Personalization: How to address different shopper types during online shopping

(1)

Personalization: How to address different

shopper types during online shopping

by

Jos van der Velde

2351617

University of Groningen

Faculty of Economics and Business

Msc Marketing Management and Intelligence

Supervisor: prof. dr. P.C. Verhoef Second supervisor: dr. J.E.M. van Nierop

(2)

2

1. Introduction

Nowadays, most firms’ top priority is creating a strong customer experience (Lemon and Verhoef, 2016) as it is an important driver in sustaining a firm’s competitive advantage because customers desire something to remember rather than just a product (Stein and Ramaseshan, 2016). Customer experience develops during the purchase journey as a collection of all points of individual contact between the firm and the customer, also called touchpoints (Lemon and Verhoef, 2016; Stein and Ramaseshan, 2016). According to Meyer and Schwager (2007), these touchpoints do not only occur directly between firm and customer but can also occur indirectly through for example reviews, news reports and word-of-mouth. In this thesis, customer experience is defined as ‘’a

multi-dimensional construct focusing on a customer’s cognitive, emotional, behavioral, sensorial, and social responses to a firm’s offerings during the customer’s entire purchase journey’’ (Lemon and

Verhoef, 2016; pp. 7). As online purchasing increases (Rose, Hair, and Clark, 2011), the web shop as a touchpoint becomes more important because the touchpoint frequency increases. For example, ‘’in 2011, a whopping 70 percent of all internet users made at least one online purchase’’ (Oracle, 2012; pp. 2). According to Baxendale, MacDonald, and Wilson (2015), both touchpoint frequency and positivity can influence brand consideration depending on the type of touchpoint. Thus, as the touchpoint frequency of the web shop is increasing, it is important to increase touchpoint positivity. However, customers nowadays are more empowered due to their significant role in the value creation process (Melero, Sese, and Verhoef, 2016), as a result they have become more demanding and expect the purchase journey to be relevant and personalized (Oracle, 2012). Minsker (2015) also states that all touchpoints that occur during the purchase journey should be personalized in order to make customers want to take that journey again and again. Therefore, it is key to personalize the customers’ web shop experience to increase touchpoint positivity.

(3)

3 Eventually, web shops can make product recommendations for the customer based on customers’ preferences. These product recommendations can be either product-related or brand-related (Jeong and Lee, 2013).

Web personalization takes into account all elements that a web shop can control; however, some elements are outside of the web shop’s control. (Verhoef, Lemon, Parasuraman, Roggeveen, Tsiros, and Schlesinger, 2009). A consumer’s shopper type is one of these elements that is outside a web shop’s control. The development of shopper types was meant to help retailers improve their decision-making regarding differentiation and targeting of their products (Westbrook and Black, 1985). Firms still use segmentation strategies to target customers but firms still need to assist customers in finding the product they want/need by guiding them the way. Web shops do this through web personalization; however, customers do have different shopping motivations (Ganesh, Reynolds, Luckett, and Pomirleanu, 2010; Rohm and Swaminathan, 2004) and have a different preference for e-store attributes (Ganesh et al., 2010). Some customers are just looking around, while others are looking to buy a specific product as cheap as possible. Due to the difference in shopping motivations, each unique customer should be shown a different way in order to find the right product. Thus, web personalization should be adjusted based on customers’ shopping motivation and their deemed importance of e-store attributes. However, current literature just identifies online shopper types (e.g. Ganesh et al., 2010; Rohm and Swaminathan, 2004) but does not address how to act on these different shopper types e.g. what type of product recommendation should be used. Therefore, this thesis will study whether customers’ shopper type may be a plausible reason as they have different shopping motivations.

The purpose of this thesis is twofold and therefore consists of two studies. Study 1 will identify online shopper types based on research by Ganesh et al. (2010) regarding what online shopper types exist as minor research on online shopper types has been done. Study 2 is an explorative study that builds upon study 1. The goal of study 2 is to determine whether the online shopper types found in study 1 have different preferences regarding product recommendation types. Thus, the research question of study 2 is:

(4)

4 The remainder of this paper will be organized in sections. In the next section, a theoretical background on shopper types and web personalization will be provided. This section will be followed by the methodology which will explain the research design and the methods used. Next, the results will be showed, and finally a conclusion will be drawn, ideas for further research will be given, and limitations will be discussed.

2. Theoretical background

Web personalization

Nowadays, it is all about creating a strong customer experience (Lemon and Verhoef, 2016). In order to do so, web shops use web personalization. Web personalization is defined as ‘’the process

of profiling consumers and generating content that matches their preferences and tastes’’ (Ho,

Bodoff, and Tam, 2011; pp. 1). The content that is generated based on consumers’ preferences and tastes is shown these consumers by using recommendation systems. Research shows that the use of recommendation systems has a positive effect on online consumer purchase behavior. Hostler, Yoon, Guo, Guimaraes, and Forgionne (2011) show that recommendation systems especially have a positive impact on product promotion which in turn leads to higher customer satisfaction. Moreover, recommendation systems increase product search effectiveness which in turn leads to an increase in unplanned purchases.

(5)

5 (2011), quality and timeliness – the step in the shopping process - cause a trade-off as it is not possible to realize the best of both worlds. The recommendations based on what is already known about the customer can be shown early on in the shopping process, but they are of lower quality than the recommendations based on learning because customers’ preferences change over time. However, learning customers’ preferences and tastes takes time so these recommendations can only be shown later on in the shopping process.

