The usage of large data sets in online consumer behaviour: A bibliometric and computational text-mining–driven analysis of previous research

The usage of large data sets in online consumer behaviour: A bibliometric and computational text-mining-driven analysis of previous research 1. Introduction In the past two decades, advances in information technology as well as the Internet and digitalisation have transformed our daily lives and shifted the focus of commerce towards online digital environments. This change has also altered the ways consumers behave, for including how they shop and buy products and services (Darley, Blankson, & Luethge, 2010; Kim & Lennon, 2008). In the online environment, consumers’ decisions to purchase certain products and services are influenced by variables beyond the actual product or service, such as the design of the website or electronic word-of-mouth (including online reviews and recommendations) (see e.g. Cantallops & Salvi, 2014). Thus, online consumer behaviour has been one of the major research areas in marketing science, and it is the subject of a vast number of studies. For example, information system and marketing scholars have examined environmental site features (Manganari, Siomkos, Rigopoulou, & Vrechopoulous, 2011; Richard & Habibi, 2016), global search features (Hausman & Skiepe, 2009), language options (Hausman & Skiepe, 2009), site design (Ha, Kwon, & Lennon, 2007; Mummalaneni, 2005), site security (Hausman & Skiepe, 2009) and culture (Richard & Habibi, 2016). In recent years, there has been an increasing trend emphasising the importance of data analytics as well as the use of larger data sets. Data analytics is not new. Indeed, various kinds of data have been collected in one form or another for a very long time. However, technological This is the accepted manuscript of the article, which has been published in Journal of Business Research, 2020, 106, 46-59. https://doi.org/10.1016/j.jbusres.2019.09.009 improvements (e.g. in storing and transmitting information) have enabled the continuous and ubiquitous collection of data. The ability to store this data is also nearly unlimited, due to everlarger local storage technologies and the rise of cloud storage solutions. Thus, the practice of collecting data on a large scale has become easy and common. However, the challenge for analytics is to go beyond collecting and storing the data and structure it in such a way that the information it contains is accessible and usable. This is where big data or big data analytics, the hot topic of the past decade, takes its place. Big data is a term that refers to both collecting and analysing large data sets; that is, it is a way to build connections within the data set so that the information it contains becomes usable (see e.g. Ruh, 2012; Press, 2013; Erevelles, Fukawa, & Swayne, 2016). However, despite the above-mentioned emergence of research on online consumer behaviour and large data sets, according to a review article by Darley et al. (2010), the research has been dominated by studies utilising survey data. Over 70 per cent (37 out of 52) of the studies included in their review were based on survey data, including some experiments (31 per cent; 16/52). Thus, the motivation in this paper is not only to discover how research publications on online consumer behaviour cover big data but also how other kinds of large data sets have developed during the last decade and a half. Using a combination of bibliometric analysis and topic modelling, this paper sets out to answer the following research questions: ▪ How has the research landscape of online consumer behaviour evolved over time? ▪ What are the dominant areas of research? ▪ Which authors and articles are most influential? ▪ Which publication outlets are influential? ▪ What kinds of topics have been studied? ▪ What are the avenues for future research on online consumer behaviour? This study advances the research of online consumer behaviour in multiple ways. First, using a bibliometric citation analysis, the study contributes to the field by identifying the key publications and authors as well as how certain topics have evolved over time. Second, in addition to the traditional bibliometric analysis, which provides mostly descriptive information about the literature, we also use more advanced topic modelling and text analytics techniques. With this more comprehensive approach, it is possible to dig deeper into the literature and see what kinds of topics can be identified in a data-driven way without, for example, restricting the analysis to a predesigned set of potential topics. This provides both new data-driven insights as well as justifications (or a lack of them) for hypotheses about the prevalent research topics. Thus, it is possible to gain more in-depth information about the existing literature. This paper is organised as follows. Following this brief introduction, Section 2 presents the concepts of large data sets and online consumer behaviour. Section 3 presents the methodology utilised in the paper. Section 4 presents the findings of the traditional bibliometric analysis as well as the results of the topic modelling. Finally, in Section 5, we discuss the results, present the conclusions and limitations and provide suggestions for further research. 