Consuming Aspect In 3d Online Environments

Published Date: 02 Nov 2017

The age distribution of the sample peaks in the groups of 18-23 years old (52.54%) and 24-39 years old (47.45%). In terms of consuming activity in 3D online stores, the percentage of younger group that do not buy or sell products is higher (25.80%) than in the group of 24-39 years old (17.85%). This cannot be explained by the internet experience, as both groups do not differ significantly. In particular, the mean of younger group for first starting using the internet is 4.63, and the mean of frequency of using the internet is 4.33. The means for the other groups are 4.64 and 4.36 respectively. While the internet experience does not significantly differ, the average income of these two groups is quite different. Only one respondent from the younger group reported earning over 500 Euros, while in the other group a percentage of 39.28% earn more than 500 Euros. Thus, the average income is likely to affect the consuming activity. Also, the reasons for not conducting purchases in this environment may shed light on the previous findings. The most important reasons for the younger group (i.e., 18-23) are lack of trust for the products displayed and the 3D online stores, while for the other age group (i.e., 24-39) the main reason is the familiarity with the medium.

In terms of gender effects on conducting purchases through 3D online retail stores, the percentages of men and women that do not buy products or services in this environment are 12.5% and 33.33% respectively. As far as internet experience is concerned, both sexes do not have significant differences, and are considered experienced internet users. In this regard, the mean for first starting using the internet for men is 4.65 and for women 4.66, and in terms of the frequency of using the internet the means are 4.34 and 4.36 respectively. As a result, women are considered slightly more experienced internet users than men. Also, the average income does not significantly differ in two samples. Half of the men that do not buy products through 3D online stores, do not buy products through the internet. The reasons are lack of trust, security and familiarity with the environment. Security and lack of trust of the internet medium was expected to lead to similar behaviour in the context of 3D online environments. Similarly, nearly half of the women (44.44%) that do not buy products through the internet follow the same behaviour in 3D environments. Security and lack of trust have been illustrated as the most critical factors, while familiarity with the environment seems to be the explanation of women that buy products through the internet and not through 3D online stores.

6.3.2. Variables and Reliability Analysis Measurement

The measurement selection and analysis of variables included in the research model was presented in chapter 4. All the items used in the questionnaire to measure the constructs of the research model should be tested to confirm the reliability of internal consistency. While the split-half reliability is regarded as the easiest for investigating internal consistency, the coefficient alpha or Cronbach’s alpha (Cronbach α) is the most popular approach (Malhotra 2007).

This approach was therefore used to test the reliability of the constructs of this study using SPSS V.17; the results are presented and discussed in the following tables of this section. Given that all the participants of the experiment evaluated the constructs of the research model in terms of five distinct store layout formats. In this regard, the reliability of internal consistency of each variable was measured in each layout type. The following tables present the results of reliability tests of the variables in each of the five store layout types. A detailed analysis of the results is included in Appendix D.

It should be mentioned that in this study, the Online Customer Experience (OCE) variable is comprised of Pleasure, Arousal, Dominance, and Flow. To test the reliability of OCE, all the items of the aforementioned components were tested. The six items used to measure pleasure, the four items used to measure arousal, the six items used to measure dominance, and the one item used to measure flow, were all tested to measure internal consistency reliability of Online Customer Experience in each store layout type. However, it should be mentioned that reliability tests were performed in each of the variables that comprise customer experience, and all values found to be above the acceptable level of 0.6.

Malhotra (2007) suggests that a value of less that 0.6 (i.e., Cronbach’s α < 0.6) is not acceptable for internal consistency variability. The Cronbach’s α results of all constructs in terms of Store#1 layout type are presented in Table 6.9 (see Appendix D.1 for a detailed presentation).

The results of Table 6.9 confirm that most of the variables scored higher than 0.6 in Cronbach’s α test. About half of them scored higher than 0.8, signaling quite satisfactory internal consistency reliability. However, the value of Cronbach α of the construct in interactivity depicts an unacceptable level on internal consistency. The total-statistic table for this construct included in Appendix D.1, shows that by omitting the item INT3, Cronbach’s α value becomes 0.716. Thus, this item was omitted from all further analysis in terms of this construct. The low value of interactivity can be attributed to the design of the laboratory experiment. Participants were provided with a video and a description of each store layout type, but they did not have a real interaction with each store. In this regard, the third item of this construct is likely to have misled their evaluation.

