Thesis MSc Marketing Intelligence

(1)

Thesis MSc Marketing Intelligence

Customer support channels: exploring drivers and the

relationship with customers’ satisfaction and loyalty

University of Groningen Faculty of Economics and Business

Department of Marketing PO Box 800 9700 AV Groningen (NL)

January 2020

Elske Willemijn van Randwijk S2667827

Wagenaarstraat 361 1093 CN Amsterdam

+31 6 51 69 93 39 elske.wvr@hotmail.com

First supervisor: dr. H. (Hans) Risselada H.Risselada@rug.nl

(2)

(3)

Abstract

Over the years, customers increasingly use new channels in all phases of the customer journey: in the pre-purchase phase, purchase phase but also in the post-purchase phase. In a time where various online support channels emerge and customers’ voice is increasing, companies realize that customer support is essential to assist their customers and differentiate from competitors. As the natural owner of a large part of the customer journey, customer support can provide invaluable insights. Not much research has appeared yet about the drivers of channel selection in the post-purchase stage, although a lot is changing within customer support and the channels through which this is offered. This study tried to fill this gap by examining task complexity and several personal characteristics as possible drivers for channel selection in customer support context. A multinomial logistic regression model is composed in order to measure these drivers on the different channels. The results show a significant and positive effect of age and

household with children on the selection of chat as customer support channel, wherefore firms

are recommended to focus on this segment while offering customer support through chat as one its main channels. Furthermore, service encounters determine the customers’ overall satisfaction and willingness to continue the relationship. To avoid churn, customer support can provide the company with a second change to meet customers’ expectations. Next to loyalty to a firm, a customer can also experience loyalty to a specific support channel. A mediation analysis is conducted to measure the presence of customer satisfaction as a mediator between

channel experience and loyalty, which shows significant and positive results for both firm loyalty and channel loyalty. If a firm clearly understands which channels are preferred by

customers and what drives these preferences, together with an understanding of how the service encounters thereafter contribute to the relationship with their customers, a firm can provide a better imbedded and satisfactory strategy among its customer support channels.

Keywords: customer support, channel selection, channel experience, customer satisfaction,

(4)

(5)

1. Introduction

The market environment is changing rapidly, which results in a fast-changing customer journey (Edelman & Singer, 2015; Rawson et al., 2013). Nowadays, companies invest substantially more in marketing touchpoints, both offline and online (Kannan et al., 2016). New opportunities for business research and practice have come up with recent developed technologies (Pauwels & Neslin, 2015). Firms have widely integrated new channels within their business, to change the way their marketers interact with their customers and to try to access competitive advantages (Venkatesan et al., 2007). This emerges one central question: “How can firms manage customer interactions and touchpoints across all channels in an integrated manner to provide a superior customer experience and gain a competitive edge?” (Melero et al., 2016, p. 21).

According to Cao & Li (2015), the development of new marketing channels created new ways for customers to get in contact with firms. These new channels make customers purchase products, but they also offer the opportunity to contact the firm for receiving information or for seeking technical advice (Neslin et al., 2006). Customers increasingly use new online channels in all phases of the customer journey: the pre-purchase phase, purchase phase but also in the post-purchase phase (Neslin & Shankar, 2009).

(8)

from the fact that the quality of customer support fundamentally affects customers’ perception of a brand (Alton, 2019), as well as that “thanks to customer care’s responsibilities, its frontline agents enable the function to hear the ‘the voice of the customer’ on a daily basis and can be used to monitor trends and overall sentiment, identify pain points, improvement levers, and success factors of the customer journey” (Lotz et al., 2018, Customer care’s vital role in the transformation section, para. 2). Here, customer care can be seen as an equivalent of customer support.

Not much literature has appeared yet about channel selection in the post-purchase stage, as prior literature mainly focusses on the pre-purchase and purchase stage of the customer journey. Prior literature includes several variables as possible drivers of channel selection in the pre-purchase and purchase stage, among which task complexity and personal characteristics. These variables are used in this study to examine whether they also drive the channel selection in a customer support context. Given the importance of understanding preferred channels and possible drivers of these preferences within customer support context, the first research question is formulated as follows:

What drives customers’ channel selection in the customer support context?

(9)

imbedded strategy among its customer support channels (Lotz et al., 2018). For this reason, a second research question is included in this study and is formulated as follows:

What is the relation between channel experience within customer support and loyalty to both firm and channel, and how is this mediated by customer satisfaction?