2. Large data sets and online consumer behaviour It has been argued that consumer behaviour in the online context is not that different from consumer behaviour in traditional settings and that patterns of general consumer behaviour are also applicable to the online context (Lee & Chen, 2010). For example, Koufaris (2002) stated that the traditional and online environments clearly share characteristics, but, within the online environment, consumers have additional unique needs and concerns because of their two-fold roles as both consumers and information system users (Lee & Chen, 2010). Montgomery, Li, Srinivasan and Liechty (2004) claimed that compared to traditional consumers, the online consumer is more convenience-oriented and innovative and seeks more variety. Consequently, online consumer behaviour can be defined as a manifestation of consumers’ decision-making and behaviour in the online environment as well as the factors that influence this behaviour when, for example, customers search for information, browse websites, evaluate products or compare information. In recent years, the average consumer has turned into a constant generator not only of traditional (i.e. structured and transactional) data but also of contemporary, unstructured and behavioural data. The magnitude of this generated data, the rapidity of its generation and its diverse richness impact how it should be treated and interpreted. As a natural consequence, big data has emerged as a topic in the related research literature, although there is a lack of a coherent use of the term (see e.g. Gandomi & Haider, 2015). In the strict sense, big data means data sets that have become too large and complex for conventional statistical measures, thus requiring advanced and unique methods for data storage, management, analysis and visualisation (Chen, Chiang, & Storey, 2012). Or, as Erevelles et al. (2016) put it (see also De Mauro, Greco, Grimaldi, & Ritala, 2018; Sivarajah, Kamal, Irani, & Weerakkody, 2017), data sets have become larger and more complex because of the unprecedented volume, velocity and variety of primary data available from individual consumers. According to Manyika et al. (2011; see also Erevelles et al., 2016), the underlying methodology for big data analytics is linked to existing disciplines, mainly statistics and computer science. Big data can be defined as high-volume, high-velocity and high-variety data assets (commonly referred to as the three Vs), which demand innovative processing methods so that the resulting information can be used to derive enhanced insights for decisionmaking (Erevelles et al., 2016; Hofacker, Malthouse, & Sultan, 2016; Lycett, 2013). Consumer behaviour produces large amounts of certain types of data. One of these is clickstream data, defined as an electronic record of Internet usage collected from the server log of a website or by third-party services. It typically contains data regarding individual navigational mouse clicks and clicks on other elements of web pages that a visitor makes when visiting a certain website. This kind of data provides information about the sequence of pages viewed by users as they navigate a website (Bucklin & Sismeiro, 2009; Montgomery et al., 2004; Moe, 2003; Senecal, Kalczynski, & Nantel, 2005). Another type is text data, which includes all kinds of natural language text, such as web pages, emails and social media posts (e.g. Zhai, 2017). A particular use of online consumer data is path analysis, which is based on the data about the navigational path a user takes through the website as well as insights based on the analysis of that data. In other words, path data analysis produces information about the sequence of events leading to a decision to purchase. This information typically contains data about the user’s goals, knowledge and interests, which can be used to predict consumer behaviour (Bucklin & Sismeiro, 2009; Montgomery et al., 2004). Meanwhile, customer analytics is defined as the overall process by which data regarding customer behaviour are used to help make business decisions. Analytics lies at the junction of data and consumer behaviour; data provide insights regarding customer behaviour, for example, and marketers turn these insights into a market advantage (see e.g. Corrigan, Craciun, & Powell, 2014; Erevelles et al., 2016). 3. Methodology 3.1 Data collection The initial search for potential articles took place in February 2019. We collected data from the academic database ISI Web of Science (also known as Web of Knowledge) and its Core Collection. According to Fetscherin and Heinrich (2015; see also Fetscherin & Usunier, 2012), the ISI Web of Knowledge is a suitable database when the object is to conduct an interdisciplinary literature review, and many notable bibliometric studies have used this database. We searched for publications that appeared from January 2000 to December 2018. The year 2000 was chosen as the cut-off year, as it could be considered the first year when the usage of large data sets emerged in the literature. December 2018 marked the most recent date for which complete citation data from the ISI Web of Science were available to us. To collect comprehensive data, we searched for journal