Besides making a decision about the quality and timeliness of the recommendations and what recommendations system to use, web shops also have to decide what type of product recommendation to use. According to Jeong and Lee (2013), there are two broad types of product recommendations, namely alternative brands product recommendation and additional products product recommendation. To make the difference between both types clear, they give the following example: ‘’when consumers search for a product (e.g. an Apple laptop) some websites provide

them with alternative brands (e.g. Dell and HP laptops) so that they can compare the alternatives with the product they originally sought. Other websites provide consumers with additional products (e.g. cordless mouse and USB) that they can buy together with their original target product’’ (Jeong and Lee, 2013; pp. 539-540). According to Jeong and Lee (2013), alternative

brands product recommendations have the highest level of preference fit followed by a combination of alternative brands and additional products recommendations. Additional products product recommendations have the lowest preference fit. However, the study by Jeong and Lee (2013) does not explain why some customers deem one type of product recommendation to be more congruent with their shopping goal than the other types of product recommendation.

Shopper types

(6)

6 Within the literature on motivation to go shopping, two broad types of shopping values are known, namely utilitarian value and hedonic value. The utilitarian value of shopping means that consumers look out for and concentrate on the most generic goal of shopping, i.e. to get the right product for the right price and at a minimum effort or cost. Hedonic values, on the other hand, represent entertainment and emotional worth (Babin, Darden, and Griffin 1994). These two shopping values are broken down into more specific dimensions in order to create shopper types. These dimensions will be discussed later.

In current literature, there are basically two approaches to establish shopper types. The first approach is based on store attributes. The goal of this approach is to detect which aspects of a store have higher or lower importance to different customers (Angell, Megicks, Memery, and Heffernan, 2014). Based on customers’ preferences of store aspects, shopper types are formed. The goal of the second approach is to detect customers’ motivations to go shopping (Angell et al., 2014). Based on these motivations, shopper types are formed.

Studies that aim to detect shopper types, whether they are based on store attributes or shopping motivations, are built upon research by Stone (1954) because this study was the first to segment consumers in order to identify different urban shopper types. This study identified four different shopper types based on their reasoning to shop at certain types of stores. Later on, Tauber (1972) conducted an exploratory study regarding customers’ motivations to shop. This study hypothesizes that these motivations are based on personal and social motives. ‘’Personal

satisfactions from shopping were obtained from: (1) the opportunity to enact a culturally prescribed role; (2) diversion from daily routine; (3) provision of self-gratification; (4) learning about new trends, fashions, and innovations; (5) obtaining physical exercise; and (6) receiving sensory stimulation from the retail environment. The principal satisfactions of a social nature were: (1) social interaction outside the home; (2) communication with others having similar interests; (3) affiliating with reference groups; (4) obtaining increases in social status; and (5) achieving success in bargaining and negotiation’’ (Westbrook and Black, 1954; pp. 85).

(7)

7 power and authority, and stimulation. However, in the online environment negotiation (bargaining with salespersons), and power and authority (being wanted by a salesperson) are irrelevant as this does not happen online. Furthermore, affiliation (interacting with others) is only relevant to a certain level because you can only interact in, for example, the review section of a product. Therefore, research regarding shopper types in the online environment was needed.

Rohm and Swaminathan (2004) were one of the first to conduct a study regarding online shopper types. They identified four different types, namely convenience shoppers, variety seekers, balanced buyers, and store-oriented shoppers. These shopper types were only based on customers’ motivations. Therefore, Ganesh et al. (2010) basically extended this study by including customers’ motivations as well as e-store attributes. The study by Ganesh et al. (2010) identified eight different online shopper types based on online shopping motivations and e-store attributes. Five of these were found to be similar to the traditional offline shopper.

The online shopping motivations are based on the following dimensions: web shopping convenience, online bidding/haggling, role enactment (looking for bargains), avant-gardism (keeping up with trends), affiliation, stimulation (interacting with interesting websites), and personalized services. Furthermore, the e-store attributes are based on the following dimensions: e-store essentials, offline presence, price orientation, website attractiveness, merchandise variety, and web security/certification.

The first type is the basic shoppers, these consumers are task-oriented and are motivated by the convenience of online shopping but do not care about merchandise variety. Secondly, the

destination shoppers were identified. This type of shopper is especially motivated to shop online

to keep up with trends and by merchandise variety. But they do not care about interacting with others or offline presence of the store at all. Thirdly, there are the bargain seekers. This type is price-oriented and proactive in searching for products. They do not like waiting to be informed about alternatives. Furthermore, they do not care about website attractiveness or interesting websites, they just want to find bargains. Fourthly, there are the interactive shoppers. These consumers enjoy looking for deals and like to be notified of new products or special deals. Moreover, they enjoy online bidding and haggling. However, they do not care about interacting with other consumers or keeping up with trends. The fifth type of shopper is the shopping

enthusiasts; these consumers like to keep up with trends, like to interact with other consumers, but

(8)

8 merchandise variety. They place value on all dimension as they like all aspects of shopping. Sixth, the e-window shopper. This type of shopper likes to surf the internet and interesting websites. Also, they enjoy just looking for what is out there and if possible buy a bargain. Seventhly, there are the

apathetic shoppers; these consumers are not specifically motivated by any motivational dimension

and do not expect web shops to have specific attributes. They do not seem to like shopping and therefore are probably task-oriented. Lastly, there are the risk averse shoppers. These consumers especially care about web security/certification and offline presence.

Goal of the thesis

Study 1 will focus on identifying online shopper types based on the study by Ganesh et al. (2010). This study is used as a basis because research shows that there are two approaches to establish shopper types, shopper types are either based on store attributes or shopping motivations (Angell et al., 2014). The study by Ganesh et al. (2010) was the first to determine online shopper types based on both approaches. However, Ganesh et al. (2010) did not identify online shopper types based on a combination of both approaches. Instead, they identified shopper types based on both approaches separately just like all other studies. However, they then compared the results of the both approaches and found out that the results of both approaches were be similar. The study showed that 5 out of the 8 different shopper types found, were found by both approaches and were similar to traditional offline shopper types. Study 1 will add to current literature by identifying online shopper types based on both store attributes and shopping motivations. As the study by Ganesh et al. (2010) shows that both approaches have similar results, this should not cause any problems.

