ONLINE FLOW AND E-SATISFACTION IN HIGH INVOLVEMENT PURCHASING PROCESSES

FLUXO ONLINE E E-SATISFAÇÃO EM PROCESSOS DE COMPRA DE ALTO ENVOLVIMENTO

Internet access became easier because of improvements in the infrastructure of telecommunications (Wachter et al., 2000). Consequently, the e-business moved forward and became more evident.

Electronic business, also known as e-business, can be defined through the use of communication technologies, among which stand out those related to the Internet, to interconnect and support business processes, and to communicate and collaborate within an organization and between it and its customers, suppliers, and other partners (Laudon and Laudon, 2010; O’Brien, 2004).

Among the different electronic business arrangements described by Gonçalves and Ferreira (2010), e-commerce is noteworthy due to the significant growth achieved by this business format in recent years. The growing competition in e-commerce is inducing more and more companies to seek the satisfaction of consumers, which, in turn, reflects the continuity of the relationship, in loyalty, favorable attitudes and positive word-of-mouth advertising (Rossi and Slongo, 1998).

The satisfaction of e-consumers or e-satisfaction results, according to Dholakia and Zhao (2010) e Bansal et al. (2004), from several factors, many of them associated with attributes of online shopping experience, which are evaluated before, during and after the finalization of the electronic purchase transaction. Those same authors argue there is no concordance between researchers about which of these attributes are more relevant to formation of e-satisfaction.

However, besides the attributes of shopping experience, there are personal and psychological aspects influencing the e-consumers behavior, including online flow, which is a state of deep immersion reached by an individual during online purchase process (Hoffman and Novak, 1996). Studies show that flow state in e-commerce increases attention, improves attitudes and increases purchase intent in a particular website (Hoffman and Novak, 2009; Van Noort et al., 2011). Nevertheless, there are no studies linking the online flow to e-satisfaction yet.

Starting from the lack of agreement among researchers about which attributes of online shopping experience influence e-satisfaction, as well as the absence of studies linking online flow of e-satisfaction, was defined as general goal for this research verify the relation between the attributes of online shopping experiences, the state of online flow and e-satisfaction in high-involvement purchases.

The importance of this work lies in three key points: (i) the relevant expansion of e-commerce in Brazil, coupled with the trend of future growth; (ii) the issue of importance of customer satisfaction in the context of e-commerce; and (iii) the lack of scientific studies that discuss this topic associated with consumer involvement and online flow.

To achieve the research goal, we collected quantitative primary data by conducting a field survey. A structured questionnaire was used, composed with questions in a structured scale format. The elaboration of the questionnaire was based on the research hypotheses.

At the beginning of questionnaire, was included a solicitation to respondents answer the questions thinking in a purchase of high involvement and a brief explanation of what was considered a high-involvement purchase. It was adopted the perspective of Solomon (2008), that high involvement of consumer is most common in situations characterized by a high cost of the product to be purchased and non repetition of purchase.

The questionnaire was divided in three parts. First, theoretical constructs identified in the literature review as antecedents of e-satisfaction (attributes that are part of the online shopping experience and influence e-satisfaction) gave rise to a set of one to four questions or statements which were included in the questionnaire, which can be explained by the fact that they are multidimensional and include more than one aspect. Second, the construct "online flow" gave rise to three statements, which also can be explained from its multidimensionality. Finally, the construct "e-satisfaction" gave rise to five statements, considering five complementary theoretical perspectives were identified that can be used to measure it (Ceribeli, 2015).

The scale used to structure the questionnaire was Likert seven-point, whose extremities were "totally disagree" and "totally agree". We chose a scale of seven points, considering that respondents tend to avoid extremes, i.e., five-point scales can become, in practice, three-point scales (Hair Junior et al., 2005).

Before proceeding to data collection, it was conducted a pretest with 10 individuals in order to identify and correct ambiguities and other problems in the drafting of the questions. Based on the results obtained, two questions were rephrased. After that, a new pretest was conducted and no additional problems were pointed.