articles written in English based on the appearance of search terms in the topic field in the Web of Science. The topic field includes the title, abstract and keywords of articles. It also includes so-called ‘keyword plus’ content, additional relevant but overlooked keywords that were not listed by the authors or publishers but which are based on the expertise of the editors of the Web of Science. We utilised our own threefacet classification for the search terms. The first level was named ‘data level’ and represented which kind of data was used in the articles. The search terms for this level were ‘big data’, ‘clickstream’, ‘text data’, ‘path analysis’ and ‘customer analytics’. The second level was labelled ‘individual level’, representing the role of the individual. The search terms ‘customer’ and ‘consumer’ were utilised for this level. Finally, the third level represents the context in which the consumer or customer operates. For this level, we included the terms ‘purchase’, ‘buying’, ‘online discussion’, ‘information search’, ‘customer journey’, ‘online’, and ‘mobile’ in our search. The search was conducted so that search criteria within each level were complementary alternatives (i.e. it was enough that one of the search terms within the level should occur); for this, we utilised the Boolean operator ‘OR’. Between the levels, the search criteria had to be fulfilled at the same time (i.e. at least one search term should occur from each of the levels); for this the Boolean operator ‘AND’ was utilised. The total set of search criteria we used in the Web of Science is presented in Table 1. INSERT TABLE 1 ABOUT HERE As a result of the search, we found 498 articles. Based on a manual inspection of these articles, we removed three articles because they were not actually related to large data sets. Eventually, 495 articles were included in the bibliometric analysis. We categorised the articles into eight different groups including ‘Business and Economics’, ‘Computer Science and Information Science’, ‘Engineering’, ‘Health’, ‘Hospitality’, ‘Environmental Sciences’, ‘Communication’, and ‘Other’ based on the types of journals to which they belonged, where the ‘Other’ group contained literature from, for example, chemistry-, mathematicsand education-related journals. Table 2 lists the number of publications from different fields. INSERT TABLE 2 ABOUT HERE 3.2 Bibliometric analysis A bibliometric data analysis was conducted to provide a quantitative analysis of the academic literature. Bibliometric analysis is the statistical analysis of written publications and citation analysis, and it is based on the construction of a citation graph, a network or graph representation of the citations between documents. Many research fields use bibliometric methods to explore the impact of their field, of a set of researchers or of a particular paper (e.g. Nicolaisen, 2010; Fetscherin & Heinrich, 2015; Hajikhani, 2017). For this study, the bibliometric data analysis was conducted using the Network Analysis Interface for Literature Studies (NAILS) tool (Knutas, Hajikhani, Salminen, & Porras, 2015). The NAILS tool analyses several essential variables from each publication, including the authors, keywords, publication forum, article type and cited articles. After the data are downloaded from Web of Science, the tool removes duplicate records and performs an exploratory data analysis, which identifies the most-cited articles and authors, the most common keywords and the journals with the most publications. The system also extracts the citation network data from the publications. In addition to the exploratory analysis report, the tool further provides data about citation and author cooperation networks, which can be visualised. This kind of data set about citation connections can be used to calculate the relative influence of publications in the network (Knutas et al., 2015). In the NAILS tool, several metrics (e.g. PageRank and In-Degree) were calculated based on the citation data. PageRank (Brin & Page, 1998) counts the number and quality of links to a paper to obtain a rough estimate of its importance, while In-Degree provides the number of citations referring to an individual paper based on the citation data records. Utilising these metrics, a report was produced, which provided an abstract and keyword analysis as well as the most productive authors and journals. 3.3 Text mining and topic modelling Text mining by topic modelling aims to discover topics that occur in a collection of documents in order to explore hidden semantic structures in the body of the texts. In this work, the text mining and topic modelling was conducted using the structural topic model (STM; Roberts, 2014). Although the most common topic modelling technique is Latent Dirichlet Allocation (LDA; Blei, 2012) and an implementation of it is even built into the NAILS tool (Hajikhani, 2017; based on a visualised application of LDA by Sievert and Shirley, 2014), we decided use the more advanced and sophisticated STM method. In short, this is because STM both turned out to be a quantitatively better fitting probabilistic model for the document corpus and because, unlike LDA, it is directly able to model the interaction of covariates like time with the