(9)

9 types might help in achieving different shopping goals. Thus, the goal of study 2 is to determine whether online shopper types have different product recommendation preferences.

Figure 1 – Visual representation of what types of product recommendations different shopper types might prefer

3. Methodology

Data collection and sample

To identify the different online shopper types and to determine whether online shopper types have different preferences regarding product recommendation types, an online survey was conducted. The survey consisted of questions to determine a person’s shopper type and what type of product recommendations utilitarian and hedonic shoppers prefer. An online survey was chosen to gather as many responses as possible to be able to base conclusions on a broad range of responses. Moreover, a survey was the most efficient way to gather responses due to time constraints. Furthermore, an online survey is anonymous so respondents will feel anonymous and thus they are less likely to fill in socially correct answers.

The respondents were invited to participate in the survey through the platform mturk.com, a website by amazon.com that helps you find respondents against a small fee. The advantage of

Online shopper type

(10)

10 using mturk.com is that the sample will be more representative because the respondents will not just be fellow students or family members/friends.

Study 1

Measurement of online shopper types

In the first part of the survey, the respondents were asked to answer statements regarding their shopping motivations and their attitude towards e-store attributes. These statements were also used in the study by Ganesh et al. (2010) and were measured on a seven-point Likert scale. It was decided to use the same statements that Ganesh et al. (2010) because this study is used as a basis but also because these statements cover the different aspects as to why customers enjoy online shopping.

The survey contained 33 items distributed over 7 dimensions regarding customers’ shopping motivations. These responses were ranging from “provides me no satisfaction at all” (1) to “provides me a great deal of satisfaction” (7). Furthermore, 22 items distributed over 5 dimensions were used to measure customers’ attitude towards e-store attributes. These responses were ranging from “not at all important” (1) to “extremely important” (7). The statement ‘’ Website

is certified by the Better Business Bureau’’ that was used by Ganesh et al. (2010) was left out

because this organization only exists in the United States of America and thus is irrelevant for this study. As a result, the statement ‘’Website is certified by an online Watchdog organization’’ was added to the dimension e-Store essentials and the dimension Web security/certification was removed. The complete list of dimensions with their corresponding items can be found in Appendix B.

(11)

11 Table 1. The analysis showed that all items loaded on the intended construct. Subsequently, a reliability analysis was performed to check the reliability of the constructs found. Table 2 shows that the Cronbach’s Alpha’s for all constructs was >.6; thus, the found constructs are reliable. It was decided to keep the item ‘’Being the winning bidder in an online auction’’, even though removing it would improve the Cronbach’s Alpha regarding the construct online bidding/haggling, because a Cronbach’s Alpha of .926 is already excellent.

Construct Eigenvalue % of variance Cumulative %

1 10,345 31,35 31,35 2 6,672 20,22 51,57 3 2,669 8,09 59,66 4 1,837 5,57 65,22 5 1,288 3,90 69,13 6 1,209 3,66 72,79 7 1,010 3,06 75,85

Table 1 – Results of the factor analysis regarding shopping motivations

Constructs Cronbach’s Alpha

Web shopping convenience .899

Online bidding/haggling .926

(could be .945 if the item ‘’Being the winning bidder in an online auction’’ was removed.)

Role enactment .850

Avant-gardism .955

Affiliation .943

Stimulation .925

Personalized services .910

Table 2 – Results of the reliability analysis regarding shopping motivations

(12)

12 individual constructs was also > 5%. The specifics can be found in Table 3. The analysis showed that all items loaded on the intended construct. Subsequently, a reliability analysis was performed to check the reliability of the constructs found. Table 4 shows that the Cronbach’s Alpha’s for all constructs was >.6; thus, the found constructs are reliable. Even though, the Cronbach’s Alpha of the construct e-store essentials could be improved by removing item ‘’Website is certified by an online Watchdog organization’’, it was decided to keep the item because the Cronbach’s Alpha is already excellent and it is the only item taking into account website security monitored by a third party.

Construct Eigenvalue % of variance Cumulative %

1 7,971 36,23 36,23

2 3,410 15,50 51,73

3 2,057 9,35 61,08

4 1,628 7,40 68,48

5 1,243 5,65 74,13

Table 3 – Results of the factor analysis regarding e-store attributes

e-Store essentials .917

(could be .923 if the item ‘’Website is certified by an online Watchdog organization’’ was removed.)

Offline presence .908

Price orientation .859

Website Attractiveness .861

Merchandise variety .886

Table 4 – Results of the reliability analysis regarding e-store attributes Analysis of online shopper types

(13)

13 variance between each other. Then the hierarchical K-means clustering procedure is used to, given the optimal number of clusters, create clusters with minimal variance within clusters and maximal variance between clusters. Furthermore, this clustering procedure distributes the respondents more evenly across the different clusters.