The questionnaire was circulated on social networks by researchers. To increase the response rate, some individuals were selected to disseminate the questionnaire at the same time as researchers (these individuals agreed to publish the link to questionnaire every two days during the period of one month. Thus, disclosure of the questionnaires was repeated 15 times in the aforementioned period, both on Facebook and Linked In.

For this research, the target population was defined as all inhabitants of the states of São Paulo and Espírito Santo, between 18 and 49 years, who carry out Internet purchases. Choosing the States of São Paulo and Espírito Santo reflects the researcher's ease of access to sample units. We opted for non-probabilistic sampling by quotas, which considers the existence of subgroups in the target population and is based on distribution of these subgroups (Hair Junior et al., 2005). To implement the sampling quotas in this study, the target population was stratified using as criteria the age group of consumers.

The percentage of each group of e-consumers in the research sample was defined from data of E-Commerce (2013) transformed to the base 100. To ensure that the population strata were represented in compliance with the defined percentage, when the pre-stipulated amount of answers to specific stratum was reached new responses for the same stratum were excluded.

To set the sample size, we followed the recommendations of Hair Junior et al. (2009), who suggested that when we adopted multivariate statistical techniques to analyze the data collected, the sample size should be 10 to 20 times the number of variables included in the survey.

The questionnaire includes 42 questions (research variables), so research sample consisted of 720 respondents, of which 10 were excluded due to errors in filling out the questionnaire. Thus, 710 responses were validated, which represents approximately 98.6% of the responses collected.

In data analysis, we used Structural Equation Modeling (SEM), merging the approaches Pathway Analysis and Confirmatory Factor Analysis. The choice for SEM is justified based on the existence of more than one dependent variable (online flow and e-satisfaction). The results obtained from the SEM were analyzed as follows: first, we examined for the existence of transgressive estimates (negative variances of error, surplus standardized coefficients or very close to 1.0, or standard errors very large); then we analyzed the overall fit of the model obtained using absolute fit measures, incremental and parsimonious; finally, we interpreted the coefficients obtained based on the theory previously consulted to validate or refute the research hypothesis.

In order to facilitate analyses, each variable received a label, as shown in Chart 1.

Chart 1.
Labels of research variables.

After carrying out a Skewness and Kurtosis analysis, it was found that only the variable CUSTOM stands out by presenting an asymmetrical behavior and a kurtosis higher than the recommended value, which may indicate the absence of normality. To confirm the existence of normality or not, we made the LM test proposed by Jarque and Bera (1987). From the results of LM tests for each variable, it was possible to confirm the analysis of skewness and kurtosis made earlier, i.e., only the variable CUSTOM did not follow a normal distribution; the other variables met the assumption of normality. Therefore, we opted to exclude the CUSTOM variable.

After the analysis of the data normality, the assumptions of homoscedasticity, linearity, and absence of autocorrelation, were tested. First, we used graphic analysis of the standardized residuals versus the predicted dependent values, as recommended by Hair Junior et al. (2009) and Gujarati (2006). To complement the graphical analysis, we used the test d of Durbin-Watson and Pesaran-Pesaran test. After testing basic assumptions for multivariate analysis, we decided to delete three of the 42 original variables of the study due to lack of linearity (SERV1, TRUST2 and VARIE2).

For statistical analysis of data, we used Structural Equation Modeling (SEM) combining Pathways Analysis and Confirmatory Factor Analysis. Confirmatory Factor Analysis was used to group the observed variables of the survey, resulting in a smaller number of latent variables. In turn, Pathways Analysis was used to analyze the relationship between the variables that measured online shopping experience and the latent variables "online flow" and "e-satisfaction" as well as the relationship between these two variables.

Whereas the Confirmatory Factor Analysis creates latent variables from the original variables, it was necessary to verify, prior to analysis, whether it would be appropriate to aggregate the original variables or not. For this, we used Bartlett's test of sphericity, associated with the KMO (Kaiser Meyer Olkin) and Cronbach's Alpha.

