Keywords

1 Introduction

Customer experience (CX) has become a central concept for many businesses. Experience became a critical element when purchasing a product or service [1, 2]. Hence, creating positive experiences allows brands to attract and retain customers [3]. The measurement of customer experience plays a key role in making insights actionable for firms and helps them remain competitive. Practitioners and academics have developed several tools and methods to understand and manage the customer experience [4]. These tools and methods focus mainly on the development of visualization and prototyping techniques as well as the implementation of measurement tools [4, 5]. The practice of CX management is characterized by the processes used to monitor and organize the series of interactions between the organization and its customers. Thus, visualization and prototyping methods allow to identify key moments of the customer experience that, once addressed, will provide a more valuable global experience.

Customer journey mapping [6, 7], service blueprinting [8, 9] and customer experience mapping [10] are among the most frequently used visualization methods. These techniques are used to visualize and structure the process that the consumer works through during their interaction with the firm [8]. Most studies use the service blueprint methodology, and few go beyond this approach in order to analyze the customer journey [8, 11, 12]. There is a need to improve customer experience measurement tools by focusing on qualitative techniques, such as narrative data collection, to seek more detailed insights [13]. All in all, progress is needed in customer journey mapping in order to identify key opportunities to influence customer experience.

By deepening the understanding of the customer experience, it strengthens the understanding of overall customer satisfaction. New technologies and data represent interesting assets to be integrated in the mapping process. Researchers have begun to incorporate neuroscientific measures (e.g., emotions and cognitive load) in order to track more precisely concurrent measures of customer experience [14,15,16]. Scholars and practitioners commonly agree that the customer experience includes cognitive, emotional, behavioral, sensorial, and social dimensions [17,18,19]. Measures such as biometrics, eye tracking and Electrocardiography (ECG) help to better understand how the customer experience is formed and the series of events that led to the overall level of satisfaction [14]. By understanding how affective and cognitive variables influence the customer in his/her journey, it contributes to identifying key drivers to overall customer satisfaction. Our study aims to propose an approach that presents, in conjunction with the actual journey of a user on a web interface, his/her implicit cognitive and emotional states along this journey. Users can deviate from the predicted path, and identifying these deviations can help select the source of intervention or influence needed on a given interface to prevent users from deviating [11]. To illustrate the relevance of the developed method, we present a methodology based on a case study of a given task and the results obtained using the visualization technique. The proposed method is discussed and validated by a group of User Experience (UX) experts to ensure its relevance and usefulness.

2 Literature Review

2.1 The Customer Journey

The analysis of a customer’s journey was first used in the areas of service management and multichannel management [8, 20]. Customer journeys are the most popular visualization techniques within the domain of service design [21]. It refers to a series of events that the consumer goes through in order to be informed, to buy or to interact with a given organization [22]. It is a visual and chronological representation of the events experienced by the user [23]. The purpose of such approach is to simulate a “walk in the customer’s shoes” [24]. The customer journey analysis allows to identify critical touch points throughout the customer journey that have the most significant impact on customer outcomes [11, 25]. The goal of such technique is to improve customer interaction by enhancing each touch point between the customer and the organization. A study by Lemon & Verhoed [11] states that the understanding the customer journey leads to better insights about the customer experience. Indeed, the customer experience encompasses the customer journey with a given organization over time through multiple touch points [11]. Thus, the customer journey approach allows analyzing the touchpoints between the organizations and its customers.

Customer journey mapping (CJM) is among the most frequently used visualization methods of the customer experience [6, 7]. This method is a visual representation of the customer experience using a given service [5, 26]. This map is a consolidated overview of all sequences of events through which customers must pass through to complete their purchase [12]. CJM establishes all touch points between the firm and its customer during a given purchasing process. The aim of such technique is to improve the quality of the overall customer experience by improving the customer experience associated with each touch point [12]. Mapping the customer experience offers information from the customer’s perspective and represents a great way toward operational improvements [27].