The construct Online Customer Experience scored higher than 0.9, implying a particularly high degree of internal consistency reliability. Malhotra (2007) reports that as the items of as construct tend to increase, so does the value of coefficient. A 17-item scale was adopted to measure the aforementioned construct in this study, explaining the high degree of reliability (0.929).

Results of Cronbach’s α tests of the constructs regarding the layout type of the Store#2 are presented in Table 6.10. All variables’ values are above the acceptable level of 0.6., allowing further analysis. The reason of the high value of OCE’s Cronbach’s alpha that was discussed earlier stands for Store#2 layout type as well. A detailed presentation of these reliability results is included in Appendix D.2.

All the constructs used for the evaluation of Store#3 layout type, scored above 0.6. Results are summarized in Table 6.11, and presented in detail in Appendix D.3.

Similarly, results of Cronbach’s α variables in terms of the Store#4 layout type are presented in Table 6.12. All other constructs apart from interactivity, scored higher than the acceptable level. However, the in-depth presentation of results included in Appendix D.4 indicates that by omitting the third item of interactivity (INT3) the value of Cronbach α is increased to a level of .637 (i.e., acceptable value of α coefficient). In this regard, this item was omitted from further research analyses.

As far as Store’s #5 layout type is concerned, the values of evaluation constructs are presented in Table 6.13. Results indicate that the constructs interactivity and safety do not meet internal consistency reliability requirements. In terms of interactivity, the in-depth presentation of results included in Appendix D.5 indicates that by omitting the third item of interactivity (INT3) the value of Cronbach α is increased to an acceptable level of .779. This item was omitted from further research analyses. The construct safety was measured by a 2-item scale, implying that an unacceptable level of reliability would exclude the construct from further analyses. In this regard, the construct safety in terms of Store#5 layout type was considered unreliable and was not used to investigate the respective research hypotheses.

The constructs hedonic shopping motivation, utilitarian shopping motivation, and telepresence were all considered as moderators in the research model. Results of coefficients α values are presented in the following table. All Cronbach’s α values confirm the internal consistency reliability of these constructs. The in-depth analysis of these constructs is included in Appendix D.6, D.7, and D.8 respectively.

Finally, a realism check was performed to identify the realism of the laboratory experiment. It should be reminded that participants were provided at the survey with a video and a description of each store layout type, but did not take part in any real shopping situation. In this regard, a realism check was adopted to investigate whether the participants fully understood the description of the store that was described and presented, and whether they could imagine an actual 3D online store doing the things described in the situation presented.

Results of Table 6.15 indicate a high level of internal consistency reliability. Taking into account the means of these two realism check items which are 4.4, and 4.6 respectively, a high level of realism of the laboratory experiment can be assumed. An in-depth analysis of realism check Cronbach α is included in Appendix D.9.

6.3.3. Assumption Testing

This section investigates the underlying assumptions regarding the statistical techniques adopted to test the research hypotheses. Hypotheses H#1-H#14 were tested through one-way repeated measures Analysis of Variance, hypotheses H#15-H#16 through Multiple Regression, and hypotheses H#17-H#19 through mixed/split-plot Analysis of Variance. The rationale for selecting these statistical techniques and the underlying assumption testing are outlined in the following three sections.

6.3.3.1. One-way repeated measures ANOVA (H#1-H#14)

Hypotheses H#1-H#14 concern the influence of five distinctive layout types (i.e., Layout#1, Layout#2, Layout#3, Layout#4, Layout#5) on efficiency, merchandise quality perceptions, online shopping enjoyment, online shopping involvement, online store perception, communication style, usefulness, ease of use, entertainment, convenience, safety, navigation, interactivity, and online customer experience. The research model presented in chapter 4 illustrates layout as the independent variable (or treatment), and the fourteen aforementioned variables as dependent. Similarly, it was noted that every participant of the study’s sample performed under every condition (i.e., layout type), following a within-subjects design. In cases where there is need to test the effects of an independent variable on a dependent, in terms of any overall differences among the means following a within-subjects design, the one-way repeated measures ANOVA is the appropriate statistical technique (Coakes et al. 2009; Laerd Statistics 2012; Khelifa 2012; Malhotra 2007).