In this thesis, however, there will also be performed a latent class cluster analysis to form shopper types because latent class analysis (LCA) is ‘’an empirically based statistical approach

for explaining the heterogeneity in response-profiles in terms of underlying latent classes’’ (Baum,

Schwens, and Kabst, 2015; pp. 761). Furthermore, latent class analysis has five advantages over other clustering methods of which three are relevant for this study: ‘’(1) LCA is based on a testable

statistical model. (2) The factor structure of the latent class allows accounting for measurement errors of the indicators. Thus, the obtained results are more likely to be valid. (3) LCA provides fit indices that allow comparing different class solutions with each other (e.g. by means of bootstrapped likelihood ration tests). (Baum, Schwens, and Kabst, 2015; pp. 761)

Ganesh et al. (2010) performed two separate cluster analyses. In one cluster analysis they formed clusters based on the shopping motivation dimensions, in the other cluster analysis they formed clusters based on e-store attribute importance. Ganesh et al. (2010) never performed a cluster analysis including both the dimensions regarding shopping motivation and e-store attribute importance at the same time. This thesis, however, will only perform cluster analyses in which both shopping motivation and e-store attribute importance dimensions are included. This is done to be able to form clusters based on both aspects and create more specific segments.

After the cluster analyses, a one-way ANOVA will be performed to determine whether or not the clusters all significantly differ on each dimension because the ANOVA table that results from the hierarchical K-means cluster analysis is limited as it can only show whether or not there is a significant difference between at least two of the clusters on a dimension, but it cannot show which specific clusters significantly differ from each other. Therefore, a post hoc test is needed that results from a one-way ANOVA.

Study 2

Measurement of utilitarian and hedonic shoppers’ product recommendation preference

(14)

14 store anyways. Subsequently, the respondents were shown alternative brands product recommendations, followed by additional products product recommendations, and finally a combination of both. Respondents were asked to answer 3 items per type of product recommendation about how congruent the product recommendations are with their shopping goal (which depends on the scenario they were given). Customers’ preference for a certain type of product recommendation will be measured in terms of congruence; in other words, the fit between a customer’s shopping goal and the type of product recommendation. The idea behind this is that the higher the fit between a customer’s shopping goal and the type of product recommendation, the more useful the product recommendation is in achieving the shopping goal. Therefore, customers will prefer the more congruent product recommendation.

(15)

15

Construct KMO Eigenvalue % of variance Cronbach’s Alpha Congruence (alternative brands) .709 2,331 77,04 .856 Congruence (additional products) .769 2,652 88,40 .934 Congruence (combination of both) .742 2,456 81,86 .887

Table 5 – Results of the factor analysis and reliability analysis regarding the three types of product recommendation Analysis of utilitarian and hedonic shoppers’ product recommendation preference

To test the hypotheses, a Multivariate Analysis of Variance (MANOVA) was performed. The independent variable in this analysis is a consumer’s online shopper type (utilitarian or hedonic). Whether a consumer is utilitarian or hedonic depends on which scenario was shown to the consumer. The dependent variable is a consumer’s perceived congruence of a certain type of product recommendation with the shopping goal. The MANOVA is used to determine whether different shopper types perceive a certain type of product recommendation significantly different.

4. Data analysis and results

In total, 145 respondents filled out the survey and everybody filled out every questions as there were no missings. Most respondents (78%) were between 26-35 years old and the majority of these respondents is male (60%). When asked how many times they visit a web shop per month, the majority (46%) answered that they visit web shops 6-10 times on average. Furthermore, 57% of the respondents stated that they buy on average 2-5 products online per month. Lastly, the average monthly gross income lies between €500 to €2000 and most respondents (21%) have an income between €1000 to €1500.

Study 1

(16)

16 were 145 respondents, the agglomeration schedule shows that there are 144 stages needed in order to fit all respondents into 1 cluster. As Table 6 and Figure 2 show, the scree plot based on the difference between coefficients between each of the last eleven stages (134-144) shows that the optimal number of clusters is 5.

Figure 2 – Scree plot resulting from the multi-step cluster analysis

After the multi-step cluster analysis, the K-means cluster analysis was performed to create 5 clusters that are more evenly distributed. The ANOVA table (Table 7a and 7b) that resulted from the cluster analysis shows that the clusters are significantly different from each other on all dimensions. However, to determine which specific clusters significantly differ from each other on a dimension, a one-way ANOVA was performed. The post hoc Tukey test (see Appendix C) resulting from the one-way ANOVA shows that not all clusters significantly differ from each on other on all dimensions. However, this is normal as clusters will never differ from each other on all dimensions. Based on the post hoc Tukey test, it can be concluded that the clusters differ enough from each other to determine that the clusters are unique. Table 7a and 7b show the dimensions of shopping motivations and e-store attribute importance and the mean score of each cluster on each dimension according to the hierarchical K-means clustering procedure.

0,000 200,000 400,000 600,000 800,000 1000,000 1200,000 134 135 136 137 138 139 140 141 142 143

Stage Coefficient Difference between coefficients 134 1207,60 54,06 135 1261,66 55,07 136 1316,74 61,88 137 1378,62 77,07 138 1455,69 84,98 139 1540,67 94,05 140 1634,72 153,82 141 1788,54 195,29 142 1983,83 242,26 143 226,10 980,06 144 3206,15

(17)

Proportion in % Web shopping convenience Online bidding Role enactment Avant-gardism

Affiliation Stimulation Personalized services Class 1 (17,24) 5,71 2,98 5,63 3,71 2,87 4,85 3,19 Class 2 (11,72) 5,74 2,52 4,61 1,51 1,35 2,99 2,08 Class 3 (18,62) 4,82 4,01 4,25 3,77 3,99 4,25 4,19 Class 4 (13,10) 6,32 2,73 5,84 2,18 2,37 5,78 5,51 Class 5 (39,32) 5,79 5,15 5,85 5,49 5,16 5,79 5,70 F-value 11,28 35,53 17,75 73,77 79,90 47,21 74,48 p-value .000 .000 .000 .000 .000 .000 .000

Table 7a – Average scores of the clusters on all shopping motivation dimensions resulting from the K-means cluster analysis, plus F and p-values

