Live customer support chats for
online shopping
A choice-based conjoint analysis for measuring consumer usage
preferences and the moderating role of utilitarian and hedonic
shopping motivations
Page | 2
Live customer support chats for
online shopping
A choice-based conjoint analysis for measuring consumer usage
preferences and the moderating role of utilitarian and hedonic
shopping motivations
Author:
Elles van der Veen Boterdiep 80 9712 LS Groningen Tel: +31 (0)619068756 ellesvdveen@hotmail.com Student number: 1966758 University: University of Groningen
Faculty of Economics and Businesses – Department of Marketing Qualification: Master Thesis
MSc Marketing Intelligence & Management Research Theme: Customer Live Support Chats Supervisors:
1st Supervisor: Dr. H. Risselada
Page | 3
Preface
This report was written as a master thesis to conclude my study Msc. Marketing at the University of Groningen. The reasoning for the topic of my thesis was based mostly on personal interest in the online shopping world. It is interesting to see how companies try to differentiate themselves online from others and compare what tools are provided to make the shopping process as fluent as possible for their customers. Writing the thesis was a very challenging job for me, where I learned more about myself and my abilities, while also gaining interesting new knowledge in this field of research. Overall, I am happy to conclude my study period with this result. I would like to thank several people for helping me through the process.
First of all, I would like to thank dr. H. Risselada for providing me with feedback and guiding me throughout the last few months. Moreover, I would like to thank my thesis group for helping me out when needed and providing constructive feedback. Also, I would like to thank all people who took their time to fill in the questionnaires; I would not have been able to write this thesis without them. Finally, a special thanks to my loving parents and dear friends for providing me with consistent support and believing in me throughout my studies.
Kind regards,
Page | 4
Executive Summary
As companies are seeking for new ways to engage with their customers and provide good customer support, many e-tailers start introducing a live chat function on their website. This tool allows consumers to ask advice and help during their online shopping process. However, there still might be barriers present, preventing consumers to make use of the chat. This report researched the preferences of consumers in using a live chat as a customer support tool in their online shopping process by utilizing a choice-based conjoint analysis. For this research, 223 respondents were questioned, where 110 respondents filled out a questionnaire for utilitarian shopping motivations and 113 respondents filled out a questionnaire for hedonic shopping motivations.
The attributes that were investigated to influence live chat preferences are: Identity disclosure company representative, Identity disclosure user, Waiting time and Intensity of communication. Two models were estimated to compare preferences across these shopping motivations. For both models, intensity of communication was found to be the most important attribute, followed by waiting time, identity disclosure user and identity disclosure company representative. The optimal live chat would have no video chat function, no waiting time before access, no personal information needed for access and a name and picture displayed of the company representative that the consumer is talking to. Furthermore, the research indicates that shopping motivations do have an effect on waiting time preferences of a live chat.
Then, for both models segments were created to see whether preferences differ across different consumers. The segments found are utilitarian anonymous shoppers, utilitarian video chat dislikers, utilitarian personal speed seekers, utilitarian privacy seekers, hedonic anonymous students, hedonic video chat haters, hedonic personal speed seekers and hedonic personal interaction doubters. These segments were based on the inactive covariates age, gender, education and frequency of online shopping. It was decided to delete the covariates country of heritage and the big 5 personality trait model from the process, as these variables were found to be inconclusive.
Page | 5
Table of Contents
1. Introduction ... 6
2. Theoretical framework ... 10
2.1 Live customer support chats 2.2 Factors influencing live chat preferences 2.3 Moderator 2.4. Control variables 2.5 Conceptual model 3. Research design ... 17
3.1 Methodology 3.2 Attributes and Levels 3.3 Research design 3.4 Sampling design 3.5 Data collection 3.6 Control variables 3.7 Model specification 4. Results ... 22 4.1 Descriptive statistics 4.2 Data preparation 4.3 Conjoint Analysis 4.4 Interaction effects 4.5 Model validation 4.6 Segment creation 4.7 Segment description 4.8 Hypotheses overview 5. Discussion & Conclusion ... 40
6. Managerial implications ... 42
7. Limitations & Directions for further research ... 43
8. References ... 44
Page | 6
1. Introduction
Nowadays, technologies have a major impact on our lives, with the internet as one of the most influential innovations. The internet provides endless possibilities, such as online search capabilities and social networking. Moreover, it enables people to shop online instead of going to a physical store nearby. In 2011, in the United States the e-commerce sales has already quadrupled since 2002, namely from $42 billion till $162 billion (Statista, 2012). There are plenty of online retailers who only engage in e-commerce and do not have bricks and mortar stores, for example the biggest online retailer in the US is Amazon, with 100 milion different visitors per month (Statista, 2012).
These e-retailers have to find alternative ways to interact with their customers, as opposed to face-to-face contact in brick-and-mortar stores. In a physical store, salespeople can be consulted by shoppers with purchase-related decisions (Olshavsky, 1973). However, in the online world other sources of information need to be found, that can assist consumers in their decision-making process. It is very important that online shoppers are assisted, as a study by LivePerson (2013) that some of the top reasons for abandonment of online purchases in their shopping basket are lack of information about a product, service or delivery and difficulties with receiving help from online customer support. Companies therefore need to find ways to prevent this from happening, a customer live support chat could be the way to do so. The live chat is currently viewed by companies as a cost-effective way to reduce purchasing risk for consumers, namely through increasing social interaction, increasing responsiveness to consumer questions, and through the personalization of the shopping experience (Elmorshidy, 2013). However, still much is unclear about how consumers think about a live chat and how they would prefer to use one.
Page | 7
Young (1999) confirmed this, as they found that a very important barrier to online shopping was the inability to talk to a salesperson. This lack of personal assistance could in turn, translate into lost sales. Therefore, online retailers created other ways to assist shoppers, to keep them from leaving the website without buying. One new solution for customer support that is introduced is an artificial intelligence program, in the form of anthropomorphic information agents (AIAs). These could provide real time product information by means of a chat program, while also giving a human touch to the website (Sivaramakrishnan et. al, 2007).
An even more personal variant of this program is the live customer support chat, where a human company representative communicates with the customer instead of a computer that generates answers to customer questions. Aslanzadeh et al. (2014) showed that a consumer’s view of the ecommerce website will be enhanced when there is the availability of a live chat or video chat. Other research in this field explains that a live chat is a very effective customer support tool, as it enhances customer trust and as a result, increases the likelihood that a consumer browsing on the website becomes a buyer (Komiak et al., 2005). A customer support life chat can act as a way to develop strong and enduring customer-brand relationships, which could in turn then have a positive influence on business outcomes and profitability (Palmatier et al., 2006). Therefore, it is important to analyze how these positive outcomes can be best reached. However, up to now little research has been performed in this field, as a live chat is an upcoming relationship management feature used by more and more ecommerce companies.