Simultaneously analyzing the KMO coefficients, the calculated values of Cronbach's Alpha, and the p-values of Bartlett's test of sphericity, we identified only three sets of variables that could effectively be grouped by Confirmatory Factor Analysis, resulting in three latent variables (INNOV1 and INNOVA2; FLOW1, FLOW2 and FLOW3; SATIS1, SATIS2, SATIS3, SATIS4 and SATIS5). The remaining variables (all of them representing different attributes of online shopping experience) were placed on Structural Equation Modeling individually.

After defining which variables were grouped by Confirmatory Factor Analysis, we structured through AMOS software, a first structural equation model, which included three latent variables (INNOV, FLOW and SATIS), being two endogenous (FLOW and SATIS) and one exogenous (INNOV), and twenty-eight observed exogenous variables. It should be mentioned that the decision to include observed variables related with attributes of online shopping experience instead of a unique latent variable to represent this construct is justified by the possibility of identify specifically which attributes influence the state of flow and which influence e-satisfaction.

For our estimation procedure, we opted initially to use the method of Generalized Least Squares, commonly used in Structural Equation Modeling. However, when a Mardia coefficient was obtained that went beyond the acceptable upper limit of 1.96, indicating possible absence of multivariate normality, we opted for the Asymptotically Distribution-Free method, due to the fact that this does not seem sensitive to absence of multivariate normality of data (Hair Junior et al., 2009). This first model had generated problem identification, i.e., the AMOS software identified that the structured model had no ability to generate single estimates.

To solve the identification problem encountered, we developed a simpler model (with only three variables: INNOV, FLOW and SATIS), which was used as a starting point to perform various tests and gradually include the twenty-eight observed exogenous variables. At the end of testing process, we maintained the three latent variables (being two endogenous and one exogenous) and only fourteen observed exogenous variables.

The quality of fit obtained was examined through three measurements. The root mean square error of aproximation or RMSEA was 0.056, lower the than borderline recommended of 0.08 (Garson, 2012; Arbuckle, 2010; Hair Junior et al., 2009). The adjusted goodness-of-fit index or AGFI was 0.813, close to 1.0, that is the value that points perfect adjusts (Garson, 2012; Arbuckle, 2010; Hair Junior et al., 2009). Finally, the normed Chi-square was 3.247, lower than the borderline recommended of 5 (Schumacker and Lomax, 2004; Joreskog, 1970).

After validating the quality of the fit of the proposed model, we validated latent variables included in the structural equation model. After validating the latent variables created, we analyze the regression coefficients of the relationship between latent endogenous variables FLOW and SATIS and the exogenous variables. These coefficients are presented in Table 1. The structural equation model was able to explain 26.6% of the variation in the latent variable FLOW and 50% of the variation in the latent variable SATIS, showing moderate explanatory power by the exogenous variables incorporated into the modeling.

Table 1.
Regression coefficients of the proposed structural equation model.

To check the statistical significance of the regression coefficients, we tested the null hypothesis of beta is equal to zero for each of the proposed relationship between endogenous and exogenous variables of the model. At a significance level of 0.05, we rejected the null hypothesis for the proposed relations between the following variables: INNOV and FLOW (b = 0,408), TRUST3 and FLOW (b = 0,199), VARIE1 and FLOW (b = 0,140), ESTHE2 and FLOW (b = 0,127), PRICE2 and FLOW (b = 0,113), FLOW and SATIS (b = 0,688), CONV1 and SATIS (b = 0,231), VARIE1 and SATIS (b = 0,190), TRUST3 and SATIS (b = 0,119), DELIVER1 and SATIS (b = 0,105), and POLIT1 and SATIS (b = 0,077).

From the performed hypotheses tests and beta coefficients analyzed, we found that the variables which positively influence the state of online flow of consumers in high involvement purchases are: (1) the innovation of website, (2) the reliability assigned to the electronic retailer, (3) the variety of the assortment available on the website, (4) the aesthetics of the virtual environment, and (5) the prices charged by the company.