There are four assumptions measuring appropriateness of data in terms of implementing repeated measures ANOVA (Coakes et al. 2009):

(1) Random selection of the sample

(2) Normality of the population

(3) Homogeneity of variance

(4) Sphericity

The first assumption is not violated, as it was considered at the designing phase of the study. The second assumption investigates whether the population’ scores follow a normal distribution. There are many different approaches to test normality; skewness, kurtosis, stem-and-leaf plots and probability plots were adopted. While there is a widespread notion that violation of assumptions does not affect the results (Schwab 2007), the aforementioned techniques are considered appropriate to test population’s normality (Schwab 2007; Cakes et al. 2009; Pallant 2001). Schwab (2007) reports that a dependent variable is considered normally distributed when scores of skewness and kurtosis fall between -1 and +1. Price (2000) introduces another way of computing whether the assumptions of skewness and kurtosis are not violated. If the value of skewness is within the range of ±(twice the value of the std. error of skewness), then the distribution is considered normal. In this study, both computation methods were adopted along with the results of stem-and-leaf plots and probability plots.

The values of skewness and kurtosis of all variables included in the research model in terms of store layout types (i.e., Store#1, Store#2, Store#3, Store#4, Store#5) indicate that are normally distributed in all cases. An in-depth analysis of normality tests is presented in Appendix E, while the summary of the results is showed in the Appendix E.17.

The third assumption underlying the repeated measures ANOVA is homogeneity of variance. Coakes et al. (2009) claim that in order to test this assumption, the largest and the smallest variances of each group should be divided to obtain the F-max score. If this score is not greater than three, then the assumption has not been violated. Results regarding this assumption can be confirmed in the SPSS outputs provided in Appendix F. This assumption has not been violated in terms of repeated measures ANOVA.

Sphericity, which is the forth assumption, delineates with the variance of the population difference scores (Coakes et al. 2009). The Mauchly’s test of sphericity was used to assess sphericity in all ANOVAs regarding hypotheses H#1-H#14. This test is included in the output of each repeated measures ANOVA presented in Appendix F. The value for Mauchly’s test was found to be significant (p<0.5) in most cases. In this regard, the F-ratio should be calculated using new degrees of freedom (Coakes et al. 2009). The corrective actions were based on the Greenhouse-Geisser and Huynh-Feldt values (Abdi 2010). In each case if the value of epsilon was >0.75 then the Huynh-Feldt correction was used (Girden 1992). If the value of epsilon was <0.75, then the Greenhouse-Geisser correction was used (Field 1998). Finally, the F-ratios evaluated indicated that were still significant when using the new dfs in most cases.

Therefore, all checks confirm the suitability of the survey data for conducting repeated measures ANOVA. The results of hypotheses testing are presented in section 6.3.4.

6.3.3.2. Multiple Regression

Hypotheses H#15-H#16 delineate with the predicting power of fourteen variables on online purchase intentions (H#15) and word of mouth intentions (H#16). In other words, following the S-O-R model, the organism and response columns of the research model presented in chapter four, are tested. The fourteen variables of the organism column serve as the independent variables, and the two variables of response column serve as the dependent variables. Malhotra (2007) suggests the use of Multiple Regression in order to investigate whether the variation of a single dependent variable can be explained by two or more independent variables. In this case two multiple regression models were used to test the variation on online purchase intentions and word of mouth intentions in terms of the fourteen variables. The two hypotheses call for the examination of the whole set of predictor variables on the dependent variables. To that end, the use of standard regression model was adopted (Coakes et al. 2009).

The assumptions underlying the use of multiple regression are (Coakes et al. 2009):

Ratio of cases to independent variables.