Research has found that website attributes affect the perceived quality of the website and satisfaction (Seethamraju, 2006). It is very likely that the presentation characteristics of a live chat might influence its perceived usefulness in helping consumers in their product decision-making process. However, it has not been researched yet what the exact preferences for this new and uprising type of consultation are.
Thus, little is still known about the live customer support chats as a tool for customer interaction, even though the use of live chats by companies is an uprising phenomenon. It especially remains unclear what the exact consumer’s preferences are for engaging in a live chat conversation with a sales employee.
Page | 8 What are the consumers’ preferences for consulting a live support chat in an online retail setting to create the best customer service tool?
For this research it will be assumed that the consumer assesses its products of interest based on information found online, without the possibility of face-to-face assistance of a salesperson or the possibility of physically assessing the product. Furthermore, it will be assumed that the consumer is still uncertain about his/her purchase and therefore is motivated to consult a live customer support chat for more information and advice. Given the fact that the shopper is interested in using the live chat option, there might be features of the chat that may influence their appreciation of the live chat.
A choice-based conjoint analysis will be performed to measure different attribute preferences for consulting a live customer support chat. The attributes that will be investigated are the identity disclosure of the company representative and identity disclosure of the user, waiting time and the intensity of communication.
Then, this paper will investigate the moderating role of frequency of online shopping on the live chat attributes and the moderating role of utilitarian vs. hedonic shopping motivations. Furthermore, it will be researched whether there are differences in live chat preferences for gender, age, education and country of heritage. Consequently, segments will be created to describe groups of consumers that have distinctive different preferences for the live chat, based on the socio-demographic information given and the big 5 personality trait model. Two models are created, one for consumers with hedonic shopping motivations and one for consumers with utilitarian shopping motivations, indicating their preferences towards a chat. Then for both models, 4 segments of consumers were found that had different preferences.
Page | 9
Page | 10
2. Theoretical framework
In this chapter, an overview of the existing literature in the field of customer live support chats will be given, to provide a deeper understanding of the topic and its relevance. Furthermore, the expected effects of the factors influencing the usability of the live customer support chat will be explained. The topics that are first of all, the literature in the field of live customer support chats. Then, the variables will be discussed that are expected to influence the consumer preferences for a live chat. Firstly, the intensity of communication of the live chat will be discussed. Then, the waiting time for a live chat interaction will be discussed. Finally, two personification factors will be discussed, namely the identity disclosure of the user and the identity disclosure of the company representative. Then, the moderating role of utilitarian vs hedonic shopping motivations will be discussed. Finally, the covariates age, gender, education, country of heritage, frequency of online shopping and the big 5 personality trait model will be discussed. The expected effects will be concluded in a conceptual model.
2.1 Live customer support chats
Page | 11
become a buyer (Komiak et al., 2005). Mainly, the use of an AIA means that consumers can ask questions about customer service, but not about the products that are sold.
Consequently, some e-commerce websites now introduced a Live Customer Support Chat, so that customers can also be helped by a company representative with product related questions, right on the spot. Online salespersons have found to be useful for reducing information overload on the website and elsewhere, they can provide online customers with recommendations on suitable products and facilitate customers in their online shopping decision-making. Which are similar roles that human salespersons have in physical stores (Komiak et al., 2005). Furthermore, a study by Liveperson (2013) has indicated that 51% of their respondents are more likely to purchase from a website if they could get answers via Live Chat. Moreover, 48% of the respondents indicated that they are more likely to return to a website with a Live Chat and 41% say they are more likely to trust a brand if a Live Chat is available on the website (Liveperson, 2013).
Page | 12 2.2 Factors influencing live chat preferences
2.2.1 Intensity of communication
For a Live Chat, there are different ways available to communicate with a company representative. The most common form of communicating is a text based chat, however some companies are now also offering an option for video chat. Research by LivePerson (2013) indicates that 7% of their respondents is interested in using a video chat to communicate with a company representative for help. Also, research by Andrews & Haworth (?) indicates that a lack of personalized interactions in the chat leads to a negative evaluation of the customer support tool. Therefore, it might be an opportunity for companies to provide a video function in their live chat interface, to personalize the interaction as much as possible. This leads to the following hypothesis:
H1: A higher intensity of communication in a live chat conversation has a positive influence on the chat preferences
2.2.2 Waiting time
Before the live chat can be an efficient customer support tool, consumers browsing the web site need to become aware of the availability of the live chat. Research has found that 51% of the online shoppers either try once or give up immediately when seeking help concerning an online purchase (LivePerson, 2013). Also, research has found that ‘accessibility’, explained by speed, ease of access, availability of the site, and ease of navigation, were the most important determinants of web quality (Seethamraju, 2003). Therefore, it is important that help is offered clearly and immediate, as it may affect how people view the website as a whole. However, as a live chat is available in real time, it might lead to some difficulties, as people are able to shop online 24/7. For instance for the e-commerce website Spartoo.nl (2014) a company representative described in a live chat conversation that “We have the live chat for a couple of months now. However, it is possible that we are not always online, but we aim to activate the chat as often as possible”. Thus, as a fast response is not always given via telephone or email, a Live Chat is found to be a quick solution in assisting consumers. For many consumers this immediacy of receiving assistance is very important, as they have increasing demands on their time. When consumers are assisted quickly, this has found to increase the likelihood of purchase, return visits and enhanced levels of customer trust in the brand (LivePerson, 2013). Therefore, the following hypothesis was proposed:
Page | 13
2.2.4 Identity disclosure user
Before a live chat can be used, companies sometimes ask for personal information of the customer. This might ensure that less prank chats will be started by customers, which is positive for the company. However, for the customer disclosing personal information might be a barrier to start the conversation. Recent studies have indicated that information privacy is one of the most important obstacles for the growth of e-commerce (Hann et al, 2007)(Andrews & Haworth, ?). The reason for this is the lack of trust of consumers in the online security and fear of lack in enforcement of privacy laws (Hann et al, 2007). Some people may have a different disposition towards privacy, dependent on the type of website. A website could achieve a good privacy disposition through pleasant design, content and functionality (Li et al, 2011). Privacy has always been an important topic of discussion in the online world. In online shopping, “most e-commerce companies collect, store, and exchange personal information obtained from individuals and use that information to support marketing strategies, gain greater insights into individuals’ behavior, and meet their needs and wants more effectively” (Boritz et al., 2011). As most research suggests that customers prefer not to disclose much information about themselves in the online shopping process, the following hypothesis is proposed:
H3: High identity exposure of the user of the live chat has a negative effect on the chat preferences.