Furthermore, we found that the variables which positively influence online consumer satisfaction in high involvement purchases are: (1) online flow state, (2) the convenience associated with the website navigation, (3) the variety of assortment available on the website, (4) the reliability attributed to the electronic retailer, (5) the fulfillment of delivery deadlines, and (6) the existence of exchange and return policies perceived as fair.

Based on these results, a model was proposed to explain online flow and e-satisfaction in high-involvement purchases, shown in Figure 1.

Figure 1.
Model of e-satisfaction and online flow in high involvement purchasing.

First analyses of the state of online flow in high involvement contexts highlights the importance of consumer perception that the website design is innovative and differentiated in relation to others, corroborating the results of studies conducted by Congwen et al. (2010) and Huang (2003).

The relationship between the website innovation and online flow can be explained by one dimension of online flow proposed by Huang (2006) and Congwen et al. (2010): the consumer curiosity during navigation. If website is different and innovative, naturally curious and engaged the consumer will be. Besides, if innovation it's seen in the website, the consumer will explore it with more concentration. Both behaviors are pointed out by Van Noort et al. (2011) as a characteristic behavior of consumers who have hit the state of online flow.

Regarding the influence that the level of reliability assigned to the website has on the state of flow, we identify that when a individual feels that a certain electronic retailer is reliable and believes he would not act in bad faith with his customers, there is a higher possibility of this individual entering into a state of flow, i.e., favorable attitudes of consumers towards a particular electronic retailer increases the chances of they experience greater immersion when browsing that website.

This finding can be understood in a more coherent way when considering that one of the obstacles to achieving a state of online flow can be the insecurity experienced by consumers while they browse on a website. Following this line of thought, when consumers feel that the company is reliable, they maintain a less defensive posture regarding the website and have the possibility of experiencing more fun and deep engagement during navigation.

Continuing with the discussion of the features of online shopping experiences that precede the online state of flow, with respect to the influence that variety of products available at website has on online flow, we identify that consumers may feel more curious when viewing a website with greater options. Furthermore, the availability of a wide assortment by the electronic retailer can maintain consumers interested and without consciousness of time while evaluating the alternatives that are available, going into a state of flow.

In addition, when exploring the relationship between online flow and website aesthetics, it appears that consumers can feel more encouraged to navigate when they are required to use less effort to understand the structure of the website. In this sense, websites that present a logical organization of its products (e.g., section of male, female, and children's items), making navigation of consumers easier, can awaken greater interest, contributing to consumer immersion during the online purchase process.

Finally, exploring the relationship between online flow and prices, we hypothesized that the perception of consumers that a website has lower prices than those charged by other physical or electronic retailers exerts a "relaxing" effect, reduces anxiety during navigation and increases interest in the website. Thus, the perception that prices are lower can increase the immersion of consumers while browsing, because they feel that don't need to pay attention to monetary matters and, consequently, start to perceive other website elements (such as range of products, information available, website aesthetics, etc.), experiencing a more complete online shopping experience, which leads them to attain online flow.

Having addressed the elements associated with online shopping experiences that exert relevant influence on the online flow state, we next discuss the elements that influence the e-satisfaction in high involvement purchases.

Starting the discussion with the relationship between e-satisfaction and online flow, it appears, based on the proposed structural equation model, the individual immersion, loss of self-consciousness, high curiosity, and natural fun experienced during navigation into a website tend to positively influence e-satisfaction in high involvement purchases. Thus, consumers coming into a flow state have a higher chance of being satisfied with the shopping experience in virtual shopping environments, which highlights the influence that cognitive-emotional states experienced during the navigation has on e-satisfaction.

In a complementary way, with regard to the relationship between convenience and e-satisfaction in high involvement contexts, it is noteworthy that, as advocated by Eid (2011), Lin and Sun (2009), Evanschitzky et al. (2004), Burke (2002) and Szymanski and Hise (2000), ease perceived by consumers in conducting transactions in virtual environments tends to positively influence satisfaction.