The assumption of having twenty times cases than the predictor variables for standard regression has been met. In this case, there are twelve predictor variables and 59 cases.

Outliers

The existence of outliers in a regression model influences the outcome of the model. The use of residual scatterplots was considered appropriate to investigate this assumption. The output of SPSS regarding each of the regression models presented in Appendix F confirms that this assumption is not violated in any of the cases.

Multicollinearity and singularity

These two issues deal with the existence of high (multicollinearity) or perfect (singularity) correlations among the independent variables. The SPSS v.17 that was used has default values for these two issues and does not allow problematic variables.

Normality, linearity, homoscedasticity, and independence of residuals

Residual scatterplots of the SPSS output included in Appendix F shed light on the normal distribution of the obtained and predicted dependent variables’ values, on the linearity of the predicted variables’ values, and on the same variance for all predicted values.

The results of these hypotheses are presented in section 6.3.4.

6.3.3.3. Mixed/split plot design (SPANOVA)

Research hypotheses H#17-H#19 investigate the moderating role of hedonic and utilitarian shopping motivation, and telepresence on the causal relationship among the independent variable (i.e., store layout) and the fourteen dependent variables presented in the research model (organism column). For this test, a mixed design of repeated measures and between-group variables is developed. The repeated measures variables refer to the variables measured in various conditions (e.g., the variable "convenience" was measured in five distinct layout types), while the between-group variables refer to the clustering of the sample in terms of the moderators (e.g., participants with high telepresence vs. participants with low telepresence).

The literature suggests that the Mixed/Split-Plot Analysis of Variance is a powerful statistical tool that is best suited for this kind of mixed designs (DeCasper and Spence 1986; Labar et al. 1995; Maroun and Richter-Levin 2003; Field 2012). Five assumptions underpin the use of split-plot ANOVA (Coakes et al. 2009). The first four are the same with repeated measures ANOVA presented in section 6.3.3.1, while the fifth assumption is homogeneity of intercorrelations (Coakes et al. 2009). It is a sensitive statistic that investigates whether the model of intercorrelations among the repeated measures levels is consistent with between-subjects levels. In SPSS v.17 this assumption can be computed by using the Box’s M statistic. When this statistic is not significant (i.e., p>.001), the assumption of homogeneity is not violated (Coakes et al. 2009). The following table summarizes the results of this assumption in terms of hypotheses H#17-H#19. The results of the other four assumptions have either been presented already or they are included in Appendix F, attached to the output of split-plot ANOVA. In case of violation occurrence, this is mentioned in the respective analysis of results.

The results of Table 6.16 indicate that there are eight cases (highlighted in the third column) which fail to meet the assumptions’ requirements either because there are fewer than two nonsingular cell covariance matrices or because the statistic has been significant (i.e., p<=0.001). The first case means that the SPSS software failed to compute this assumption’s test. It does not mean that failed to do the analysis. In this case, the output of the analysis is amenable to criticism. To overcome this issue, Stack (2008) states that the control group should be reduced to a sample of equal size. This is not an option is this study, because there are cases where the one group comprised of 21 participants and the other comprised of 38 participants. In this regard, it was not considered appropriate to arbitrarily exclude cases from the sample’s population.

6.3.4. Hypotheses Testing Results and Discussion

This section presents the analysis of results and discussion of the hypotheses testing. One-way repeated measures ANOVA was used to test H#1-H#14 in terms of significant differences among the five layout evaluation means. Results of ANOVA show whether there is an overall significant difference among means but do not specify where the difference exists (Malhotra 2007). The Bonferroni post-hoc test was used to identify where those differences occurred (Laerd Statistics 2012). However, in case the overall result of ANOVA is not significant, the Bonferroni post-hoc test should not be considered. In other words, if the null hypothesis is not rejected (i.e., F statistic), there is no reason to discover any differences among means of the five layouts.