2.2.5 Identity disclosure company representative
Page | 14
competent when a name is disclosed. A review from a reviewer with a name disclosed has proven to get a more positive judgment, leading the review to be perceived more credible (Chaiken, 1980). Also, according to Swan et al. (1991), a pleasant interface or appearance of a medium can help to build trust, while an unpleasant interface does the opposite. This may indicate that including a pleasant picture of the company representative in the interface may increase trust, and may therefore be more preferred by consumers than a more anonymous interface. Therefore the following hypothesis will be proposed:
H4: High identity exposure of the company representative in a live chat has a positive effect on the chat preferences
2.3 Moderator
Consumers can shop online for different reasons. According to Childers et al. (2001), it is important to consider the motivations that people have to utilize different sites and tools on the web, as only then success could be achieved through electronic ecommerce. The authors state that the most important two consumer motivations for online shopping are utilitarian and hedonic shopping motivations. The former can be described as consumers that purchase products in an efficient and timely fashion to achieve their goals with minimum irritations. The latter motivation for online shopping relates more to enjoying shopping as an experience and looking around for products to buy (Childers et al., 2001).
Page | 15
be the desire for entertainment; informants looked up information for fun (Schlindler & Bickart, 2005).
Having these utilitarian and hedonic motivations for online shopping and searching online information seeking in mind, differences might be present between consumers’ preferences when they want assistance via a live chat. As hedonic shoppers are more experience-seeking, they are expected to have more patience when using the live chat for help. While on the other hand, the utilitarian shoppers may sooner dismiss the chat when there is waiting time involved. Therefore, the following hypoteses are proposed
H5a Utilitarian shopping motivations will increase the negative effect of waiting time on consumer preferences
H5b Hedonic shopping motivations will diminish the negative effect of waiting time on consumer preferences.
2.4 Control variables
Page | 16 2.5 Conceptual Model
De hypotheses can be concluded in the conceptual model below:
+ - - + +/- Intensity of Communication Video chat Text chat Text chat with
video function
Waiting Time
No waiting time
2nd person in line
5th person in line
Identity exposure user
No information
Enter Name
Create account
Identity exposure company representative
No information
Name
Name & Picture
Live customer support chat preferences
Controls:
Page | 17
3. Research design
In this chapter, the method of data collection will be discussed, then it will be explained how the data was measured. Finally, it will be discussed how the data will further be analyzed.
3.1 Methodology
In this report the features of a live chat will be analyzed to discover what people prefer in using this tool when shopping online. The most suitable method for understanding consumer preferences of multi-attribute products, services or of other objects is the choice-based conjoint analysis (Hagerty, 1985). As a live chat can be seen as a product that cannot be actually bought, but can be used, and exists of a set of attributes, a choice-based conjoint analysis would be suitable in researching consumer preferences for the live chat design. By letting respondents of this study choose between 3 different profiles, the optimum combination of attribute levels will be determined. 2 surveys will be used, one including the moderator for utilitarian shopping motivations and the other for hedonic shopping motivations. Furthermore, interaction effects will be measured with the data collected through the choice-base conjoint analysis. Finally, segments will be defined to clarify whether certain groups of respondents have significantly different preferences.
3.2 Attributes and levels
For the choice-base conjoint analysis, different attributes and levels of the customer live support chat need to be defined; these are depicted in table 1. To keep the standard error for each parameter as low as possible, the number of levels should be kept low. Also, the number of levels per attribute should be balanced evenly, to avoid the number-of-levels effect. This effect namely occurs when the number of levels are not distributed equally across attributes, and leads to higher importance of the attributes with more levels.
Attributes Level 1 Level 2 Level 3
Waiting time Start chat now! 2nd person in line, please wait..
5th person in line, please wait..
Intensity of Communication Video chat Text chat Text chat with video option
Identity disclosure user No personal
information required
Please enter name for access
Please create account for access
Identity disclosure company representative
Name & Picture Name No information
Page | 18 3.3 Research design
To establish a successful experimental design it is decided that the attribute ‘Intensity of communication’ and ‘Waiting time’ will be described separately in the conjoint analysis choices, while the other attributes will be incorporated in pictures of live chat designs in the choice sets. The amount of combinations possible for the conjoint analysis is 3x3x3x3= 81, which would be the full factorial design. However, for the respondents it is tiring too evaluate that many stimuli, therefore a fractional factorial design will be used in this study. This means that respondents only need to evaluate a subset of all possible combinations of attributes. In order to create a successful fractional factorial design, the design needs to be balanced and orthogonal. The former, as explained before, relates to the amount of levels per attribute, while the latter implies that each level combination should appear an equal number of times. This pair wise balance assures that no correlation between attributes exists. To create a balanced and orthogonal fractional factorial design, orthogonal arrays need to be assessed. A template fractional factorial design was found on the website of W.F. Kuhfeld, which can be found in appendix A. This coded orthogonal array was proven to be successful in previous conjoint studies and is developed for a design with 4 attributes with each 3 levels. Correspondingly, the design needed 9 choice sets to be created with each 3 choices of live chats.
This design was used for both studies, as two studies were created to manipulate utilitarian and hedonic shopping motivations. For the former, respondents were asked to imagine a situation where they need to shop for a coffee machine and do not enjoy looking for one. For the latter, respondents were asked to imagine a situation where they wanted to shop for a shirt and enjoyed shopping online. The different surveys are displayed in appendix B.
Page | 19
Figure 1 Personality traits measurement scales (Gosling et al., 2003)
For the second and final moderator ‘frequency of online shopping’ literature by Cao et al (2010), Shergill & Chen (2005) and Cao, Zu, Cleaveland & Douma (2010) and Farag et al (2007). The scale items and their corresponding scale points are depicted in table 3. Shergill & Chen (2005) asked their respondents how many times the respondent had made purchases on the web in the recent year, these were then put into categories. Respondents by the study from Cao et al (2010) were asked to indicate their online buying frequency on a scale from “never” to “more than once a month”. Also they were asked to indicate their online-searching frequency concerning product information or stores. Farag et al (2007) used an 8-point Likert scale for their respondents to indicate the frequency of online searching of information about products and stores and used a continuous question on the amount of online purchases for private use in the recent year. The scale items and scale points of these articles were combined to get a complete as possible image on the frequency of online shopping of the respondents. Respondents were asked to rate three items on a 7-point Likert scale, the scale items and scale points are depicted in table 2.