topics; additionally, the specific NAILS implementation of LDA has further restrictions. We next as abstracts of the articles were scrutinised. Based on this, the number of articles including clickstream, path analysis and text mining increased modestly from 2000 to 2018. Notably, however, the number of articles containing references to big data or discussing it experienced rapid growth between 2013 and 2018. In 2013, there was only one such article; in 2014, there were nine articles; in 2015, there were 20 articles; and in 2018, there were over 60 articles related to big data. Another growing area of research seems to be related to social media. Since 2013, there has been substantial growth in the number of articles, and in 2018 there were over 20 articles related to social media. INSERT FIGURE 3 ABOUT HERE When looking only at keywords, the popularity of big data was similarly evident. As Figure 4 shows, big data was listed as a keyword in 113 articles. In contrast, the more traditional objects of interest in this research field, such as e-commerce (26), clickstream data (16 articles) and online reviews (13), were much less in focus in the articles included in our study. INSERT FIGURE 4 ABOUT HERE In addition to the pure occurrence of the keywords, we also inspected their importance. To do this, we used the number of citations for the articles containing certain keywords. As can be seen in Figure 5, ‘Big data’ was the most-cited keyword, with 951 citations. ‘E-commerce’ (763 citations), ‘Internet’ (607) and ‘Clickstream data’ (544) were also widely cited keywords. Other keywords that had over 200 citations included ‘Internet marketing’ (310), ‘marketing’ (302), ‘Social media’ (287), ‘Consumer behavior’ (276) and ‘Computer-mediated environments’ (248). INSERT FIGURE 5 ABOUT HERE 4.2 Development of the number of publications Figure 6 shows the development of publications in terms of the number of articles over time. The number of articles published per year varied from one to 15 in the years between 2000 and 2010. After that, the growth was substantial, and the number of articles grew each year, from 10 published articles in 2011 to 128 in 2018. INSERT FIGURE 6 ABOUT HERE Also, looking at the relative publication volume (i.e. the fraction of publications included in our study of all publications within the Web of Science) reveals how the importance of the topics included in our study emerged. As Figure 7 shows, the fraction of all publications included in the study started to grow mildly from 2010 and turned into practically vertical growth from 2013 onwards. INSERT FIGURE 7 ABOUT HERE 4.3 Noteworthy authors Figure 8 provides the list of the most-cited authors. Of all the authors in the 495 articles, the six most cited are C. A. Lin , R. Larose, M. S. Eastin (all with 298 citations), S. J. Newell, B. A. Lafferty and R. E. Goldsmith (all with 285 citations). In addition to these six authors, four had over 200 citations: W. W. Moe, P. S. Fader, A. L. Montgomery and S. Bellman. INSERT FIGURE 8 ABOUT HERE In addition to focusing purely on the absolute number of citations, we also inspected the most important papers. They were identified utilising three importance measures: first, the In-Degree in the citation network; second, the citation count provided by the Web of Science (only for papers included in the data set); and finally, the PageRank score in the citation network. The topten highest-scoring papers were identified using these measures separately. The results were then combined, and duplicates were removed. The results presented in Table 3 are sorted by InDegree, and ties are broken first by citation count and then by PageRank score. INSERT TABLE 3 ABOUT HERE Based on this analysis, the three most important papers in our data set were as follows: Montgomery et al. (2004), discussing the modelling of online browsing and path analysis using clickstream data; Bucklin et al. (2002), discussing research progress and future opportunities for modelling consumer choice on the Internet using clickstream data; and Johnson, Moe, Fader, Bellman and Lohse (2004), examining online search behaviour. 4.4 Important publication outlets As Figure 9 shows, the most influential journals in terms of article count are the leading marketing journals, led by Marketing Science (14 articles) and followed by the Journal of Marketing (10). In addition, technology-oriented journals are potential outlets in this field: Electronics Commerce Research and Applications (10 articles), IEEE Access (8) and Expert Systems with Applications (10). INSERT FIGURE 9 ABOUT HERE The vital role of marketing journals is more apparent when the focus is on the volume of citations for different journals (see in Figure 10). Marketing Science is by far the most-cited journal with 613 citations, followed by Journal of Marketing (469 citations) and Journal of Marketing Research (320). The next most-cited ones have been cited between around 300 times: Media Psychology (298 citations), Journal of Advertising (291), Management Science (265) and Journal of Business Research (260). INSERT FIGURE 10 ABOUT HERE 5. Topic modelling Next, we analysed the content of the text, including the titles, keywords and abstracts of the articles, to discover what kinds of