Proportion in % e-Store essentials Offline presence Price orientation Website attractiveness Merchandise variety Class 1 (17,24) 6,35 4,20 4,41 5,44 5,61 Class 2 (11,72) 5,68 2,69 3,53 3,84 4,51 Class 3 (18,62) 4,66 3,79 4,54 4,12 4,46 Class 4 (13,10) 6,13 2,63 5,77 4,12 4,93 Class 5 (39,32) 6,15 4,98 5,76 5,66 5,97 F-value 26,95 16,40 26,33 20,64 14,72 p-value .000 .000 .000 .000 .000

(18)

As an extension to the study by Ganesh et al. (2010), this thesis also performs latent class cluster analysis to identify the different online shopper types. To compare the different models that result from the latent class cluster analysis, the information criteria are used. The closer an information criterion is to zero, the better the model. Usually the AIC is favored over the BIC because this thesis makes use of a relatively small sample size and the AIC is better suited in small sample cases. However, Table 8 shows that the AIC is closest to zero in an 8 class solution and that is surprising as the multi-step cluster analysis showed that 5 clusters would be optimal. As the difference between the multi-step cluster analysis solution and the solution according to the BIC is this huge, the other information criteria were also taken into account. However, taking into account the other information criteria did not help in getting a conclusive solution either because each information criterion shows a different optimal solution. However, because both the BIC and the multi-step cluster analysis show that a 5 class solution is optimal, it was decided to further analyze the 5 class solution.

Number of latent classes BIC AIC AIC3 CAIC 4 class solution 5416,89 5122,19 5221,19 5515,89 5 class solution 5397,36 5028,25 5152,25 5521,36 6 class solution 5453,13 5009,60 5158,60 5602,13 7 class solution 5463,84 4945,89 5119,89 5637,84 8 class solution 5520,62 4928,25 5127,25 5719,62 9 class solution 5618,49 4951,71 5175,71 5842,49

Table 8 – Overview of the different information criteria used to determine the optimal number of clusters

(19)

Proportion in % Web shopping convenience Online bidding Role enactment Avant-gardism

Affiliation Stimulation Personalized services Class 1 (39,01) 5,90 3,24 5,63 3,36 3,17 5,22 4,20 Class 2 (21,98) 5,52 5,29 5,69 5,59 5,26 5,57 5,66 Class 3 (14,02) 5,68 2,68 4,59 1,75 1,54 3,15 2,39 Class 4 (13,12) 4,41 4,01 4,04 4,03 4,31 4,21 4,29 Class 5 (11,85) 6,47 5,13 6,26 5,46 4,74 6,37 6,34 Wald 102,22 196,74 116,75 398,08 398,08 228,00 222,04 p-value .000 .000 .000 .000 .000 .000 .000

Table 9a – Average scores of the clusters on all shopping motivation dimensions resulting from the latent class cluster analysis, plus Wald and p-values

Proportion in % e-Store essentials Offline presence Price orientation Website attractiveness Merchandise variety Class 1 (39,01) 6,13 3,77 5,01 4,74 5,29 Class 2 (21,98) 5,95 5,19 5,66 5,55 5,70 Class 3 (14,02) 5,45 2,68 3,72 3,78 4,44 Class 4 (13,12) 4,54 4,07 4,42 4,23 4,49 Class 5 (11,85) 6,65 4,42 6,24 6,47 6,68 Wald 130,44 80,01 77,52 166,54 194,17 p-value .000 .000 .000 .000 .000

(20)

A comparison of the results of both the hierarchical K-means cluster analysis (Table 7a and 7b) and the latent class cluster analysis (Table 9a and 9b) shows that both methods have led to a similar distribution of the respondents across the different clusters. In both instances, the largest cluster takes account for of around 39% of the respondents and the smallest cluster takes account for around 11,5% of the respondents. The sizes of the other clusters are also similar. However, the similar sized clusters have completely different mean scores. Take the largest cluster of both methods for example, the results show that these consumers have completely different views regarding online bidding/haggling, avant-gardism, and affiliation. Thus, the similar sized clusters of both methods cannot be seen as the same; therefore, it was decided that the clusters that will be elaborated in this thesis will be the clusters that resulted from the K-means cluster analysis. These clusters are preferred over the clusters as a result of the latent class cluster analysis because the four information criteria in Table 8 regarding the latent class cluster analysis all show a different optimal solution. Thus, even though a latent class cluster analysis is supposed to be more valid than other clustering methods (Baum, Schwens, and Kabst, 2015), the latent class cluster analysis does not seem to be reliable in this case.

Elaboration of the classes

The elaboration of these classes is based on the results of the K-means cluster analysis that can be found in Table 8a and 8b.

Class 1 – Store attribute shoppers (17,24%)

(21)

21

Class 2 – Task-oriented shoppers (11,72%)

These shoppers are labelled as task-oriented because of their relatively high scores on web shopping convenience (5,74) and e-store essentials (5,68) compared to their scores on all other dimensions. This shows that it is important to them that they do not need to spend any more time shopping than necessary and probably know what kind of product they want to buy when they go shopping. They do not care about avant-gardism or affiliation with others shown by the 1,51 and 1,35 scores on these dimensions, respectively.

Class 3 – Basic shoppers (18,62%)

These shoppers are labelled as basic because they do not particularly like any of the shopping aspects, but they also do not dislike any of the shopping aspects. They score above average on all dimensions. They do shop online, but do not necessarily do so because of the convenience as they have the lowest score on web shopping convenience (4,82) of all classes.