Scale items Scale points
I look for information about products and/or stores online I browse for products to buy online
I buy products online
Never
Less than once a month Once a month
More than once a month Once a week
More than once a week Daily
Page | 20 3.4 Sampling design
To be able get significant results by the survey, a rule-of-thumb was used to determine the right sample size (Johnson and Orme, 1996). The formula used to calculate the sample size that was established is:
𝑛𝑡𝑎
𝑐 ≥ 500
Where n is the number of respondents, t is the amount of choice sets, a is the number of alternatives per choice set and c is the number of analysis cells. However, critiques found that it would be better to have more than 1000 representations of the main level effects, instead of 500 (sawtoothsoftware.com, 2010). Therefore the amount of 113 would be optimal where: 113x9x3/3=1,017. For the other survey 110 respondents would be sufficient where: 110x9x3/3=990 representations.
3.5 Data collection
For the data collection, two online surveys were distributed, one for utilitarian shopping motivations and one for measuring hedonic shopping motivations. The website used to create the surveys is preferencelab.com, which is a website specifically designed for conducting CBC analyses. The surveys were distributed via Facebook and e-mail and respondents were asked to share the survey link with their friends, which created a snowball effect. People from different age categories and different backgrounds were approached to get a diversified portfolio. The completion rate of the hedonic survey was 41.1% and the average time to complete the survey was 7.3 minutes. The completion rate of the utilitarian survey was 38.3% with an average completion time of 8.6 minutes. Overall, 110 respondents filled in the utilitarian survey and 113 people the hedonic survey.
Page | 21 3.6 Control variables
For this study some control variables were used to be able to create segments and see whether the different segments have different preferences for a live chat. The control variables used are gender, age, education and country of heritage. Moreover, respondents are asked to fill in their frequency of online shopping and lastly, respondents were asked to fill in some questions that would determine their personality types based on the big 5 personality traits model. With these personalities, segments can be characterized even more.
3.7 Model specification
Before starting the analysis the utility function will be depicted that is derived from the attributes and levels established for the CBC analysis. The live chats are presented in combinations of attributes, where consumers attach part-worth utilities to each attribute level. This means that the systematic utility of consumer n for live chat i, is the sum of the part-worth utilities and can be presented in the following utility function:
𝑉𝑖 = 𝛽0+ 𝛽𝐼𝐷𝐶𝑅𝑁𝑃𝐼𝐷𝐶𝑅𝑁𝑃𝑖 + 𝛽𝐼𝐷𝐶𝑅𝑁𝐼𝐷𝐶𝑅𝑁𝑖+ 𝛽𝐼𝐷𝐶𝑅𝑁𝐼𝐼𝐷𝐶𝑅𝑁𝐼𝑖+ 𝛽𝐼𝐷𝑈𝑁𝐼𝐼𝐷𝑈𝑁𝐼𝑖 + 𝛽𝐼𝐷𝑈𝑁𝐼𝐷𝑈𝑁𝑖 + 𝛽𝐼𝐷𝑈𝐴𝐼𝐷𝑈𝐴𝑖+ 𝛽𝑊𝑇𝑁𝑊𝑇𝑁𝑖 + 𝛽𝑊𝑇2𝑊𝑇2𝑖 + 𝛽𝑊𝑇5𝑊𝑇5𝑖 + 𝛽𝐼𝐶𝑉𝐼𝐶𝑉𝑖 + 𝛽𝐼𝐶𝑇𝐼𝐶𝑇𝑖+ 𝛽𝐼𝐶𝑇𝑉𝐼𝐶𝑇𝑉𝑖 Where: V= Utility β0 = Constant
βk = part-worth utility of consumer n for attribute level k
IDCRNP = Identity disclosure company representative: name & picture IDCRN = Identity disclosure company representative: name
IDCRI = Identity disclosure company representative: no information IDUNI = Identity disclosure user: no information
IDUN = Identity disclosure: name IDUA = Identity disclosure: account WTN = Waiting time: none
WT2 = Waiting time: 2 minutes WT5= Waiting time: 5 minutes
ICV = Intensity of communication: video ICT = Intensity of communication: text
ICTV = Intensity of communication: text & video
Page | 22
4. Results
In this section the CBC analysis results will be discussed to provide a clear view on consumer preferences for Customer Support Live Chats. First of all, some descriptive statistics will be given, to create some insights on the survey respondents. Then, the results of the CBC analysis will be given by creating two models, one for consumers with hedonic shopping motivations and one for consumers with utilitarian shopping motivations. The results will indicate the relative importance of the live chat attributes, and segments will be created for both motivations to display different groups of people with different preferences concerning the live chat.
4.1 Descriptive statistics
To measure the customer preferences for customer life support chats, two surveys were conducted with 223 respondents in total. Of the respondents, 42.6 % were male and 57.4% were female. Most respondents were Dutch, namely a percentage of 92%. Moreover, the majority of the respondents had some level of higher education, namely 53% has a university degree or higher and 34% has a HBO (higher education) diploma. The age of the respondents ranged from 16 till 74, with an average age of 32. An overview of the descriptive statistics can be found in table 2 below.
Gender Country of Heritage
Male Female Total Frequency 95 128 223 Percentage 42.6% 57.4% 100% The Netherlands Germany Belgium France Spain Italy Other Total Frequency 205 6 1 0 1 1 9 223 Percentage 91.9% 2.7% 0.4% 0% 0.4% 0.4% 4.0% 100% Education Age
High school or lower MBO HBO University or higher Total Frequency 11 17 76 119 223 Percentage 4.9% 7.6% 34.1% 53.4% 100% Minimum Maximum Mean Standard Dev. 16 74 31.91 13.938
Page | 23
Furthermore, some descriptive statistics can be shown related to the prior use of a customer support live chat, their attitude towards customer support live chats in general and the frequency of shopping of the respondents. First of all, the results indicate that the vast majority of the respondents has not used a customer support live chat before in their shopping process, namely 70%, the other 30% did make use of a live chat before. Furthermore, respondents were asked to rate their attitude towards a live chat on a list of scales from 0-6. The average results per scale item are depicted in table 4. For all scale items, the rating are quite high, indicating that most people have a good attitude towards a live chat. The scale item undesirable – desirable with an average rating of 3.66 indicates that the respondents did not find it a must to have a live chat on a website for online shopping. However, most people do have a positive attitude when a chat option is available.