Mapping is a commonly used tool in the service design methodology [26]. However, A study by Rosenbaum [12] mentions that CJM assumes that all customers of a given organization experience the same touch points and assigns the same importance to each of these touch points. Hence, the understanding of the participants’ overall experience does not necessarily reflect discreet experiences with specific phases, events, or activities. Currently, there is a lack of methods that combine the users’ behaviors with their affective and cognitive states. To date, some studies have incorporated neuroscientific data into the analysis of the customer journey in order to have a richer understanding [14,15,16].

2.2 Measuring Emotional and Cognitive Responses

Psychophysiological data are physical signals measured in real time, which are generated due to psychological changes [28]. These measures allow researchers to assess the user’s reaction to a specific stimulus [29]. Several tools are available to measure psychophysiological measures, such as electrocardiography (ECG), skin-based measures including electrodermal activity (EDA), ocular measures, brain measures including electroencephalogram (EEG), respiration rates, and blood pressure (Charles & Nixon [79]). Furthermore, psychophysiological measures are unobtrusive, which allows collecting data about the user’s experience without affecting their decision-making [28, 30,31,32]. These measures therefore allow for more natural reactions from the user while also offering uninterrupted reports of emotions [31, 33, 34]. The advantage of having an uninterrupted report of emotions is that it gives access to the unconscious emotional reactions of users [35]. Indeed, research has shown that interrupting users while they are completing a task often leads to biased results [34, 36,37,38]. Psychophysiological measures offer a more complete view of the human-computer interaction and allow to assess events of cognitive and emotional relevance to the users [39,40,41,42].

In the UX literature, emotions are often defined according to two complementary dimensions, namely valence and arousal [43, 44]. The circumplex model of affect is a very popular dimensional approach to defining several emotional states according to these two psychophysiological constructs [45]. In addition, valence varies from pleasure to displeasure [30, 46, 47], which is equivalent to a variation from positive to negative emotions [48, 49]. Valence can be defined as “what a user feels” and the most reliable way to measure such constructs is with facial coding [50]. Individuals tend to express their emotions with micro-movements of the facial muscles [51]. Arousal is the second and complementary dimension to the circumplex model of affect. This construct refers to “level of arousal which ranges from calm to excited” [46, 49]. Thus, psychophysiological measures translate emotions into measurable constructs.

Moreover, cognition, which refers to the process of reasoning and the mental effort required to understand and complete a given task [52] is also a measurable physiological construct [53]. Several studies have shown that pupil dilation is a significant measure of cognitive load [54,55,56,57,58,59]. Indeed, the degree of the pupil’s dilation correlates with the workload of the task [54]. Hence, pupil size offers an immediate measure of cognitive effort.

A study by Gentile et al. (2007) suggests that customers interpret information from the interface from a cognitive and affective point of view which leads to a personal impression of the given website. Epstein’s Cognitive Experiential Self-Theory (CEST) suggests that there are two systems which operate in parallel when exposed to a stimulus event. These two systems are affective and cognitive. Thus, combining cognitive and emotional measures provides a comprehensive understanding of the user’s experience. The customer journey map illustrates every touchpoint the consumer has with the company in order to complete a given task [60]. Hence, by adding affective and cognitive measures to the customer journey, it allows UX researchers to identify critical interface elements influencing the user’s journey. Thus, the convergence of these three perspectives (i.e., emotions, cognition and behavior) affords a complete view, which few methods allow for [61]. Therefore, we pose the following proposition:

P1: Enriching customer journey visualizations with information about the cognitive and emotional states of the user facilitates the identification of critical touch points shaping customer experience.

3 Method

The study is presented in two phases and consists of a collection of user tests and a focus group. The user tests were performed to collect data for the purpose of developing an illustrative case of the customer experience of all users. The focus group was conducted to validate the research proposal. First, we collected psychophysiological data about each user’s experience in order to track their cognitive and emotional responses during an e-commerce interaction. The goal was to develop a method which triangulates behavioral data with psychophysiological data in order to gain key insights into the user journey and to produce a complete overview of the user’s experience. Thus, we present a new visualization method that contextualizes the user’s physiological and behavioral data while interacting with a web-based information system in the financial services industry, an example of FinTech. We illustrate the methodology by presenting in detail one task and the results obtained using the visualization method. Second, the new visualization method is reviewed by a group of experts in order to confirm whether it enriches contemporary understanding of critical touch points shaping customer experience, and therefore offers support for the research proposition.