6.3.4.1. Research hypothesis#1: Efficiency

This hypothesis investigates whether there is a significant difference among the layout types in terms of efficiency. The detailed SPSS output of this test is included in Appendix F1. The assumption of sphericity was violated in this test (Mauchly’s W=.250, sig=.000). As aforementioned in the assumption testing section, the Greenhouse-Geisser and Huynh-Feldt values of epsilon should be considered (Abdi 2010). In this case, the value of epsilon is under 0.75 (i.e., 0.550) leading to the use of the Greenhouse-Geisser correction (Field 1998). Results of within subjects effects presented in the following table show a significant value of the F statistic (i.e., F=3,138, sig.=0.42). In this regard, there are not significant differences among the layout means, and the null hypothesis is not rejected.

However, the store#1 layout type achieved the highest mean value (i.e, 3.55), while the store#2 layout type the lowest (i.e., 3.15). Although the outcome of this analysis was not expected, the nature of 3D online retail stores in conjunction with the items used to measure efficiency could provide insightful points. The process of accessing a 2D online store could be considered much easier in relation to a 3D online store. Consumers are much more familiar with 2D retailing than 3D. Also, 2D online retail stores are accessible from multiple devices (i.e., mobile phones, tablets etc.), while the access in a 3D online retail store requires more advanced technological requirements (e.g., graphics engine). In this regard, the measurement items regarding the efficiency of time management and making the life easier could be considered the same in various store layout types. It is likely that a visit in a 3D retail store would make life easier than a visit in a traditional store, but the efficiency of visiting 3D online retail stores does not seem to be influenced by store layout. This outcome does not confirm prior claims regarding the influence of store layout on efficiency (Titus and Everett 1995; Puccineli et al. 2009).

6.3.4.2. Research hypothesis#2: Merchandise quality perceptions

The second research hypothesis investigates the influence of store layout on merchandise quality perceptions. The detailed SPSS output included in Appendix F.2 shows that because of Mauchly’s test of Scphericity values (W=.421, sig.=.000), the Greenhouse-Geisser correction should be used (sig.=.000). The Bonferroni post-hoc test presented in the following table was used to determine which layout means differed in terms of merchandise quality perceptions. The merchandise quality perceptions elicited a slight reduction from store#4 layout (mean=4.34) to store#5 layout (mean=4.14) but this reduction is not statistically significant, according to the results of Table 6.19. However, the reduction from store#5 layout to store#1 layout (mean=3.31) is statistically significant. Similarly the reduction from store#1 layout to store#3 layout (mean=3.24) is not statistically significant, while the reduction from store#3 layout to store#2 layout (mean=2.75) is significant.

The results of this hypothesis testing were expected and are line with prior knowledge regarding traditional and online retail stores. The "boutique" store layout (i.e., store#4) emphasizes interesting architecture, attractive materials, appealing allocation and display of products, and gives distinctive names to products in order to help customers to differentiate them. In this regard, it was expected that the quality of the products displayed in a "boutique" store would be considered of high quality (Wang et al. 2011).

In the same vein, it was expected that the products displayed in "large-warehouse" stores (i.e., store#2: lowest mean value) would not be considered of high quality, as this layout type does not emphasize the quality of the products. The non significant difference between the "boutique" store layout and store#5 layout (i.e., department store) can be attributed to the fact that the department layout type is similar to traditional department stores, and consumers are familiar with the quality of products in traditional retail stores. Regarding the non significant difference between store#1 (i.e., "medium-size" stores) layout and store#3 layout (i.e., "image-reliant" stores), their main difference in terms of products’ display is that "medium-size" stores use models to display the products, while "image-reliant" stores use only images. According to the study’s results, the use of models does not seem to affect the perceptions of products’ quality.

6.3.4.3. Research hypothesis#3: Online shopping enjoyment

The third research hypothesis tests the influence of store layout on online shopping enjoyment. The repeated measures ANOVA using Greenhouse-Geisser correction showed that the mean scores for online shopping enjoyment were statistically significantly different (F(2.852, 165.429)=7.720, sig.=.000). The Greenhouse-Geisser correction was used because of the Mauchly’s test of Scphericity values. The Bonferroni post-hoc test was used to discover which means were different. According to Table 6.19, online shopping enjoyment for store#1 layout is not significantly different with the other stores’ layouts (see detailed SPSS output in Appendix F.3). The store#1 layout achieved the third place (i.e., mean=3.34) in terms of mean scores, while the store#4 layout achieved the first (i.e., mean=3.73), the store#5 layout the second (i.e., mean=3.55), the store#3 layout the forth (i.e., mean=3.21), and the store#2 layout the fifth place (i.e., mean=2.97). The store#2 layout significantly differs from store#4 and store#5 layouts; the store#3 layout significantly differs from the store#4 layout; the store#4 layout significantly differs from the store#2 and store#3 layouts; and the store#5 layout significantly differs from the store#2 layout.