topics might occur within our data set. Here, we conducted the text analysis by utilising a more advanced text analysis method called structural topic modelling, following the procedure described in section 3.3. We finally extracted 14 meaningful topics and labelled them based on the top words of each topic, and the identified labels and corresponding top words are briefly described in the following. The topic Shopping Experience and Satisfaction contains top words including ‘consumer’, ‘study’, ‘intention’, ‘satisfaction’, ‘experience’ and ‘shop’. The topic Online Behavior Pattern includes the top words ‘online’, ‘use’, ‘time’ and ‘pattern’. The topic Privacy in Digital Community consists of words such as ‘information’, ‘social’, ‘datum’ and ‘privacy’. The topic Information Search comprises top words like ‘consumer’, ‘information’, ‘online’ and ‘search’. The topic Challenges in Retail Market contains words like ‘market’, ‘big data’, ‘research’, ‘retail’ and ‘challenge’. The topic Quality of Customer Experience puts more emphasis on words related to ‘customer’, ‘service’ and ‘quality’. The topic Online as a Sales Channel includes top words such as ‘online’, ‘study’, ‘relationship’ and ‘sale’. The topic Analysis of Social Networks encompasses ‘datum’, ‘analysis’, ‘mobile’, ‘user’ and ‘network’. The topic Brands and Online Reviews consists of top words ‘product’, ‘use’, ‘review’ and ‘brand’. The topic Purchasing Behavior contains top words such as ‘purchase’, ‘behavior’ and ‘model’. The topic Data Analytics and Food includes the top words ‘datum’, ‘analytics’, ‘system’, ‘big data’, ‘business’, ‘value’, ‘process’, ‘new’ and ‘food’. The topic Advertising Performance contains top words such as ‘model’, ‘effect’, ‘use’ and ‘advertise’. The topic Perceived Attitudes Towards Companies is expressed by the top terms ‘model’, ‘attitude’ and ‘perceive’. The topic Digital Age and Services contains the terms ‘technology’, ‘service’, ‘trust’ and ‘knowledge’. The top word list for each topic and the corresponding word cloud figure can be found in Table 4 and Figure 11. Out of the 14 identified topics, the topic Privacy in Digital Community has received the highest total number of citations (21.58). The topic Shopping Experience and Satisfaction has received the highest number of relative citations (2.93). The topic Analysis of Social Networks has received the fewest relative citations (2.06) compared to other topics. One possible explanation for this could be that this topic is related to research interests that have emerged only recently, and the small number of citations may be due to the nature of the accumulation of citations, although we have adjusted the citations over time. The prevalence of each topic over each year is shown in Figure 12, where the red dotted line is the five-year moving average value. Based on the plot, although the specific prevalences vary year to year, there is some indication that the topics Privacy in Digital Community, Quality of Customer Experience and Digital Age and Services may similarly have a rising trend. Other topics are either unstable or stable over time, and no obvious trends are found. Note that there is only one article in 2000; therefore, the drastic decreases or increases in topic prevalence during that year can only be due to the lack of data. We further conducted a Spearson non-parametric test between the topic prevalence and time on the above-mentioned three topics. The Spearman’s rank correlation coefficient values are 0.58, 0.53 and 0.75, respectively, with corresponding pvalues of 0.01, 0.02 and 0.0003. Another thing worth noticing is that the topic Analysis of Social Networks may have experienced a change point in the year 2006, with the trend of the topic prevalence decreasing until 2006 and increasing afterward. The Spearman’s rank correlation coefficient value between time and topic prevalence of this topic from 2000 to 2006 is -0.85 (pvalue: 0.02) and 0.84 (p-value: 0.0006) from 2006 to 2018. We further examined the topic prevalence in different categories by specifying a linear regression model where the output was the proportion of the nine identified topics while the input was the category of the publication. The comparison of topic prevalences in different groups is displayed in Figure 13. As shown in the figure, the topic Privacy in Digital Community is more prevalent in the category ‘Communication’. Fiore-Gartland and Neff’s work (2015) focuses on ‘data valence’, which is relevant to the notion of this topic. Meanwhile, the topic Shopping Experience and Satisfaction was frequently discussed in the literature under both the categories ‘Hospitality’ and ‘Communication’. Studies in the above two categories have shared common interests regarding this topic. For example, Yuan et al.’s work (2008) from the category ‘Hospitality’ (which is categorised in the Web of Science as Hospitality, Leisure, Sport & Tourism) discussed customer attitudes and wine purchasing. The work of Bobkowski (2015), which belonged to the ‘Communication’ category, discussed how the personalities of news consumers affect their news-sharing behaviours. For the topic Online as a Sales Channel, there was little obvious difference between the prevalences in different categories. One possible