Class 4 – Bargain hunters (13,10%)

These shoppers really value e-store essentials (6,35) and web shopping convenience (6,32). This is probably because these aspects of web shops speed up the shopping process which is highly valued because these shoppers do not seem to like shopping considering their low scores on avant-gardism (2,18) and affiliation (2,37). Most importantly, they highly value role enactment (5,84), personalized services (5,51), and price orientation (5,77) which shows that these shoppers are in search of bargains.

Class 5 – Online shopping enthusiasts (39,32%)

(22)

22

Study 2

A MANOVA was performed to determine whether different shopper types perceive a certain type of product recommendation significantly different. The multivariate tests (Table 10) show that the clusters have significantly different preferences regarding the types of product recommendation. However, these multivariate tests can only show whether or not there is a significant difference between at least two types of shoppers regarding a type of product recommendation, but it cannot show which specific clusters significantly differ from each other. Furthermore, Table 11 also shows that the clusters significantly differ regarding their product recommendation type preference.

Value F Sig.

Pillai’s Trace .320 4.173 .000

Wilks’ Lambda .704 4.321 .000

Hotelling’s Trace .387 4.411 .000

Roy’s Largest Root .265 9.286 .000

Table 10 – Results of the multivariate tests resulting from the MANOVA

Alternative brands recommendation type Additional products recommendation type Combination of both types Store attribute shoppers 3,79 2,52 3,17 Task-oriented shoppers 3,47 1,98 2,71 Basic shoppers 3,35 2,91 2,99 Bargain hunters 3,98 2,25 2,93 Online shopping enthusiasts 4,02 3,29 3,54 F-value 2,480 0,963 3,368 p-value 0,001 0,000 0,003

(23)

23 The post hoc Tukey test that resulted from the MANOVA shows that there are only a couple of significant differences between the clusters regarding their preference for a type of product recommendation. Table 12 shows between which clusters a significant difference was found and shows that there are only 8 significant differences. It is striking that cluster 2 (Task-oriented shoppers) and cluster 5 (Online shopping enthusiasts) significantly differ on all three product recommendation types. This could be explained by the fact that these are two completely different shopper types with different shopping goals. Table 12 shows that cluster 2 (Task-oriented shoppers) only value alternative brands product recommendations and that cluster 5 (Online shopping enthusiasts) value all types of product recommendation. Furthermore, it is striking that cluster 5 (online shopping enthusiasts) is involved in 6 out of 8 differences found. This is most likely due to the high average scores of this cluster on all three product recommendation types as shown in Table 11. Thus, cluster 5 stands out from the other clusters, and there are only minor differences between the other clusters.

Significant difference between clusters p-value Alternative brands recommendation type 2 & 5 3 & 4 3 & 5 .050 .028 .001 Additional brands recommendation type 1 & 5 2 & 3 2 & 5 4 & 5 .020 .036 .000 .002 Combination of both types 2 & 5 .009

Table 12 – Overview of the clusters that significantly differ from each other

(24)

24

5. Discussion, managerial implications, and future research

Ganesh et al. (2010) found in their study eight different online shopper types based on two separate cluster analyses, one based on shopping motivations and one based on e-store attribute importance. However, by performing two different types of cluster analyses in which both the shopping motivations and e-store attribute importance were included at the same time, this thesis found five different types of online shopper types. This in itself does not cause major implications; however, in contrast to Ganesh et al. (2010), this thesis found that all online shopper types really value the convenience that web shops offer. This might be due to the fact that 83,4% of the respondents indicated that they prefer online shopping over offline shopping. And thus the convenience a web shop offers might be the cause of this preference for online shopping in the first place. This may also be the cause of the striking fact that online shopping enthusiasts form the largest cluster with 39,32%. The size of this cluster is quite surprising even though Ganesh et al. (2010) also found a large cluster of shopping enthusiasts (31,04%). However, this was only regarding the e-store attributes. Regarding the online shopping motivations, Ganesh et al. (2010) only found a shopping enthusiasts cluster with a size of 16,12%. Furthermore, all online shopper types deem e-store essentials important; thus, web shops should at least meet the basic essentials in order to attract customers. Moreover, the results show that most online shopper types enjoy looking for deals and bargains. This might be due to the fact that the internet makes it very easy to compare the price of a product across different web shops. However, it is interesting that this thesis did not find an online shopper type that is comparable to the ‘’Apathetic shopper’’ that Ganesh et al. (2010) identified. According to Ganesh et al. (2010), this is the type of shopper that shops out of necessity because their results show that has low values on all dimensions and thus does not seem to like shopping at all.

(25)

25 should preferably make use of the alternative brands product recommendation to satisfy its customers. This finding might speed up the collaborative filtering recommendation system that is used to identify relationships between new and existing customers in order to find similarities in their preferences in order to make product recommendations (Wang, Luse, Townsend, and Mennecke, 2014) because this recommendation system can narrow down its search to alternative brands or models that a customer is searching for.

Furthermore, this is interesting finding because web shops do still make use of the other types of product recommendations too. Coolblue, for example, only makes use of additional products product recommendations when searching for a laptop. Furthermore, bol.com makes use of both alternative brands product recommendations and a combination of both alternative brands and additional products product recommendations when searching for a laptop. Both Coolblue and bol.com are two very successful web shops in the Netherlands, yet they do still use the less preferred types of product recommendations. Therefore, an idea for future research is to study why web shops do make use of the less preferred types of product recommendations. Does the preference for a type of product recommendation not just depend on a consumers’ online shopper type, but could it also depend on the type of product a consumer is looking for? These questions still need to be answered and are food for thought.

(26)

26 shopper types of their consumers as this can really help satisfying their customers. If a web shop uses additional brands product recommendation even though their consumers are mainly task-oriented shoppers. These consumers will have a harder time achieving their shopping goal, this might cause dissatisfaction, but may also cause the consumers to leave the web shop.