Phase 1

3.1 Participants

A total of 38 participants (42% female) were recruited in this study and their age ranged from 23 to 62 years old. Prior to the study, participants were asked to rate the financial institution on the Net Promoter Score (NPS) Scale [62]. NPS is a one-question metric used to assess a customer’s overall perception of a brand using a Likert scale from 0 to 10, with 0 being not at all likely to recommend the brand, and 10 being extremely likely to. To reduce the potential influence of brand equity (either extremely high or low) on the participants’ experience, only participants who rated the financial institution between a score of 3 to 8 were invited to participate in our study. Each participant received a moderate financial compensation to participate. To participate, users needed normal vision and were screened-out for glasses, laser eye surgery, astigmatism, epilepsy, neurological and psychiatric diagnoses.

3.2 User Test Procedure

The test sessions took place in a university usability laboratory based on standard practice for UX enriched with psychophysiological measures [63]. Upon arrival, participants were asked to complete a consent form, after which they were informed about the purpose of the study and fitted with physiological sensors. Participants were told that their participation was requested in order to evaluate a financial website. Each key segment of the financial institution (e.g., early retirees, midlife adults, youth, and entrepreneurs) was assigned a set of tasks specific to their group. To control the task type, each participant was assigned to a total of nine tasks completed frequently on the website. The nine tasks were grouped evenly into three main categories representing frequent actions taken by users: searching for specific information on the website, using a tool to find answers, and find ways to contact the financial institution. This approach minimizes the effect of the task type by having a varied number of different tasks. Participants were asked to complete tasks for which an optimal navigation path was first established by the authors vis-à-vis the smallest number of clicks necessary to complete a task.

3.3 Measures

For the user test, data were collected using a variety of non-intrusive tools. Behavioral data were captured from the recording of the screen interface allowing the user’s actions to be monitored. A Tobii X-60 eye-tracker (Stockholm, Sweden) sampled at 60 Hz was used to capture participant eye movements and Tobii Studio was used to record the experience. The screen recording along with the eye-tracking device allows having the user’s perspective by identifying precisely where the participant was looking at every second [64]. Moreover, pupil size measurement allowed for tracking each user’s cognitive load [65, 66]. Arousal was measured using electrodermal activity (EDA) [46, 67, 68] with the AcqKnowledge software at MP150 sampled at 500 Hz (BIOPAC, Goleta, USA). EDA represents an indication of the variation of physiological arousal [69]. In fact, it measures the skin conductance, which measures current flow between two points of skin contact after an electrical potential has been applied to them [68]. Hence, it indicates cognitive and emotional stimulation throughout the experience [70]. Furthermore, valence was measured using facial expressions [51, 71] performed with the FaceReaderTM Software (Noldus, Wageningen, Netherlands). Indeed, this tool is used to observe facial reaction and assess emotional valence with a score ranging from -1 (most negative emotions) to +1 (most positive emotions) and a temporal precision of 30 inferences per second [51, 71]. More specifically, the valence score is calculated as the intensity of a happy score minus the intensity of the highest negative responses [72]. Then, Observer XT and CubeHX [73, 74] were used to synchronize all the signals between Acknowledge and FaceReader (Noldus, Wageningen, Netherlands) as recommended by Léger et al. [44].

3.4 Preparation of Visualization Models

In order to triangulate collected data, webpages visited are first grouped and assembled to construct each user’s journey consisting of its various navigation paths. Using the eye-tracking device and Tobii Studio, these tools help contextualize the behavioral data of users and group all the webpages used for a given task. This compilation leads to a visual representation of all paths used by the sample for a given task. Moreover, using the complete psychophysiological dataset generated by all users, an average of each measure (i.e., arousal, valence, and cognitive load) was calculated for each webpage; this generated an overview of the page-by-page evolution of the users’ cognitive and emotional states. Moreover, the averages for each measure are grouped by webpage, tasks and segments to get an overview of an average customer journey for a given segment.