explanation for this could be that it is a general topic, so it received equal attention in different categories. This topic also contained a certain number of commonly used terms in this field, such as ‘study’, ‘sale’, ‘finding’ and ‘literature’. The prominent prevalence of the topic Digital Age and Services in the categories ‘Health’ and ‘Environmental Sciences’ is also worth mentioning. It reveals the academic interest in digital services. Ramkumar et al. (2017) explored the potential of using mobile health data for orthopaedic surgeons, while Beal and Flynn’s work (2015) is an example of research on ‘smart metering’ in water utility, an important issue in environmental sciences. The STM model also enabled us to compare the wordings of two different categories with respect to the same topic, in terms of which words or bigram terms were prevalent in the use of the topic within the different categories. We mainly compared the categories ‘Business and Economics’ and ‘Computer Science and Information Science’, as they had the largest number of publications (141 and 63, respectively, See Table 2), making the comparison meaningful. The wording difference for the topic Shopping Experience and Satisfaction is displayed in Figure 14. The literature in the ‘Business and Economics’ category mentioned more terms related to consumers, such as ‘consumer’, ‘experience’ and ‘attitude’. In contrast, the terms mentioned in the ‘Computer Science and Information Science’ category seem to focus on the discussion of the overall trend and data sources, such as ‘positive’, ‘negative’, ‘sns’ and ‘news’. There was also a difference between the categories ‘Business and Economics’ and ‘Computer Science and Information Science’ with respect to the topic Privacy in Digital Community (shown in Figure 15). Researchers in ‘Computer Science and Information Science’ tended to place more emphasis on technical discussions, thus using terms such as ‘Communication’, ‘datum’ and ‘digital’ frequently. Meanwhile, terms related to human–computer interaction, such as ‘information’, ‘interaction’ and ‘social’, were found in the literature in the ‘Business and Economics’ category. INSERT TABLE 4 ABOUT HERE INSERT FIGURE 11 ABOUT HERE INSERT FIGURE 12 ABOUT HERE INSERT FIGURE 13 ABOUT HERE INSERT FIGURE 14 ABOUT HERE INSERT FIGURE 15 ABOUT HERE 6. Discussion In this study, we report the evolution of scientific research with regard to the usage of large data sets on online consumer behaviour between 2000 and 2018 in terms of publications available in the ISI Web of Science database. The results based on 495 articles provide an overview of the existing information regarding research on online consumer behaviour utilising or discussing large data sets. This study provides information on the evolution of the field by identifying key publications and authors and examining how certain topics have evolved over time. By utilising topic modelling and text analytic techniques, we also identified certain research themes from the papers included in our data. This analysis offers a guide to those who want to contribute to this field of online consumer research by providing information, for example, about which journals should be consulted, which authors are the most prominent and what topics are relevant in different fields. With this paper, we also contribute to the methodology related to literature surveys and bibliometric analyses. Notably, we utilise a more advanced methodology beyond traditional bibliometric analyses—or even beyond basic LDA modelling. We conducted topic modelling to extract the latent topics from the collected literature. We demonstrated the usage of STM, a more advanced text analysis tool in literature review studies, showing that this automated framework enables researchers to gain more interesting insights than with other off-the-rack text-mining tools such as LDA. Our results show that the research landscape of online consumer behaviour utilising large data sets has evolved during the selected time period (2000 to 2018), so the topic has evidently become more popular and important in recent years. The number of articles published per year has increased rapidly from 2011 onwards, and the share of the articles from all publications within the Web of Science has skyrocketed since 2013. Perhaps the most notable change relates to the occurrence of the term ‘big data’ in the papers. This is particularly obvious from 2013 onwards, and it appears to be a highly dominant area of research. Moreover, the number of citations related to ‘big data’ is dominant, even though researchers have only relatively recently started to focus on big data. In addition, articles focusing on the Internet in general (Internet, ecommerce) and clickstream data as well as on Internet marketing and consumer behaviour are among the most-cited articles. However, the popularity of big data is so evident that we consider it likely the number of citations for such publications will continue to increase considerably. When we look at the publication outlets, it seems that the leading marketing journals are those that conscious academics should target in order to exert as much influence within the research community as possible. Our results show that journals such as