(10)

2. Theoretical Framework

This section starts with an explanation of the customer journey and the customer support stage specifically. Thereafter, one can find two parts of the theoretical framework: part A provides theoretical background for the first research question, part B for the second research question. In part A, firstly the channels within customer support context are explained in more detail. Then, possible drivers of channel selection in the customer support context are explained, containing task complexity and personal characteristics. Part B starts with an explanation of channel experience, followed by its relationship with both loyalty to a firm and loyalty to a channel. Finally, the mediating role of customer satisfaction is explained. The conceptual model at the end of the section graphically explains the relationships of both parts A and B.

2.1 The Customer Journey and Customer Support

The customer journey consists of three stages: the pre-purchase stage, purchase stage, and post-purchase stage (Lemon & Verhoef, 2016). The pre-post-purchase consists of all the customer interactions with the brand or firm before a purchase transaction is made (Lemon & Verhoef, 2016). To this stage belong need recognition, search, and consideration (Lemon & Verhoef, 2016). The purchase stage consists of all the customer interactions with the brand or firm during the actual purchase event (Lemon & Verhoef, 2016). To this stage belong the making of a choice, ordering, and payment (Lemon & Verhoef, 2016). The post-purchase stage consists off all the customer interactions with the brand or firm after the purchase of a product or service (Lemon & Verhoef, 2016). To this stage belong the usage of the product or consumption, post-purchase engagement, and support requests (Lemon & Verhoef, 2016).

(11)

helping to define journeys, identify pain points, and spur collaboration across functions” (Lotz et al., 2018, Introduction section, para. 3).

Next to that, “for more and more organizations, customer care is playing an essential and natural role in mapping the customer journey” (Lotz et al., 2018, Customer care’s vital role in the transformation section, para. 1). When customer support started, it was only provided through call centres (Lotz et al., 2018). However, nowadays, “customer care is increasingly becoming a major contributor to customer satisfaction across a broad range of customer channels, from chat and social media to service apps and self-service channels” (Lotz et al., 2018, Customer care’s vital role in the transformation section, para. 1).

Not all organizations use customer support as a strategic unit (Lotz et al., 2018). However, “customer care should be closely involved in designing customer journeys: the function encompasses all touchpoints and organizational units with a clear service component, such as branches, field service, and contact centres that handle calls, e-mails, chats and back-office tasks” (Lotz et al., 2018, Customer care’s vital role in the transformation section, para. 2). Furthermore, “thanks to customer care’s responsibilities, its frontline agents enable the function to hear the ‘voice of the customer’ on a daily basis, monitor trends and overall sentiment, and identify pain points, improvement levers, and success factors” (Lotz et al., 2018, Customer care’s vital role in the transformation section, para. 2).

2.2 Part A

2.2.1 Channel Selection in Customer Support

Because of the developments of new online support channels, the importance of customer support through multiple channels has increased (Kumar, 2010). For the management of customer relationships across channels, it is very important to understand the consequences of customer behaviour regarding the different channels (Thomas & Sullivan, 2005). This holds especially because firms experience increasing pressure to demonstrate the contribution of marketing investments to the profitability and growth of the firm (Kumar & Shah, 2009).

(12)

customers’ expectations of being able to communicate online (Greenberg, 2010), wherefore many service providers are now using new technologies to provide customer support online (Truel & Connelly, 2013). However, the different support channels may differ in terms of human contact, search and switching costs and support level (Neslin & Shankar, 2009). One of the biggest differences between online and offline ways of providing customer support is the element of social interaction (Nass & Moon, 2000). With the usage of offline channels, the customers and the firm interact on a face-to-face manner, which may increase the feeling of a relationship with support staff and thus with the firm (Macintosh & Lockshin, 1997). The next section explores possible drivers of channel selection within customer support context.

2.2.2 Drivers of Channel Selection in Customer Support

As mentioned earlier, if a firm clearly understands which channels are preferred by customers and what drives these preferences, this may increase customers’ satisfaction across the preferred channels (Neslin et al., 2006). Prior literature, which particularly focuses on the search and purchase phase of the customer journey, determined several drivers of channel selection such as channel attributes, task complexity, marketing activities, social influence and

personal characteristics (Melero et al., 2016; Neslin et al., 2006). Task complexity and personal characteristics are the drivers of which the impact on channel selection in the

customer support context will be explored in this study, as these factors seem most appropriate to possibly drive channel selection in customer service context.

2.2.2.1 Task Complexity

(13)

support channels because they are more personal (Pieterson & Van Dijk, 2007). When less complex problems appear or when background information is already available, the preference for online support channel usage increases (Pieterson, 2010).

Based on this literature, the following hypothesis is formulated:

H1. Task complexity has a negative effect on the probability that an online support channel is preferred.