The online shopping enjoyment construct measures the appeal, entertaining, fun, interesting, and exciting aspect of the shopping experience provided by the store layout. In this regard, the artistic items that appear in a boutique store layout (i.e., store#4), and the shift of this layout to provide a unique and of high quality experience were expected to lead this layout to scoring the highest. On the contrary, the "large-warehouse" store layout (i.e., store #2) emphasizes displaying a great variety of products and the easiness of finding products without paying particular attention to the enjoyable side of customer experience. Similarly, the "department" store layout (i.e., store#4) includes all the characteristics that appear in the "medium-size" store layout (i.e., store#1) and "image-reliant" store layout (i.e., store#3) that could influence the shopping enjoyment. For example, the use of images, the use of models/avatars to display the products, and the theme-based/similar-based display of products, are characteristics included in both store layout types that are included in store#5 layout type as well. In addition, the store#5 layout type emphasizes the appealing and exciting aspect of various departments within the store. The positive influence of the excitement on the shopping enjoyment is also confirmed by Kim et al.’s (2007) study.

6.3.4.4. Research hypothesis#4: Online shopping involvement

The forth research hypothesis tests the influence of store layout type on online shopping involvement. Similarly to the previous cases, the use of Greenhouse-Geisser correction was considered appropriate (Mauchly’s test of Scphericity: W=.237, sig.=.000). The significance of Greenhouse-Geisser correction (sig.=.004<.005) determined that the means of online shopping involvement differ statistically significantly (i.e., F(14.468, 160.172)=5.239) among the store layout types (see detailed SPSS output included in Appendix F.4). The store#1 layout type elicited the highest mean (i.e., mean=3.53), and in descending order, the store#4 layout (i.e., mean=3.52), the store#5 layout type (i.e., mean=3.32), the store#3 layout type (i.e., mean=3.21), and the store#2 layout type (i.e., mean= 2.93). The Bonferroni post-hoc test indicated that the store#1 layout type differs from store#2 layout type, while the store#2 layout type differs from the store#4 layout type besides the store#2 layout type. Finally, the store#4 layout type differs from the store#2 layout type (see Table 6.20).

The online shopping involvement investigates the role of store layout on providing an important, meaningful, relevant, and concerning experience to consumers. Characteristics of the "medium-size" store layout type (i.e., store#1) such as the insertion of screens in the floor plan to increase the amount of the display space, and the need of avatars to move through the store rather than just being able to pan the walls with the camera are likely to prompt users-consumers to get more involved with the store.

Conversely, characteristics of "large-warehouse" stores, such as the ability to teleport into specific product-related areas, the ability to get into the building through alternative entry points and the existence of a virtual salesman that guides customers to find the products they are looking for, were expected to contribute to a relevant and meaningful experience. This expectation was not confirmed from the research results. The study’s results indicate that as far as the shopping involvement is concerned, consumers place more emphasis on the layout characteristics that prompt them to visit most of the store’s places.

Finally, the difference between the "boutique" store layout type (i.e., store#4) and the "large-warehouse" store layout type (i.e., store#2) was expected, as the artistic items that are displayed are likely to be expensive, and the consumers prefer to try them on before buying them. Along these lines, the process of reading carefully the characteristics of a "special"/unique product, and the process of visiting the dressing room to try on the product that can take place in a "boutique" store are likely to make customers feel more involved with the store. The first argument is in line with the Fiore and Jin’s (2003) and Li et al.’s (2001) studies which state that the visual presentation of products in conjunction with rich visual information, have a positive influence on involvement. Also, the process of buying a unique and an expensive item is likely to be considered as an important and meaningful experience.