Scale item Average Rating
Bad - Good 4.46 Unpleasant - Pleasant 4.15 Unfavorable – Favorable 3.92 Useless - Useful 4.22 Undesirable - Desirable 3.66 Negative - Positive 4.34
Irritating – Not Irritating 3.93
Table 4 Attitude towards a Live Chat
Figure 2 Frequency of online shopping
In figure 2 the frequency of online shopping has been presented for the respondents per scale question. It shows that most people do use the internet often for looking up product information and information on stores, namely 53% of the respondents does this more once a week or more. Also, quite some people browse for products to actually buy online monthly,
1% 9% 14% 24% 26% 21% 5% I look for information about products and/or stores 1% 14% 17% 30% 17% 18% 3% I browse for products to buy online 1% 42% 29% 21% 5% 1% 1%
I buy products online
Never
Less than once a month
Once a month More than once a month
Once a week More than once a week
Page | 24
namely a majority of 30% indicated to do this more than once a month. However, not many people actually buy products online that often as well, namely 71% indicated to buy products online once a month less. Only very little people indicate to never shop and browse for products and stores online, namely 1% of the respondents.
4.2 Data preparation
In order to successfully analyze the conjoint data, the data needed to be prepared for the Latent Gold program. First of all, the data of the two studies needed to be combined, and the data from the conjoint studies and additional data about the demographics needed to be combined. Then, the shopping motivations were coded into a dummy variable, one for indicating utilitarian shopping motivations and zero for indicating hedonic shopping motivations. Then two factor analyses were performed to create the variables Attitude towards a live chat & Frequency of Online Shopping. Both are unrotated analyses, as the factors directly loaded well. The factor loadings can be found in Appendix C.
4.2.1 Attitude towards the Live Chat
Page | 25
4.2.2 Frequency of Online Shopping
Secondly, a factor analysis was performed to create the variable Frequency of Online Shopping. For this factor the KMO = .654, and is therefore appropriate. Also, Bartletts test of sphericity indicates that the variables are significantly correlated, with a p-value of .000. Then, the Eigenvalue is 2.146, which is again sufficient and the amount of variance explained is 71.542%, which is also sufficient. Finally, the factor loadings are all above >.5, so all are sufficient for the factor. The scree plot confirms that creating one factor for these variables is appropriate. Then, when performing a reliability analysis, it was found that for the variable Frequency of online shopping the factor is also strong enough. For this variable the Cronbach’s alpha gives a value of .796, which is good as it is above .6. Also this value does not rise when items are deleted. Therefore, this factor will be used in defining segments. 4.2.3. Big Five Personality Trait Model
Page | 26 4.3 Conjoint Analysis
The program Latent Gold has been used to analyze the CBC conjoint outcomes. All attributes are nominal, therefore parth-worth utilities can be estimated per attribute. To research whether there are some differences concerning attribute importance for the two shopping motivations, Two separate models are estimated for each shopping motivation. Then, it can be determined what attribute of the Live Chat is most important for consumers. For each choice set consumers had to make decision in which live chat they most preferred using. Also, after indicating their decision they had to answer a question whether they would actually make use of the chosen chat if it was presented in that way. By the means of SPSS, a crosstabulation was created for both surveys to indicate how often people would actually use the live chat of their preference, which can be found in appendix E. The results show that in for the consumers with the utilitarian shopping motivations, in 659/990 * 100% = 66.6% of the cases indicated that they would use the chat of their preference. For the hedonic shoppers, this percentage is 75,5%. This indicates that hedonic shoppers are more lenient and thus more likely to make use of the chat, even though it is not exactly what they would prefer. Where for the utilitarian shoppers, they are less accepting and would quicker not make use of the chat if there are some barriers present.
Page | 27
Attributes Class1 Wald P-value Range Importance
Company Representative 25.1453 3.5e-6 0.457 10%
1 Name and picture 0.2653*
2 Name only -0.1917*
3 No info -0.0736
Personal information disclosure 117.6334 2.9e-26 1.2664 28%
1 No info 0.4895*
2 Name only 0.2874*
3 Create account -0.7769*
Waiting time 163.2375 3.6e-36 1.3463 29.8%
1 Start chat now! 0.6987* 2 2nd person in line -0.0511 3 5th person in line -0.6476*
Intensity of communication 179.9061 8.6e-40 1.4547 32.2%
1 Video chat only -0.8618*
2 Text chat 0.5929*
3 Text chat with video function 0.2689*
* significant with an significance level of 95%, Z-value >1,95
Table 5 Parameter estimates and attribute importance Utilitarian shopping motivations
Then, the same was done for the study with hedonic shopping motivations. As can be seen in table 6, all variables are significantly contributing the live chat preferences, as the P-values of all attributes are <0.0001 again. Then, by calculating the relative importance of the attributes, it can be seen whether there are some differences between the two shopping motivations. Again, the parameter ranges are calculated followed by the calculation of the relative attribute importance where each range is divided by the sum of the total ranges (0.5255+1.3648+1.1053+1.8813=4.8769).
Attributes Class1 Wald P-value Range Importance
Company Representative 34.4761 3.3e-8 0.5255 11%
1 Name and picture 0.3306*
2 Name only -0.1356*
3 No info -0.1949*
Personal information disclosure 138.7866 7.3e-31 1.3648 28%
1 No info 0.4927*
2 Name only 0.3794*
3 Create account -0.8721*
Waiting time 94.6233 2.8e-21 1.1053 23%
1 Start chat now! 0.5262* 2 2nd person in line 0.0528 3 5th person in line -0.5791*
Intensity of communication 257.9661 9.6e-57 1.8813 39%
1 Video chat only -1.1334*
2 Text chat 0.7479*
3 Text chat with video function 0.3855*
* significant with an significance level of 95%, Z-value >1,95
Table 6 Parameter estimates and attribute importance Hedonic shopping motivations
Page | 28
studies show that the intensity of communication was the major driver for the live chat choices.
Based on the information presented in the tables above, graphs can be created that depict the part-worth utilities of the attributes (Figure 3). For both shopping motivations, waiting time has an almost linear effect, namely the longer the waiting time, the less preferred the live chat is by consumers. The direction of the parameters are quite similar for both shopping motivations, only the variable company representative shows a slight difference, the figures indicate that customers with utilitarian shopping motivations prefer no information on the company representative over a name only, while for consumers with hedonic shopping motivations, the less is disclosed about the company representative the least preferred the live chat is. From the findings it can be concluded that the optimal live chat would consists of the following attribute levels: Name and picture of the company representative visible, no personal information required for accessing the chat, no waiting time and a text chat.
Figure 3 Part-worth utilites per attribute Utilitarian shopping motivations
-0,3 -0,2 -0,1 0 0,1 0,2 0,3 1 Name and picture
2 Name only 3 No info
Ut ility Company representative -1 -0,8 -0,6 -0,4 -0,2 0 0,2 0,4 0,6 1 No info 2 Enter Name 3 Create account Ut ility
Personal information disclosure
Page | 29
Figure 4 Part-worth utilites per attribute Utilitarian shopping motivations
4.4 Interaction effects
Now that the main-effect model has been estimated, it might be possible that some interactions between the variables occur as well. As the variables Intensity of Communication, Identity Disclosure Company Representative and Identity Disclosure user are all related to the degree of personalization of the live chat interaction, interactions could be present between these variables. Therefore, some interaction variables were calculated by creating new variables; the different attribute variables were multiplied with each other. Then, these new variables were inserted as attributes in Latent Gold and tested. The outcomes are presented in table 7 & 8 below.