Future research on product recommendation types should focus on two aspects. The first aspect is more theoretical in nature and should focus on the reasoning behind the preference for a type of product recommendation. Qualitative research could help in clearing up what it is that consumers like about the different types of product recommendation. The second aspect future research should focus on is, is more practical in nature and should focus on whether alternative brands product recommendation actually results in the highest profits for companies. Even though this thesis shows that alternative brands product recommendation is most preferred, it might not actually result in the highest profits for companies. In collaboration with companies, all three types of product recommendation should be tested in similar circumstances to find out what type is most profitable for companies.

6. Limitations

(27)

27

7. References

Adomavicius, G., and Tuzhilin, A. (2005) Towards the next generation of recommender systems: a survey of the state-of-the-art and possible extensions. IEEE Transactions on Knowledge & Data

Engineering, 17(6), pp. 734–749

Angell, R.J., Megicks, P., Memery, J., and Heffernan, T.W. (2014). Older shopper types from store image factors. Journal of Retailing and Consumer Services, 21, pp. 192-202.

Babin, B.J., Darden, W.R., and Griffin, M. (1994). Work and/or fun: measuring hedonic and utilitarian shopping value. Journal of Consumer Research, 20(4), pp. 644–656.

Baum, M., Schwens, C., and Kabst, R. (2015). A latent class analysis of small firms’ internationalization patterns. Journal of World Business, 50, pp. 754-768.

Baxendale, S., MacDonald, E.K., and Wilson, H.N. (2015). The Impact of Different Touchpoints on Brand Consideration. Journal of Retailing, 91(2), pp. 235-253.

Benlian, A. (2015). Web Personalization Cues and Their Differential Effects on User Assessments of Website Value. Journal of Management Information Systems, 32(1), pp. 225-260.

Ganesh, J., Reynolds, K.E., Luckett, M., and Pomirlenau, N. (2011). Online Shopper Motivations, and e-Store Attributes: An Examination of Online Patronage Behavior and Shopper Typologies.

Journal of Retailling, 86(1), pp. 106-115.

Ho, S.H., Bodoff, D., Tam, K.Y. (2011). Timing of Adaptive Web Personalization and Its Effects on Online Customer Behavior. Information Systems Research, 22(3), pp. 660-679.

Hostler, R.E., Yoon, V.Y., Guo, Z., Guimaraes, T., and Forgionne, G. (2011). Assessing the impact of recommender agents on on-line consumer unplanned purchase behavior. Information &

Management, 48, pp. 336-343.

Jeong, H.J., and Lee, M. (2013). Effects of recommendation systems on consumer inferences of website motives and attitudes towards a website. International Journal of Advertising, 32(4), pp. 539-558.

Johar, M., Mookerjee, V., and Sarkar, S. (2014). Selling vs. Profiling: Optimizing the Offer Set in Web-Based Personalization. Information Systems Research, 25(2), pp. 285-306.

Lemon, K.N., and Verhoef, P.C. (2016). Understanding Customer Experience throughout the Custoemr Journey. Journal of Marketing, JM-MSI Special Issue.

(28)

28 Melero, I., Sese, F.J., and Verhoef, P.C. (2016). Recasting the Customer Experience in Today’s Omni-channel Environment. Universia Business Review, 50, pp. 18-37.

Minsker, M. (2015). The Path to Personalized Customer Journeys. Customer Relationship

Management, pp. 20-24.

Oracle, 2012. Powering the Cross-Channel Customer Experience with Oracle’s Complete

Commerce

Rohm, A.J., and Swaminathan, V. (2002). A typology of online shoppers based on shopping motivations. Journal of Business Research, 57, pp. 748-757.

Rose, S., Hair, N., and Clark, M. (2011). Online Customer Experience: A Review of the Business-to-Consumer Online Purchase Context. International Journal of Management Reviews, 13, pp. 24-39.

Stein, A., and Ramaseshan, B. (2016). Towards the identification of customer experience touch point elements. Journal of Retailing and Consumer Services, 30, pp. 8-19.

Stone, G.P. (1954). City Shoppers and Urban Identification: Observations on the Social Psychology of City Life, American Journal of Sociology, 60(1), pp. 36-45.

Tam, K. Y., and Ho, S. H. (2006). Understanding the impact of web personalization on user information processing and decision outcomes. MIS Quarterly, 30(4), pp. 865-890.

Tauber, E.M. (1972). Why do people shop? Journal of Marketing, 36, pp. 46-59.

Verhoef, P.C., Lemon, K.N., Parasuraman, A., Roggeveen, A., Tsiros, M., and Schlesinger, L.A. (2009). Customer Experience Creation: Determinants, Dynamics and Management Strategies.

Journal of Retailing, 85(1), pp. 31-41.

Wang, Y., and Luse, A., Townsend, A.M., and Mennecke, B.E. (2014). Understanding the moderating roles of types of recommender systems and products on customer behavioral intention to use recommender systems. Information Systems & e-Business Management, 13(4), pp. 769-799.

Westbrook, R.A., and Black, W.C. (1985). A Motivation-Based Shopper Typology. Journal of

(29)

29

8. Appendix

(30)

(31)

(32)

(33)

(34)

(35)

(36)

(37)

37

Scenario 1

(38)

(39)

(40)

(41)

41

Appendix B – Overview of the dimensions and the corresponding Cronbach’s Alpha’s

Items Cronbach’s Alpha

All 33 items used to measure shopping motivations.

