org/10.1007/978-3-319-40171-3_6
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-40170-6
Online ISBN: 978-3-319-40171-3
eBook Packages: EngineeringEngineering (R0)