6.3.4.5. Research hypothesis#5: Online store perception

Repeated measures ANOVA was used to test the influence of the five distinct layout types on online store perception. Because of the violation of the assumption of sphericity (Mauchly’s W=.250, sig=.000), the Greenhouse-Geisser correction (epsilon= .663< .75) was used. The detailed output included in Appendix F.5 showed that the mean scores for online store perception were statistically significantly different (i.e., F(36.997, 150.587)= 14.250, sig.= .000). In order to discover where the differences occurred, the Bonferroni post-hoc test was used, and the results are presented in Table 6.21. The first place in terms of mean scores elicited the store#4 layout (i.e., mean= 3.89), and the second place the store#5 layout (i.e., mean= 3.79) but there are no statistically significant differences between these stores. The third place obtained the store#1 layout (i.e., mean= 3.49), which differs significantly from the store#2 layout which has the lowest mean (i.e., mean= 2.91). the forth place is taken up by the store#3 layout (i.e., mean= 3.28) which differs significantly from the store#4 layout. Finally, store#2 layout differs significantly from all other layout types, except for store#3 layout (detailed SPSS output: Appendix F.5).

The online store perception construct measured how each layout was evaluated in terms of navigation, browsing, attractiveness, and the interesting character of the store. One of the components of the "boutique" store layout (i.e., store#4) is the interesting architecture, walls of glass, and attractive and appealing materials. In this regard, it was expected that the boutique store layout would obtain a high value in relation to the other store layout types. Also, it would be expected that the "boutique" layout would be considered differently from the "large-warehouse" store layout (i.e., store#2), and the "Image-reliant" store layouts (i.e., store #3). The former is characterized by a large variety of products rather than design, while the latter by the simplicity of images of the products in order to keep the system requirements much lower. For the same reasons, it was expected that the "large-warehouse" store layout would be considered differently from all the other store layouts, but also, was expected that would not be considered differently from "image-reliant" store layout. "Large-warehouse" and "image-reliant" store layouts follow more or less the same rationale as far as attractive and interesting issues/architecture are concerned. Finally, the outcome of this study, in general confirms prior knowledge about the effects of store layout on online store perception (Mummalaneni 2005).

6.3.4.6. Research hypothesis#6: Communication style

The influencing role of the five distinct layout types on communication style of the store is investigated in this hypothesis. The repeated measures ANOVA indicated statistical significant differences. Because of violation of Macuhly’s test of sphericity (Mauchly’s W=.310, sig=.000), the Greenhouse-Geisser correction (epsilon= .640< .75) was used, showing an overall significant difference among the means at the different store layouts (i.e., F(29.447,161.437)=10.580). The Bonferroni post-hoc test presented in the following table presents which specific means of layout types differed. The store#4 layout type elicited the highest mean (i.e., mean=3.79), and in descending order, the store#5 layout (i.e., mean=3.58), the store#1 layout type (i.e., mean=3.34), the store#3 layout type (i.e., mean=3.27), and the store#2 layout type (i.e., mean= 2.85). However, only the store#2 layout type differs from all the other store layout types. The detailed SPSS output is included in Appendix F.6.

In this hypothesis, the communication style of the store layout was measured in terms of its friendliness, calmness and knowledge (i.e., less vs. more knowledge). The kindness of the approach on the store, the social orientation, and the demonstration of expertise and competence which are determinants of communication style (Verhagen and Dolen 2011), are in line with the result that the "boutique" store layout (i.e., store#4) scored highest. However, it was not expected that only the "large-warehouse" store layout (i.e., store#2) would be considered different from all the other layout types, as it includes some dimensions that should be of its advantage. For example, this kind of the store layout provides customers with the ability to compare similar products, the designer can be contacted for further information about the products, and one or more virtual salesmen in the store guide customers to find the products they are looking for. With reference to the last dimension, the fact that participants did not recognize it as an advantage of this kind of store layout, could be attributed to the design of the laboratory experiment. The video that was presenting this layout type to participants also presented a virtual salesman within the store, but participants did not really interact with the store. Participants are therefore likely to have missed evaluating the presence of virtual salesmen to a large extent, or they did not place importance due to the lack of real interaction.