In order to facilitate the understanding of all collected data, the averages obtained from both cognitive and emotional measures are translated into symbols and visual codes according to pre-established thresholds. Arousal, emotional valence and cognitive load are rescaled to (−1 to +1) and adjusted to the baseline. The baseline allows distinguishing the user’s psychophysiological sensitivity since each participants’ emotional reaction to a stimulus does not have the same direction and intensity [34]. Hence, the baseline allows participants’ emotional responses to be standardized on the same scale.

First, the user’s emotional responses are translated into four main categories of the circumplex model: enthusiasm, frustration, serenity and tolerance [75]. Indeed, all affective states are characterized by two fundamental neurophysiological constructs which are valence and arousal [45]. Consequently, Russell (1989) proposed the Affect Grid, which is a scale based on the circumplex model of affect in order to describe emotions according to the crossing of the dimensions of valence on the horizontal axis and arousal on the vertical axis (see Fig. 1). The center of the grid represents a neutral valence and a medium level of arousal. For example, when the valence value is below 0, it refers to the lower part of the Affect Grid (i.e., tolerance and serenity). Moreover, when the arousal is below 0, it refers to the left side of the Affect Grid (i.e., frustration and tolerance). Thus, when the arousal and emotional valence are below 0, the user ends up in the quadrant of tolerance. With the Affect Grid, it is possible to analyze whether the participant feels pleasant/unpleasant or active/inactive. Hence, the symbolization of emotions on the visualization models respects the coding of the Affect Grid. Therefore, each quadrant (i.e. frustration, enthusiasm, tolerance and serenity) is represented by an icon of a facial expression. Hence, it is easy to visually recognize what emotion is felt on average by users for each webpage.

Fig. 1.
figure 1

The Affect Grid

Full size image

Second, cognitive loads are categorized into three levels such as easy, moderate or difficult and these levels are represented using a gear icon (e.g., one gear symbolizes a relatively easy task whereas three gears represent a difficult task). Since the arousal variable varies from −1 to 1, the difference is divided into three equal parts to distinguish the different levels of difficulty. Given that the arousal values are rescaled between −1 and 1, a cognitive load value of less than zero (i.e., inferior to the baseline) means that cognitive effort is low. Thus, cognitive loads were divided into three main categories which are low cognitive load (< −0.3), moderate cognitive load (between 0.3 and −0.3) and high cognitive load (>0.3) (See Table 1).

Table 1. Thresholds for Psychological Data.
Full size table

The thresholds for cognitive loads and emotional responses are presented in the table below (see Table 1). Such visual codes allow simplifying the visualization of the page-by-page evolution of the user’s cognitive and emotional states.

Phase 2

3.5 Focus Group Procedure

Once all tasks were completed, two visualization models were created using psychophysiological data with the objective of representing the overall experience of users. These two models were then discussed in a focus group with UX experts. Experts are impartial individuals selected on a voluntary basis, who understand usability and have experiences in the analysis of the user journey. The focus group is a collective interview where participants meet to discuss a defined topic [76]. Since our approach aims at facilitating the UX expert’s analysis of the customer journey, the focus is on the acceptance of this system by ensuring it meets UX experts’ needs. The type of focus group selected is called “prioritizing functionalities”, as the main objective is to identify the most attractive functionalities for UX experts in order to guide and optimize the method design.

The focus group procedure lasted 90 min from the reception to the closing of the interview. A total of six (6) UX experts participated. The number of participants in a focus group is recommended to be between six and 10 individuals [77], as a smaller number of participants promote interdependence among members [78]. First, participants were informed about the purpose of the focus group, the activities planned and their roles as UX experts. Then, UX experts had to evaluate two visualization models for which they were asked a series of questions. UX experts used the technique of associating ideas to express what first comes to mind when looking as the visualization model [77]. UX experts were then asked to state their analysis of the customer journey. By stating their understanding, it would reveal whether the model represents an appropriate and relevant analysis tool. Afterwards, UX experts were asked to elicit both strengths and weaknesses for each visualization scheme in order to identify which features to keep or remove. Lastly, following the evaluation of both visualization models, UX experts were asked to define what an ideal visualization scheme should be according to the two models shown previously. Once the focus group was finished, the visualization model chosen by UX experts as the best option is modified according to their recommendations and sent back to UX experts in order to obtain their final impressions.