Attribute Parameter z-value Wald p-value
Identity Disclosure Company Representative +
Identity disclosure user -0.0342 -0.0000 0.0000 1.00
Identity Disclosure Company Representative +
Intensity of communication 0.0218 0.1072 0.0115 0.91
Identity Disclosure User + Intensity of
Communication -0.3072 -1.3927 1.9396 0.16
Indentity Disclosure User + Identity Disclosure Company Representative + Intensity of Communication
-0.0220 -0.2019 0.0408 0.84
Table 7 Interaction attributes Utilitarian shopping motivation
-0,3 -0,2 -0,1 0 0,1 0,2 0,3 0,4 1 Name and picture
2 Name only 3 No info
Ut ility Company representative -1 -0,5 0 0,5 1
1 No info 2 Name only 3 Create account
Ut
ility
Personal information disclosure
-0,8 -0,6 -0,4 -0,2 0 0,2 0,4 0,6 1 Start chat now! 2 2nd person in line 3 5th person in line Ut ility Waiting time -1,5 -1 -0,5 0 0,5 1 1 Video chat only
2 Text chat 3 Text chat with video function
Ut
ility
Page | 30
Attribute Parameter z-value Wald p-value
Identity Disclosure Company Representative +
Identity disclosure user 0.0020 0.0000 0.0000 1.00
Identity Disclosure Company Representative +
Intensity of communication 0.2130 0.9589 0.9195 0.34
Identity Disclosure User + Intensity of
Communication -0.4084 -1.8011 3.2440 0.072
Indentity Disclosure User + Identity Disclosure Company Representative + Intensity of Communication
0.0024 0.0202 0.0004 0.98
Table 8 Interaction attributes Hedonic shopping motivation
From the analysis, it can be seen that none of these interaction variables for consumers with utilitarian shopping motivations are significant with a significance level of 95%, as the z-values are not > 1.95 and the p-z-values are not <0.05. However, for consumers with hedonic shopping motivations, the interaction between identity disclosure of the user and the intensity of the communication is almost significant with a p-value <0.05, namely with a p-value of 0.072 and a corresponding z-value of -1.8011. This means that consumers who do not like to disclose their information before entering are also more likely to not choose a high intensity in intensity of communication like a video chat.
4.5 Model validation
Now that the attribute importance for the live chat design has been determined. It needs to be checked whether the models and information produced is valid. First, it will be checked how well the models fit the data by calculating the goodness of fit with the likelihood ratio test. Then, the hit rate will be calculated to see whether the model is good in predicting choices.
First, in the likelihood ratio test, it will be checked whether the model is significantly better than the null model. The formulas presented that were used for this test can be found in appendix F, with the outcomes of the calculations presented in table 9, where Model 1 represents the model for utilitarian shopping motivations and model 2 the model for hedonic shopping motivations.
The outcomes of the Chi-square calculations were looked up in a Chi-square distribution table and were found to have p(675.3538) < 0.0001 and p(607.87013) < 0.0001, therefore the estimated models’ parameters are significantly different from zero. Furthermore, R2 and R2adj
Page | 31
Then, the hit rate will be calculated, by use of the Latent Gold prediction table (depicted below), to indicate how good the model is in predicting consumer choices.
Prediction Table Estimated Observed 1 2 3 Total 1 211.0 124.0 22.0 357.0 2 67.0 257.0 14.0 338.0 3 52.0 59.0 184.0 295.0 Total 330.0 440.0 220.0 990.0
Table 10 Latent Gold Prediction Table Utilitarian shopping motivation
The prediction table shows that this 1-class (aggregate) model for utilitarian shopping motivations correctly predicts 211 of the 357 alternative 1 responses, 257 of the 338 alternative 2 responses and 184 of the 295 alternative 3 responses. This means that overall 211+257+184 = 652 of the observed 990 choices were predicted correctly. The corresponding hit rate can then be calculated as followed: (211+257+184)/990*100%=65,86% hit rate. The hit rate should be at least above 25%, so this model is quite good in predicting the respondents’ choices. Then, the model for the hedonic shopping motivation indicates that overall 220+239+232=691 of the observed 1017 were predicted correctly. This means the corresponding hit rate for the hedonic model is 691/1017*100%=67.94%, which is also quite good. Therefore, based on likelihood ratio test and the hit rate, we can conclude that both models are sufficient in predicting consumer’s choices.
Model 1 Model 2
Number of respondents (n) 110 113
Number of Choice sets (c) 9 9
Number of alternatives (m) 3 3 Number of parameters (df) 4 4 LL(0) -1087.626166 -1117.288698 LL(β*) -783.6911 -779.6118 Chi-square 675.3538 607.87013 R² 0.27577 0.298648772 R²adj 0.27577 0.298648772
Table 9 Goodness of fit
Prediction Table Estimated Observed 1 2 3 Total 1 222.0 61.0 73.0 356.0 2 69.0 239.0 34.0 342.0 3 48.0 39.0 232.0 319.0 Total 339.0 339.0 339.0 1017.0
Page | 32 4.6 Segment creation
As different people have different preferences, there is heterogeneity in the data. Therefore, segments will be created based on socio-demographic information, namely the control variables age, gender, education and people’s frequency of online shopping. The country of heritage will not be used for segmenting, as too little foreign consumers filled out the surveys to create clear segments based on that. In order to create segments, a preference-based segmentation with latent classes will be used, to compare a range of one to ten class models, to find out which model would provide the most optimal results. In order to decide which model is best, Bayesian information criteria (BIC), Akaike information criteria 3 (AIC3) and Consistent information criteria (CAIC) will be compared. These criteria indicate the relative fit of the models, while penalizing the models for the amount of parameters added. This will be done for both models: utilitarian shopping motivations and hedonic shopping motivations. An overview of the models with their scores can be found in appendix G.