Marketing Science and the Journal of Marketing and the Journal of Marketing Research have published most of the articles, and they are also among the journals that are cited most often. However, general management and business journals (e.g. Management Science and Journal of Business Research) as well as journals with a technological orientation (e.g. IEEE Access and Expert Systems with Applications) are other potential publication outlets for the researchers in this field. In terms of topic modelling, by utilising the above-mentioned novel and more advanced STM methodology, we managed to identify certain topics under which the papers included in the data set could be classified. Using this kind of modelling, it is possible to show hidden common content in papers by studying latent topics. It is also possible to compare citations of different topics or the occurrence of the topics in different fields of research. As a result, we discovered that papers related to Privacy in Digital Community had received the highest number of citations, followed by papers under the topics Challenges in Retail Markets and Shopping Experience and Satisfaction. Thus, it seems that these are the most influential topics in the field. A bit surprisingly, papers under the topic Analysis of Social Networks have received fewer citations. Again, however, we believe that due to the popularity of articles related to data analytics and social media, together with the nature of the accumulation of citations, this topic could receive a notable number of citations in the future (indeed, since 2006, the trend of the topic prevalence has been increasing). One interesting finding related to the topic Shopping Experience and Manyika, J., Chui, M., Brown, B., Bughin, J., Dobbs, R., Roxburgh, C., & Byers, A. H. (2011). Big data: The next frontier for innovation, competition, and productivity. McKinsey Global Institute. Retrieved from http://www.mckinsey.com/insights/business_technology/big_data_the_next_frontier_for_ innovation Marketing Science Institute (2016). Research priorities 2016–2018. Cambridge, MA: Marketing Science Institute. Moe, W. W. (2003). Buying, searching, or browsing: Differentiating between online shoppers using in-store navigational clickstream. Journal of Consumer Psychology, 13(1/2), 29– 39. Moe, W. W. (2006). An empirical two-stage choice model with varying decision rules applied to internet clickstream data. Journal of Marketing Research, 43(4), 680–692. Montgomery, A. L., Li, S., Srinivasan, K., & Liechty, J. C. (2004). Modeling online browsing and path analysis using clickstream data, 23(4), 579–595. Mummalaneni, V. (2005). An empirical investigation of web site characteristics, consumer emotional states and online shopping behaviors. Journal of Business Research, 58(4), 526–532. Murty, D. (2008). Digital ethnography. An examination of the use of new technologies for social research. Sociology, 42(5), 837–855. Nicolaisen, J. (2010). Bibliometrics and citation analysis: From the Science Citation Index to cybermetrics. Journal of the American Society for Information Science and Technology, 61(1), 205–207. Orgad, S. (2009). How can researchers make sense of the issues involved in collecting and interpreting online and offline data? In A. N. Markham & N. K. Baym (Eds.), Internet inquiry. Conversations about method (pp. 33–53). Los Angeles, CA: Sage. Philander, K., & Zhong, Y. (2016). Twitter sentiment analysis: Capturing sentiment from integrated resort tweets. International Journal of Hospitality Management, 55(2016), 16– 24. Press, P. (2013). A very short history of big data. Retrieved from http://www.forbes.com/sites/gilpress/2013/05/09/a-very-short-history-of-big-data/ Ramkumar, P. N., Muschler, G. F., Spindler, K. P., Harris, J. D., McCulloch, P. C., & Mont, M. A. (2017). Open mHealth architecture: A primer for tomorrow's orthopedic surgeon and introduction to its use in lower extremity arthroplasty. The Journal of Arthroplasty, 32(4), 1058–1062. Richard, M.-O., & Habibi, M. R. (2016). Advanced modeling of online consumer behavior: The moderating roles of hedonism and culture. Journal of Business Research, 69, 1103–1119. Roberts, M. E. (2014). Fear or friction? How censorship slows the spread of information in the digital age. Unpublished manuscript, 26. Ruh, B. (2012). The industrial internet: Even bigger than big data. Retrieved from http://www.forbes.com/sites/ciocentral/2012/10/04/the-industrial-internet-even-biggerthan-big-data/ Senecal, S., Kalczynski, P. J., & Nantel, J. (2005). Consumers' decision-making process and their online shopping behavior: A clickstream analysis. Journal of Business Research, 58(11), 1599–1608. Shlomo, N., & Goldstein, H. (2015). Editorial: Big data in social research. Journal of the Royal Statistical Society, 178(4), 787–790. Sievert, C., & Shirley, K. (2014). LDAvis: A method for visualizing and interpreting topics. Proceedings of the Workshop on Interactive Language Learning, Visualization, and Interfaces: 63–70. Retrieved from http://www.aclweb.org/anthology/W/W14/W14-3110 Sivarajah, U., Kamal, M. M., Irani, Z., & Weerakkody, V. (2017). Critical analysis of big data challenges and analytical methods. Journal of Business Research, 70, 263–286. Torres, E. N., & Singh, D. (2016). Towards