4 Method

4.1 Customer Journey Visualization

Two models were designed and discussed in this study. Each model is a distinctive visual representation of a task using a variety of pre-established visual codes and symbols. This study aims at comparing both models in order to define the optimal approach to understand the concurrent user experience. Through the representation process, there are seven key elements found in each model which support the representation of customer journeys. These elements are presented in the table below and the visual representation of some elements vary between both models (see Table 2). Indeed, the representation of the optimal path, which refers to the optimal navigation vis-à-vis the smallest number of clicks necessary to complete a task, is represented differently between the two models. The optimal path layout varies across models. It is not aligned or centered in the first model but is in the second model. Also, individual customer journeys are displayed on the second model while the first model does not distinguish between each customer journey.

Table 2. Elements found in each visualization model.
Full size table

The first visualization scheme is presented in Fig. 2. All pages are aligned in order to provide a structured representation of the different pages used. The most problematic pages can be identified with the indicators of emotional value and cognitive effort. Indeed, these pages are shown in red and represent frustration (key located in top-left box) and high cognitive effort (top-right box) respectively. Moreover, traffic and used paths are indicated in blue to emphasize this information. In short, this model aims to highlight travel and psychophysiological data.

Fig. 2.
figure 2

First Visualization Model. (Color figure online)

Full size image

The second visualization model is presented in Fig. 3. Unlike the first model, the journeys are represented in the form of swim lanes, where psychophysiological data (i.e., cognitive load, emotional valence, and arousal) are associated to each webpage visited. The most problematic areas are identified using the same indicators of emotional valence and cognitive effort. Indeed, the problematic areas are illustrated by the red zones and represent frustration (left zone) and high cognitive effort (right zone). In Fig. 3, several pages seem to be problematic regarding negative emotional valence. However, no page seems to require high cognitive effort. Since the paths are linear, it is possible to see where participants have deviated from the original goal and where problems have emerged. For example, most users used the link at the bottom of the homepage to go to the “Savings” page, deviating from the expected route.

Fig. 3.
figure 3

Second Visualization Model. (Color figure online)

Full size image

4.2 Results of the Focus Group

Regarding the first visualization model, the UX experts mentioned that the beginning of the experiment is not clearly indicated and thus, the starting point is difficult to identify. Moreover, the fact that the optimal path is not aligned and centered makes the reading of the customer journey less intuitive. In general, UX experts find this model not sufficiently refined and the understanding of the users’ experience is not obvious. According to the participating UX experts, this model seems suitable for client reporting through a matrix format. This first visualization model allows to identify problematic areas on the interface.

Regarding the second visualization model, the UX experts mention that the model allows for a better understanding of the sequence experienced by users. Indeed, as mentioned previously, the first model does not distinguish between each journey. Thus, UX experts are more likely to be interested in physiological data than in the context surrounding the use of the webpage. In short, the second method allows us to have more context around the physiological data in order to prioritize the source of intervention on the interface to improve the overall experience. For example, the “log in” page seems problematic according to the first model, because it indicates frustration. However, looking through the second model, it is possible to see that this page has only been viewed by one participant. Hence, this frustration can be caused by several factors and the “log in” page may not be a priority when updating the website. Therefore, UX experts suggest that model 2 represents a good tool to have a summary of the situation and can also serve as a benchmark to measure progress through improvements made on the website.

The results of the Focus Group suggest that the research proposition be accepted since the experts mention that the visualization models facilitate the identification of problematic areas on the interface. Indeed, by bringing together the cognitive and emotional data of the users, it allows for an enriched customer journey and a more complete understanding of the associated experience.