4.6.1 Utilitarian segmentation
Page | 33
4.6.2 Hedonic segmentation
Then, for the hedonic shoppers the same was done with all covariates set as active. Here, the BIC and CAIC were also lowest for the 3-class model, with values of 1,499.1797 and 1,537.1797 respectively. The AIC3 is lowest for the 6-class model, with the value of 1,411.1272 and the classification error is lowest for the 4-class model, with the value of 0.0383. However, the covariates again were insignificant, thus models were estimated for setting age as active, as this variable was significant on the 90% significance level and for setting all variables as inactive. An overview of the scores can again be found in appendix G. The results indicate that the best model is the 4-class model, with all covariates set as inactive, with a BIC of 1443.0048, an AIC3 of 1382.5462 and a CAIC of 1478.0048, also the standard error was the lowest. Moreover, the segment classes are all about the same size. Thus, 4 segments will be created for the hedonic shoppers.
Next, now that it is determined that 4 segments for both studies are most sufficient. These segments need to be described in terms of their preferences. In appendix H, the corresponding parameters are displayed per segment and the covariates for both utilitarian and hedonic shopping motivations. As the covariates were not significantly in predicting consumer preferences, they can be used as inactive variables to describe the segments. In figure 5 on the next page, the directions of the attributes are displayed per segment class. These will be further discussed in the following section: describing segments. Finally, the relative attribute importance is displayed in table 13, which is important for describing the segments.
Relative Importance Attributes Class 1 Class 2 Class 3 Class 4
Company Representative 8.11% 11.39% 33.94% 16.14%
Personal information disclosure 40.85% 3.04% 16.71% 50.42%
Waiting time 17.40% 26.09% 42.37% 23.56%
Intensity of communication 33.65% 59.48% 6.98% 9.88%
Segment size 36 28 23 23
Table 12 Relative attribute importance utilitarian
Relative Importance Attibutes Class 1 Class 2 Class 3 Class 4
Company Representative 11.07% 3.56% 19.05% 47.43%
Personal information disclosure 53.31% 8.07% 12.87% 6.43%
Waiting time 5.32% 21.50% 58.09% 7.38%
Intensity of communication 30.30% 67.86% 9.99% 38.76%
Segment Size 57 29 17 10
Page | 34
Figure 5 Attribute directions per segment class, utilitarian shopping motivation
Figure 6 Attribute directions per segment class, hedonic shopping motivation
-1 -0,5 0 0,5 1 1,5 1 Name and picture 2 Name only 3 No info Ut ility Company representative Class1 Class2 Class3 Class4 -4 -2 0 2 4
1 No info 2 Name only 3 Create account
Ut
ility
Personal information disclosure
-1,5 -1 -0,5 0 0,5 1 1,5 2 1 Start chat now! 2 2nd person in line 3 5th person in line Ut ility Waiting time -4 -3 -2 -1 0 1 2 1 Video chat only
2 Text chat 3 Text chat with video function Ut ility Intensity of communication -2 -1 0 1 2 1 Name and picture 2 Name only 3 No info Ut ility Company representative Class 1 Class 2 Class 3 Class 4 -3 -2 -1 0 1 2
1 No info 2 Name only 3 Create account
Ut
ility
Personal information disclosure
-2 -1,5 -1 -0,5 0 0,5 1 1,5 1 Start chat now! 2 2nd person in line 3 5th person in line Ut ility Waiting time -6 -4 -2 0 2 4 1 Video chat only
2 Text chat 3 Text chat with video function
Ut
ility
Page | 35 4.7 Segment descriptions
Based on the utility scores and attribute importance, segments were created. They are further described by the means of the covariates, which can be found in appendix I.
Utilitarian anonymous shoppers
The first segment describes shoppers with an utilitarian shopping motivation finds personal information disclosure is the most important attribute (40.85%), followed by intensity of communication (33.65%), then waiting time (17.40%) and finally company representative (8.11%). From the parameters it can be seen that these shoppers really find it important that they do not have to create an account for accessing the live chat, while they do not really mind about entering a name; they would prefer that no information is required for entering the chat. Moreover, this segment really also don’t want to use a video chat for communicating. Therefore, this segment can be described as utilitarian anonymous shoppers, as they do not want to disclose their information while asking for assistance in their shopping journey. The average age of these shoppers is 30 years old and this segment is the biggest, as 36 people fit into this segment. This segment can further be described by the fact that the vast majority exists has a university education or higher, namely 62.31%. Moreover, they are average in their frequency of online shopping and are almost just as many males as females (48.83% & 51.17%.
Utilitarian Video chat dislikers
Page | 36 Utilitarian Personal speed seekers
Then, the third utilitarian shopper segment can be described by consumers that really want assistance as soon as possible, as waiting time was the most important attribute in their decision-making (42.37%). Then, the attribute that is most important is company representative (33.94%), followed by personal information disclosure (16.71%) and lastly intensity of communication (6.98%). They prefer to start the chat right away and do not really mind about entering personal information to start the chat or using a video chat to get an immediate response. Furthermore, these people find it important that they know who they talk to on the chat and indicate to want to see a name and picture of the company representative. Therefore, this segment can be described as personal speed seekers. This segment exists mainly of relatively older people, as the average age for this segment is 39. Also, these consumers are infrequent online shoppers and are somewhat lower educated than the other segments, namely 47.5% indicates to have a HBO diploma. Finally, it was found that the majority of the 23 survey respondents are female (58.14%).
Utilitarian Privacy seekers
The final utilitarian segment can be described as consumers that really do not want to disclose personal information to enter the chat (50.42%), even more than segment 2. This is followed by the attribute waiting time (23.56%), company representative (16.14%) and intensity of communication (9.88%). This segment exists of 23 persons, who do not mind if the company representative is visible or not, as long as their personal information can stay private. However, they would not mind about the intensity of the communication, because they rather would be helped quickly. This segment can further be described as averagely aged (33) and quite highly educated. Also, these are mainly consumers that shop online very frequently and are almost 50/50 in terms of gender.
Next, the four hedonic segments will be discussed, which are very similar to the utilitarian segments:
Hedonic anonymous students
Page | 37
These shoppers also really find it important that they do not have to create an account for accessing the live chat, while they do not really mind about entering a name; they would prefer that no information is required for entering the chat. Moreover, this segment really also don’t want to use a video chat for communicating. Therefore, this segment can be described as hedonic anonymous shoppers, as they do not want to disclose their information while asking for assistance in their shopping journey. Furthermore, they do not mind about the identity of the company representative and also do not care so much about waiting time. This is a quite young segment, with an average age of 28 years old. Also, this segment is the biggest, as 57 people fit into this segment. Moreover, the vast majority of this segment has a university education or higher, namely 58.87% and can therefore be characterized as students. These students shop online averagely and are predominantly female (60.16%).