Web shopping convenience - Shopping from my home - Avoiding regular shopping

- Avoiding having to deal with salespeople - Having products delivered right to my home - Shopping any time of day or night

- Avoiding standing in line

- One-stop shopping, i.e. buying everything you need at one shop - Avoiding crowds

- Completing my shopping tasks quickly - Not having to travel from store to store

- Finding exactly what I want in the least amount of time

Online bidding/haggling

- Bargaining over the price of an item through an online auction - Being the winning bidder in an online auction

- Haggling over the price of a product - Submitting online bids for products

- Bargaining with a website on the price of a product

Role enactment

- Looking for great deals

- Hunting for and finding a real bargain

- Comparison-shopping to find the best product for my money

Avant-gardism

- Keeping up with new trends

- Getting to create a new “image” for myself or my home

- Being one of the first to have the latest in new fashions or new products - Keeping up with the newest fashions

Affiliation

- Chatting with other consumers who share my own interests

- Finding other consumers who are interested in the same product as I am - Interacting with other Web shoppers

Stimulation

- Interacting with websites that I am interested in - Seeing interesting websites while shopping - Just looking around at interesting websites - Finding entertaining websites

.899

.926

(could be .945 if the item ‘’Being the winning bidder in an online auction’’ was removed.)

.850

.955

.943

(42)

42 Personalized services

- Being notified of new products that interest me - Being alerted to special deals or sales

- Having emails sent to me about new products, upcoming sales events or both

All 22 items used to measure e-store attribute attitude.

e-Store essentials - Safety/security of site

- Website is certified by an online Watchdog organization - Confirmation of order/delivery

- Ease of ordering - Ease of payment

- Ease of returning merchandise - Quality of information

- Ease of contacting company

- Low-cost shipping and delivery charges - Deliveries are made in a timely manner

Offline presence

- Website company also has physical store - Physical store for website located nearby - Ability to return purchases to a physical store

Price orientation - Special deals

- Notifications about sales or new products - Frequency of sales or special deals

Website attractiveness - Attractiveness of website - Cutting-edge site

- Well-designed website

Merchandise variety

- Availability of a wide variety of products - Availability of brand-name products - Availability of latest products

All 3 items used to measure congruency between product recommendations and the shopping goal

Alternative brands product recommendation

- The products recommended by the website fit my shopping goal well - The products recommended by the website are consistent with my shopping goal

- The products recommended by the website are relevant to my shopping goal

(43)

43 Additional products product recommendation

Combination of both product recommendations

.934

(44)

44

Appendix C – Results of the post hoc Tukey test

Multiple Comparisons

Tukey HSD

Dependent Variable

(45)

(46)

(47)

(48)

(49)

(50)

(51)

(52)

(53)

53 2 1 -1,10353*_,31853 _,006 _-1,9838 _-,2232 3 ,05301 ,31372 1,000 -,8140 ,9200 4 -,42002 ,33827 ,727 -1,3549 ,5148 5 -1,46096*_,28001 _,000 _-2,2348 _-,6871 3 1 -1,15654*_,28123 _,001 _-1,9338 _-,3793 2 -,05301 ,31372 1,000 -,9200 ,8140 4 -,47303 ,30341 ,526 -1,3116 ,3655 5 -1,51397*_,23672 _,000 _-2,1682 _-,8598 4 1 -,68351 ,30839 ,180 -1,5358 ,1688 2 ,42002 ,33827 ,727 -,5148 1,3549 3 ,47303 ,30341 ,526 -,3655 1,3116 5 -1,04094*_,26842 _,001 _-1,7827 _-,2991 5 1 ,35743 ,24306 ,583 -,3143 1,0292 2 1,46096*_,28001 _,000 _,6871 _2,2348 3 1,51397*_,23672 _,000 _,8598 _2,1682 4 1,04094*_,26842 _,001 _,2991 _1,7827

(54)

54

(55)

1/30/2017

Jos van der Velde S2351617

Msc Marketing Management & Intelligence

University of Groningen

30 January 2017

| 1

Master thesis

(56)

Introduction

Most firms’ top priority is creating a strong customer

experience (Lemon and Verhoef, 2016)

(Web)personalization

Current literature on online shopper types

(57)

Research question

What type of product recommendation do the

different shopper types prefer?

(58)

Web personalization

‘’the process of profiling consumers and generating

content that matches their preferences and tastes’’ (Ho,

Bodoff, and Tam, 2011; pp. 1).

Content-based & collaborative filtering recommendation

systems (Wang, Luse, Townsend, and Mennecke, 2014)

Two product recommendation types: alternative brands

& additional products (Jeong and Lee, 2013)

(59)

Shopper types

Utilitarian & hedonic shopper types (Babin, Darden, and

Griffin 1994).

Two broad approaches to form shopper types:

- store attributes

- shopping motivations (Tauber, 1972)

(60)

Shopper types

First study regarding online shopper types (Rohm and

Swaminathan, 2004)

Extended by Ganesh, Reynolds, Luckett, and Pomirleanu

(2010)

(61)

Hypotheses

Hypothesis 1: Utilitarian online shopper types prefer

alternative brands product recommendations over additional

products product recommendations or a combination of both.

Hypothesis 2: Hedonic online shopper types prefer a

combination of alternative brands and additional products

product recommendations over solely alternative brands or

additional products product recommendations

(62)

Methodology

Web shopping convenience .899

Online bidding/haggling .926

(could be .945 if the item ‘’Being the winning bidder in an online auction’’ was removed.)

Role enactment .850 Avant-gardism .955 Affiliation .943 Stimulation .925 Personalized services .910 1/30/2017 | 8

e-Store essentials .917

(could be .923 if the item ‘’Website is certified by an online Watchdog organization’’ was removed.)

Personalization: How to address different shopper types during online shopping