5 Method

Our results suggest, through an illustrative case, that the use of the proposed method facilitates the identification of critical touch points to customer experience. This model is useful in order to have an overview of the experience and different paths taken by users. It allows one to see the problematic webpages when diagnosing the customer journey. Following the focus group, the UX experts agreed that the second model is the more appropriate tool for customer journey analysis. The visual representation has been described as clear and simple to use. A UX expert mentioned that “the red color makes it easier to read problematic pages and the optimal path is well highlighted without cluttering up the parallel paths” (UX expert 5). Moreover, this model has been selected because it has several strengths. First, this model is self-supporting, which means that it presents a “simple and quick visualization, without the need for explanation” (UX expert 3). Secondly, the information presented is clear, thanks to “a clear legend and the use of icons and colors that are easy to understand” (UX expert 2). It is easy to identify successes and failures as well as areas of frustration or high cognitive load. Finally, this model also allows to follow the customer journey of each user thanks to the color tablets. It gives an overview of the experience while differentiating the various paths. This model is also useful for understanding the main errors made during experience sequence (i.e., to understand where participants deviated from the optimal path). Simply put, the model allows to quickly identify where participants are getting lost.

However, the model also has a limitation. Indeed, it is easy to analyze for a certain number of participants. The visual representation represents the paths of 8 participants. However, the larger the number of participants, the more complex the visual representation becomes. The use of color tablets allows the distinction between participants, but leads to limits in terms of sample size.

Our study contributes to the UX literature by presenting a comprehensive method allowing for the visualization of a user’s emotions and behaviors during his/her navigation on a website. Thus, this study adds to the literature which focuses on the modelling of the consumer decision journey [12]. Additionally, it provides more precision into the analysis and the interpretation of results, aiding to reveal problematic areas on the interface. This method serves as a complementary approach to other methods available such as questionnaires and interviews that enable collection of self-reported data and adding the convergence of these three perspectives (i.e., emotions, cognition, and behaviors), which few methods allow for [61]. The relative simplicity of this method, which enables the visual representation of the evolution of a user’s cognitive and emotional states throughout their online journey and experience, should be particularly useful to both UX researchers and practitioners. Modelling service delivery from the customer’s perspective is an important topic for service providers seeking to improve their services [23].

Our method allows contrasting several user journeys against the planned journey. We contend that most customer journey maps can potentially be critically flawed. They assume all customers of an organization experience the same organizational touchpoints and view these touchpoints as equally important [12]. In contrast, our method provides an accurate report of several user journeys. The results also pose managerial implications. First, this new method allows both practitioners and researchers to identify psychophysiological pain points on a webpage easily and the visualization helps to analyze and interpret results more efficiently [79]. Moreover, our method is useful for comparing user experiences on various interfaces, which can be used to compare the user experience of a specific task on competing interfaces.

Furthermore, some limitations need to be acknowledged. First, the method focuses on one section of the website. Hence, the evaluation of the customer experience does not reflect that of the entire website but depends on the selected task. Second, this model is limited to a certain number of users. As there were only 8 individuals user journey studied, this is not a large-scale study, mostly due to the high cost of obtaining the data. As the number of users studied increases there will be different user journeys. In this way, the visualization becomes more complex to analyze due to the large amount of information. Thus, it would be interesting to study the possibility of creating segments from user navigation when the sample size is too large to distinguish each path using a colored pad.

6 Conclusion

Our results show that the models proposed adequately identified and interpreted psychophysiological data supporting an understanding of the user’s experience. By representing the various customer journeys through the webpages visited, it makes it possible to have a real report of the experience lived by users. Using this new visualization method generates a complete overview of users’ experience and produces key insights. Indeed, it facilitates the understanding of the evolution of the experience since it shows critical touch points of the interface where the user experienced a given cognitive and emotional state. It also helps to identify the main differences between the planned customer journey and the user’s decision-making. This method serves as a complementary approach to other methods available such as questionnaires and interviews that enable the collection of self-reported data and adds the convergence of these three perspectives (i.e., emotions, cognition, and behaviors), which few methods allow for [61].