a model of electronic word-of-mouth and its impact on the hotel industry. International Journal of Hospitality & Tourism Administration, 17(4), 472–489. Tvinnereim, E., & Fløttum, K. (2015). Explaining topic prevalence in answers to open-ended survey questions about climate change. Nature Climate Change, 5(8), 744–747. Van Auken, S. (2015). From consumer panels to big data: An overview on marketing data development. Journal of Marketing Analytics, 3(1), 38–45. Van den Poel, D., & Buckinx, W. (2005). Predicting online-purchasing behaviour. European Journal of Operational Research, 166(2), 557–575. Yuan, J., Morrison, A. M., Cai, L. A., & Linton, S. (2008). A model of wine tourist behaviour: A festival approach. International Journal of Tourism Research, 10(3), 207-219. Zhai, C. (2017). Probabilistic topic models for text data retrieval and analysis. Proceedings of the 40th International ACM SIGIR Conference on Research and Development in Information Retrieval, 1399–1401. Table 1 Search Criteria Level Search term Data TOPIC: (“big data” OR “clickstream” OR “text data” OR “path analysis” OR “customer analytics”) AND Individual TOPIC: (customer OR consumer) AND Related TOPIC: (“purchase” OR “buying” OR “online discussion” OR “information search” OR "customer journey” OR “online” OR “mobile”) Table 2 Number of Publications in Each Category Category Number of publications Business & Economics 195 Computer Science and Information Science 133 Engineering 33 Health 26 Hospitality 26 Environmental Sciences 16 Communication 14 Other 35 Table 3 Most Important Papers Year Author Title Journal InDegree Times Cited PageRank 2004 Montgomery et al. Modeling online browsing and path analysis using clickstream data MARKETING SCIENCE 48 192 0.0000810 2002 Bucklin et al. Choice and the Internet: From clickstream to research stream MARKETING LETTERS 23 55 0.0000626 2004 Johnson et al. On the depth and dynamics of online search behavior MANAGEMENT SCIENCE 22 206 0.0000621 2003 Chatterjee et al. Modeling the clickstream: Implications for Web-based advertising efforts MARKETING SCIENCE 20 167 0.0000574 2005 Van den Poel et al. Predicting online-purchasing behaviour EUROPEAN JOURNAL OF OPERATIONAL RESEARCH 13 89 0.0000538 2009 Bucklin et al. Click Here for Internet Insight: Advances in Clickstream Data Analysis in Marketing JOURNAL OF INTERACTIVE MARKETING 13 76 0.0000544 2005 Senecal et al. Consumers decision-making process and their online shopping behavior: a clickstream analysis JOURNAL OF BUSINESS RESEARCH 10 71 0.0000508 2014 Li et al. Attributing Conversions in a Multichannel Online Marketing Environment: An Empirical Model and a Field Experiment JOURNAL OF MARKETING RESEARCH 10 62 0.0000514 2006 Moe et al. An empirical two-stage choice model with varying decision rules applied to Internet clickstream data JOURNAL OF MARKETING RESEARCH 10 47 0.0000495 2009 Huang et al. Searching for Experience on the Web: An Empirical Examination of Consumer Behavior for Search and Experience Goods JOURNAL OF MARKETING 9 201 0.0000498 Table 4 Extracted Topics from STM Topic (Prevalence (%)/ Average Citations/Relative Citation*) Top words Shopping Experience and Satisfaction (6.86/20.12/2.93) consumer, study, intention, satisfaction, experience, shop, examine, usage, positive, understand Online Behavior Pattern (7.68/17.89/2.33) online, use, time, pattern, base, propose, choice, good, framework, across Privacy in Digital Community (7.91/21.58/2.73) information, social, datum, privacy, digital, design, content, internet, socialmedia, change Information Search (7.91/20.40/2.58) consumer, information, online, search, price, user, behavior, internet, use, website Challenges in Retail Markets (8.19/21.32/2.60) market, bigdata, research, retail, challenge, store, author, management, source, potential Quality of Customer Experience (7.72/16.50/2.14) customer, service, quality, product, mine, approach, buy, strategy, algorithm, loyalty Online as a Sales Channel (6.50/16.30/2.51) online, study, relationship, sale, behavior, channel, finding, demand, role, literature Analysis of Social Networks (6.91/14.23/2.06) datum, analysis, mobile, user, network, method, customer, paper, sentiment, text Brands and Online Reviews (7.19/18.79/2.61) product, use, review, brand, impact, predict, show, research, result, variable Purchasing Behavior (6.80/15.03/2.21) purchase, behavior, model, customer, website, decision, visit, ecommerce, market, provide Data Analytics and Food (7.16/16.94/2.37) datum, analytics, system, bigdata, business, value, process, new, food, approach Advertising Performance (6.61/18.80/2.84) model, effect, use, advertise, performance, result, firm, increase, purchase, web Perceived Attitudes Towards Companies (6.16/14.89/2.42) model, attitude, perceive, company, industry, test, mobile, toward, retailer, socialmedia Digital Age and Services (6.42/15.08/2.35) technology, service, trust, knowledge, health, study, paper, collect, app, factor *Relative Citation = Average Citations/Prevalence Figure 1. Top Terms Figure 2. Word Cloud of All Literatures Figure 3. Term Frequency Over Time Figure 4. Keyword Occurence Figure 5. Keyword Citations Figure 6. Number of Publications Between 2000 and 2018 Figure 7. Relative Publication Volume