Items that increase the ease perceived by consumers buying at e-commerce identified in the literature review include tools of price/product comparison (Schaupp and Belanger, 2005), website maps, FAQ, instructions available, easy registration (Kim et al., 2009), and search tools (Kim et al., 2009; Zhao and Dholakia, 2009; Chung and Shin, 2008; Lee and Lin, 2005). All these elements tend to reduce the effort required for consumers navigate the website and complete their purchases in less time.

Analyzing the dimensions of convenience, Berry et al. (2002) reiterate that e-commerce stands out for convenience of decision and transaction because it reduces the time and effort involved in finding information regarding specific products and services, comparison of competitors, and evaluation of shopping alternatives.

Analyzing the relationship between website's assortment and e-satisfaction, it appears that a large assortment significantly influences the satisfaction experienced by consumers in online purchasing processes of high involvement. This finding is consistent with studies of Dholakia and Zhao (2010), Holloway and Beatty (2008), Evanschitzky et al. (2004), Burke (2002), and Szymanski and Hise (2000), which relate the number of options available to consumers and online satisfaction. Furthermore, as explored by Liu et al. (2008) and Fasanghari and Roudsari (2008), individuals feel more satisfied when they realize they have ample choice.

This finding is also in agreement with the results obtained by Martín and Camarero (2008) and Posselt and Gerstner (2005), who concluded that consumers, after finalizing the online purchase, feel more satisfied when they see different purchase options, many of which are not available in traditional retail, optimizing their process of choice.

In turn, the relationship between e-satisfaction and reliability is supported by the studies of Rose et al. (2011), who found that websites that are associated with greater confidence tend to reduce the perception of risk associated with e-commerce and elicit favorable attitudes from consumers, increasing satisfaction. Therefore, as greater the confidence in the retail brand, lower is the cognitive dissonance or discomfort of the individual after placing the purchase order, which tends to positively influence the online satisfaction (Rose et al., 2011; Fasanghari and Roudsari, 2008; Schaupp and Belanger, 2005).

The credibility of electronic retailers is directly related to consumer perception about the privacy and security of the website, as well as the guarantees offered (Kassim and Abdullah, 2010; Cheung and Lee, 2005; Ribbink et al., 2004; Wang et al., 2003). When there are adequate security and privacy policies, the consumer discomfort tends to be lower after placement of purchase orders, which increases the sense of control and e-satisfaction (Martín and Camarero, 2008).

The relationship between the quality of delivery service and e-satisfaction was also identified, which is in line with the studies of Parasuraman et al. (2005) and Collier and Bienstock (2006), who argue that e-consumers overestimate the quality of delivery service in electronic environments, because they are affected in a very negative way by failures in this regard. According to Dholakia and Zhao (2010), Fasanghari and Roudsari (2008), Holloway and Beatty (2008), Liu et al. (2008), Posselt and Gerstner (2005), Schaupp and Belanger (2005), and Wang and Huarng (2002), the delivery time is essential in the evaluation of consumers. If the time between order placement and delivery of the product does not correspond to the deadline previously informed by the retailer, the consumer tends to become very dissatisfied (Fasanghari and Roudsari, 2008; Schaupp and Belanger, 2005).

Looking at the issue of delivery time from the perspective of cognitive dissonance, we consider that, when the deadline previously informed by the e-retailer is not respected, the consumers’ post-purchase discomfort tends to increase, which can be explained by geographical distance between buyer and seller and the intangibility of the online purchase process. When this occurs, the post-purchase satisfaction levels tend to suffer large losses.

Finally, we cite the importance attributed to the exchange and return policies. The structural equation modeling shows that the perception of fair exchange and return policies tends to positively influence the e-satisfaction, corroborating studies Holloway and Beatty (2008) and Wang and Huarng (2002).

Exchange and return policies of an e-retailer can transmit greater security to an e-consumer that there will be no major obstacles for the return of products that do not meet expectations, and/or easy exchange of the purchased product if a problem occurs. Therefore, as exchange and return policies reduce risks and discomfort associated by individuals to e-commerce, tend to increase e-satisfaction (Martín and Camarero, 2008; Cheung and Lee, 2005).