Hedonic Video chat haters
The second hedonic segment finds the intensity of communication the most important attribute (67.86%) and has an even stronger negative attitude towards a video chat than the utilitarian video dislikers, therefore their name is hedonic video chat haters. The attribute importance is followed by waiting time (21.50%), personal information disclosure (8.07%) and finally company representative (3.56%). This segment further prefers to have no waiting time, however a little waiting time is still ok for them. Moreover, they have a negative attitude towards creating an account for accessing the chat. The average age of these shoppers is 34 and are mostly female consumers (69.28%). The frequency of online shopping is quite highly educated.
Hedonic Personal speed seekers
Page | 38
often. Furthermore, their level of education is average and similar as the hedonic video haters. The segment consists of 17 people, with an almost equal amount of males and females.
Hedonic personal interaction doubters
The last hedonic segment is quite small, namely consists of 10 respondents. These respondents find it very important to see who they are communicating with, as the company representative is the most important attribute (47.43%). This is followed by the attribute intensity of communication (38.76%), waiting time (7,38%), and personal information disclosure (6.43%). This segment differs largely from the fourth segment of the utilitarian segments: Utilitarian privacy seekers. As this segment wants to know who they are communicating with, does not mind about entering personal information but does not particularly want to use a video chat to communicate, this segment will be described as hedonic personal interaction doubters. This segment does not mind about waiting time and can be further be described as above average aged (35) and averagely educated. Also, these are mainly consumers that shop online very frequently and are surprisingly mostly men.
When comparing these segments, it can be seen that similar groups of people exist. However, what is noticeable is that for utilitarian shoppers waiting time is an important attribute for all segments, while for the hedonic shoppers, waiting time is not important for the majority of the consumers (segment 1 and 4 combined are 67 respondents).
4.7 Hypotheses overview
Now that attribute importance is measured, interactions are tested and segments are created. An overview will be created to show which hypotheses were accepted and which were rejected. The hypotheses are displayed in table 14 below, where U stands for utilitarian and H stands for hedonic.
Hypotheses
H1: A higher intensity of communication in a live chat conversation has a positive influence on the chat preferences
H2: Low waiting time in a live chat has a positive effect on the chat preferences
H3: High identity exposure of the user of the live chat has a negative effect on the chat preferences.
H4: High identity exposure of the company representative in a live chat has a positive effect on the chat preferences
H5a Utilitarian shopping motivations will increase the negative effect of waiting time on consumer preferences
Page | 39 Conclusions Aggregate U- model Aggregate H- model U- segment 1 U-segment 2 U-segment 3 U-segment 4 H-segment 1 H-segment 2 H-segment 3 H-segment 4 H1 Reject Reject Reject Reject Accept Reject Reject Reject Reject Reject
H2 Accept Accept Accept Accept Accept Accept Accept Accept Accept Reject
H3 Accept Accept Accept Accept Reject Accept Accept Accept Accept Reject
H4 Accept Accept Accept Accept Accept Reject Accept Reject Accept Accept H5a Accept Accept
H5b Accept Accept
Table 14 Hypotheses overview
Page | 40
5. Discussion & Conclusion
This report will be finalized by concluding previously discussed results, first by describing some general findings of the research, then by discussing the attributes and their importance, followed by discussing the role of the moderator and finally mentioning the segments that were created.
First of all, it was found that 70% of the respondents of the surveys have not used a live chat before in their online shopping process. However, most people indicate to have a quite positive attitude towards a live chat, namely by giving an average rating of 3.66 on a scale from 1-6. Respondents were asked to fill in a set of choice sets of live chats to indicate what chat they would prefer best, afterwards they were asked if they would actually use the chat. By the use of a cross tabulation in SPSS, it was found that for the consumers with the utilitarian shopping motivations, 66.6% of the cases indicated that they would use the chat of their preference. For the hedonic shoppers, this percentage is higher, namely 75,5%. This implies that in general shoppers with hedonic motivations are more likely to make use of the customer support chat in their shopping process. Then, for both studies it was researched how consumers felt about the attribute that effect the live chat use, no interactions between the attributes were found to have an influence on the chat’s preferences.
Identity disclosure company representative
This attribute was found to has the smallest influence on consumers preferences for a live chat. For utilitarian shoppers the attribute importance is 10%, while for hedonic shoppers this is 11%. The direction of the effect for both surveys shows that consumers prefer to see a name & picture of the company representative, only a visible name is not much better for them than no information.
Identity disclosure user
Page | 41
Waiting time
The effect of waiting time on the live chat preferences was found to be almost linear for both shopping motivations. Waiting time is an important determinant for consumers in whether they would prefer using the chat or not, with an attribute importance of 29.8% for utilitarian shoppers and 23% for hedonic shoppers. The hedonic have a slightly positive attitude towards ‘2nd
person in line’, indicating that they would allow for some waiting time to enter the chat. However, just as the utilitarian shoppers they have a very negative preference for long waiting time and a strong positive preference for no waiting time.
Intensity of communication
The research indicated that the intensity of communication was the most important factor in consumer’s decision making, namely because for the utilitarian model the attribute importance is 32.2 % and for the hedonic model even 39%. The parameters show that consumers have a strong negative attitude towards the video chat, and mostly prefer using a text chat only. Only the segment utilitarian personal speed seekers of 23 people have positive attitude towards a video chat, which is about 10% of all respondents. They also have a somewhat positive attitude towards a text chat with the possibility of a video, however most people would most likely not make use of this.
Segmenting customers
Page | 42
6. Managerial implications
This paper contributes to the academic literature about online customer support; however it may also be particularly useful for managers. As the research is quite practically oriented, managers of e-commerce websites can learn more about consumers’ needs and want and take advice on how and if they should incorporate a live chat function on their website. Therefore, this paper will investigate the consumer preferences for a live chat conversation, so that companies can design their chat in the most successful way. It was found that consumers have a very positive attitude towards the possibility of using a live chat as a customer support tool; therefore it can be advised to managers to incorporate a chat on their e-commerce website. As a live chat is seen as a successful way to assist consumers and engage better with customers. Also, from previous literature it was found that effective help via live chat can decrease the likelihood of product basket abandonment. Moreover, incorporating a live chat can lead to higher customer satisfaction about the website and then increase sales indirectly. However, it is important that it is done properly; otherwise people will not use it.
First of all, the chat needs to be available as often as possible, managers should investigate at what time per day most shoppers are active, so that then the chat can be provided. This is important as most consumers find it important that there is no or little waiting time for using the chat. Thus, more than enough company representatives should be available for helping the consumers right away. Also, the chat should be accessible without too much personal information needed, most consumers are fine with entering a name but more information than that is not appreciated. Then, from the research is found that most people have a very negative attitude towards video chatting with a company representative, thus this does not have to be provided in a live chat. Finally, it was found that many people do not really mind about seeing who they are communicating with, however it is always more positive if the identity of the company representative is shown.