6.3.4.7. Research hypothesis#7: Ease of use

The seventh research hypothesis investigates the influencing role of store layout on ease of use. The detailed SPSS output included in Appendix F.7, showed that because of the violation of the Mauchly’s test of Scphericity (W=.273, sig.=.000), the Greenhouse-Geisser correction used (epsilon= .612< 0.75) indicated an overall significant difference among the means (i.e., F(12.664, 131.696)=5.577). The Bonferroni post-hoc test’s results presented in Table 6.23, uncovers the differences among means. The ranking of store layout types in terms of ease of use is: (a) store#1 (i.e., mean=3.82), (b) store#3 (i.e., mean=3.78), (c) store#4 (i.e., mean=3.45), (d) store#2 (i.e., mean=3.37), and (e) store#5 (i.e., mean=3.33). The mean scores that statistically differ are across store#1 and store#2, store#4, and store#5 respectively, and between store#3 and store#2.

The ease of use construct was measured in terms of ease of use, operation, interaction, understandability, skillfulness. While the mean scores’ difference is low, the "image-reliant" store layout (i.e., store#3) was expected to elicit the first place in terms of ease of use. This store layout type emphasizes simple product management for the end user, and on displaying wall-only items in order to keep the systems requirements lower and avoid lag of the store. To achieve this, it does not support complicated services and extra skills are not needed. On the other hand it lacks interaction, and this could be the reason for its second place in the ranking. Similarly, it was expected that the "department" store layout (i.e., store#5) would not be considered easy enough to use, as it contains stairs, ramps and various stores inside the department that should facilitate the customers’ visit.

As aforementioned, the "medium-size" store layout (i.e., store#1), differs from the "large-warehouse" store layout (i.e., store#2), from the "boutique" store layout (i.e., store#4), and from the "department" store layout. A probable explanation is that this difference could be attributed to the fact that a characteristic of "medium-size" stores is the insertion of screens in the floor plan and highlighted signs. Prior knowledge (e.g., Wei and Ozok 2005) confirms that similar kind of design characteristics positively influence ease of use.

The significant difference between the "large-warehouse" stores and "image-reliant" stores could be attributed to the teleporting stations that exist in the "large-warehouse" stores. A user-consumer who is not familiar with operating teleporting stations could find this functionally difficult to use.

6.3.4.8. Research hypothesis#8: Usefulness

The influencing role of the layout types on usefulness is investigated in this hypothesis. The detailed SPSS output of this test included in Appendix F.8 showed that the assumption of sphericity was violated in this test (Mauchly’s W=.250, sig=.000), and the Greenhouse-Geisser value of epsilon (i.e., 0.598) was considered appropriate for testing the hypothesis. The results of within-subjects effects in light of this correction, presented in the following table, show that the mean scores for usefulness are not statistically significantly different. The value of the F statistic of the overall model is significant (i.e., F(13.092,180.027)=4.218, sig.=0.12). In this regard, as there are not significant differences among the layout means as far as the usefulness is concerned, the null hypothesis is not rejected. However, in light of the mean scores, the store#1 elicited the highest score (i.e., mean= 3.73), and store#3 (i.e., mean= 3.50), store#4 (i.e., mean= 3.27), store#5 (i.e., mean= 3.24), store#2 (i.e., mean= 3.15) follow.

The results of this hypothesis were not expected as there has been considerable research in traditional and online retail environments highlighting the influencing role of store layout on usefulness (Lee et al. 2003; Chen et al. 2002; Vrechopoulos et al. 2004). In this study the construct was measured in terms of usefulness, effectiveness, easiness, improvement on searching and buying products. The characteristics of each store layout type in terms of searching and buying products should influence the evaluation of users. For example, in the "large-warehouse" stores there are teleporting stations helping customers to find groups of products, in the "department" stores someone is able to find a great variety of products, and in the "image-reliant" stores the products management is very simple for the end-user-consumer. The aforementioned reasons indicate that this result merits further investigation.