To conclude, we made a comparison between the elements of online shopping experiences that influence e-satisfaction and those that influence the state of flow. Our results show that only two elements influence both: the variety of products available on the website and website credibility. In this regard, it is noted that, in addition to increasing interest and immersion of consumer when browsing a certain website, the availability of several product also guarantees that the online shopping experience become more satisfying. In addition, it appears the credibility of the electronic retailer reduces psychological resistance associated with the online buying process and the website, which may leaves a consumer to experience the state of flow and also increases his e-satisfaction.

Factors that were not significant for the e-satisfaction or the online flow include the information available on the website, the degree of customization of offers, and the response time of the website, which may be the subject of future studies, considering that, although they do not affect the e-satisfaction, they could influence consumer choice between different websites.

Contextualized by the significant growth that Brazilian e-commerce has shown in the last decade, this study aimed to analyze the relation between the attributes of online shopping experiences, the state of online flow and e-satisfaction in high-involvement purchases. To achieve this goal, we made a literature review in order to identify all theoretical elements of online shopping experience that could influence e-satisfaction.

From the statistical analysis of data on high involvement purchases, we found there are some attributes of online shopping experiences that directly influence e-satisfaction (convenience associated with the website, exchanges and return policies, and quality of delivery service), while others influence online flow state (perceived innovation on the website, prices, and website aesthetics), which, in turn, influence the e-satisfaction. Additionally, we identified two attributes that influence both online flow and e-satisfaction (reliability attributed to the website and variety of products available for sale on the website).

Analyzing the relationships identified, it appears that more involved consumers tend to entry into a deep state of immersion while shopping when faced with more innovative and organized virtual environments, and when they realize that the prices of website are attractive.

In addition, more involved consumers tend to be more satisfied when the delivery of purchased products is realized within the time specified by the electronic retailer at time of purchase, when they perceive that exchanges and returns policies of the company are fair, and when they can explore the website without major difficulties.

Thus, we can accept the three research hypotheses, i.e., some attributes of online shopping experience influence online flow and/or e-satisfaction of Brazilian e-consumers, and the state of online flow influences e-satisfaction of these same consumers. This findings show there is relationship between online shopping experience, online flow and e-satisfaction in high involvement purchases.

As implications of this research, we have that e-companies should prioritize three aspects of its business: structure and design of online store (website), service level to consumers and reputation. These, added to prices, can be pointed as critical success factors to companies seeking to compete in the virtual market adopting a strategy focused on the customer.

As a managerial contribution of this research, we identified attributes of shopping experiences of high involvement that should be monitored by virtual retailers to increase online flow and satisfaction of e-consumers. Thus, this research presents a vision that combines strategic decisions and operations management in e-commerce.

Additionally, as an academic contribution, this research adds to the study of consumer behavior in e-commerce. Moreover, considering the study explored e-satisfaction and online flow simultaneity and incorporated a perspective based on involvement of the e-consumer during purchase process, it can be considered innovative.

One limitation of this study is the fact that the sample was not selected randomly, reducing the generalizability of the results. However, it should be added that, by opting for sampling by quotas, the researcher was able to form a sample that had age distributions that represented the actual distribution of the population studied. Therefore, we consider that, despite not having been formed by a random selection process, the research sample represents, at least peripherally, the target population of the study.

For future research, it is suggested that the questionnaire developed be tested in other samples, enabling the comparison of results. Additionally, it is recommended that the reflections arising from the analysis of the results for high involvement purchases be the subject of future studies, consolidating the discussion started at this work.

It is also suggested that future studies analyze attributes of online shopping experiences that influences e-satisfaction in low involvement purchases. Besides, another issue to be addressed in future studies is the discussion of how demographic and cognitive characteristics of consumers can influence their buying behavior in e-commerce, based on the models proposed here. Finally, it is recommended that the construct “perceived innovation”, which proved to be positively associated with online flow in high involvement purchases, be divided into dimensions and the influence of each dimension to online flow be examined separated.