2.2.2.2 Personal Characteristics

According to Hemmer (2012) and Pieterson (2009), channel choice is also affected by personal characteristics. Bilgicer et al. (2015) also state that customer demographic characteristics explain difference in adoption of various channels. Other prior literature states that demographics such as gender, age, education and household influences channel choice (Ansari et al., 2008; Gupta et al., 2004; Inman et al., 2004; Kushwaha & Shankar, 2005; Verhoef et al., 2005). As online channels contain more technological innovations than offline channels, specific demographics may determine their adoption. Past research has shown that older people are less likely to adopt new technologies which could be explained by the fact that older people have a less positive attitude towards new technologies than younger people (Brickfield, 1984). Furthermore, prior research shows a negative relationship between age and the acceptance of new technologies (Trocchia & Janda, 2000; Karjaluoto et al., 2002; Lee et al., 2002; Zhang, 2005). This can be explained by the fact that older consumers have lower cognitive capabilities to learn to work with these new technologies (Hertzog & Hultsch, 2000). Finally, Ansari et al. (2008) state that older people are less likely to use online channels.

(14)

Other reviews of literature show that online consumers are better educated (Bellman et al., 1999; Li et al. 1999; Swinwyard & Smith, 2003). In the study of Ansari et al. (2008), they show the difference between a migration group towards online channels and a non-migration group. This study also indicates that customers in the migration group towards the online channels are more likely to have children than customers in the non-migration group (Ansari et al., 2008).

Based on this literature, the following hypotheses are formulated:

H2. Age has a negative effect on the probability that an online support channel is preferred. H3. Men are more likely to select online support channels than women.

H4. Education has a positive effect on the probability that an online support channel is preferred.

H5. Household with children has a positive effect on the probability that an online support channel is preferred.

2.3 Part B

2.3.1 Loyalty to Firm and Channel

(15)

However, a customer can not only experience loyalty to a firm, but also loyalty to a specific channel. Channel loyalty is defined as the likelihood that a customer will remain loyal to the previous used channel (Yoshida et al., 2013; Zeithaml et al., 1996). Channel loyalty leads to the usage of the same channels over time (Johnson et al., 2003). When a customer is not loyal to a channel, he or she may not have a strong intrinsic preference for any specific channel, and switches among different channels (Gensler et al., 2007). Independent of a channel’s performance with respect to the channel attributes, the previous use of a channel increases the likelihood of using that channel again in the future, causing consumers to be locked into the specific channel and increasing channel loyalty (Johnson et al., 2003). The next paragraph explains this concept in more detail.

2.3.2 Channel Experience

According to Melero et al. (2016), a customers’ channel choice is determined by a customers’ channel experience. Usually, past customer behaviour predicts future behaviour (Venkatesan & Kumar, 2004). This is also agreed by Boulding et al. (1999) and Thomas et al. (2004), who state that prior experiences affect current choices. Valentini et al. (2011) state that experience leads to learning and habit, which increases the probability that the same channels will be used over time. This is also agreed by Ansari et al. (2008), who state that consumers’ user experience (or prior channel experience) affects the consumers’ future channel choice. Channel experience might thus determine channel loyalty (Johnson et al., 2003). The opposite can also occur, as unsatisfying experiences can also have an impact on future behaviour (Mattila, 2003). An unsatisfying experience may lead to usage of other channels (Mattila, 2003).

(16)

Based on this literature, the following hypothesis is formulated:

H6. Channel experience has a positive effect on both firm and channel loyalty.

2.3.3 Customer Satisfaction

According to Oliver (2014), customer satisfaction can be seen as the emotional response to desired fulfilments. Kotler (2003) states that this emotional response is derived from a product’s perceived performance compared to the expectations of the customer. As Schneider et al. (2008) state, companies make use of customer survey to gather customer feedback data in order to measure and improve customer satisfaction and loyalty. To measure this customer loyalty, single-questions have become a popular metric to measure among customers (Wiesel et al., 2012). One sufficient measurement question is proposed by Reichheld (2003), containing the Net Promoter Score (NPS). The NPS is a popular customer feedback metric that measures the extent to which customers are prepared to recommend the company to friends and family (De Haan et al., 2015). Depending on the score, customers can be categorized as ‘detractors’ (numbers between 0 and 6), ‘passives’ (numbers 7 and 8) or ‘promoters (numbers 9 and 10) (De Haan et al., 2015). According to Schneider et al. (2008), likelihood of recommendation is a better measurement of satisfaction than simple like or dislike questions, as it asks the respondent to his or her future behaviour. Forward-looking customer feedback metrics focus on what customers plan to do in the future and may signal something about the future performance of the relationship (Zeithaml et al., 2006). Reichheld’s (2003) NPS is an example of a forward-looking customer feedback metric because it considers the willingness to recommend a firm in the future, which may also signal one’s future relationship with the firm (Zeithaml et al., 2006). Although the NPS might be a suitable measure for loyalty, it does not explain the root cause or causes of a low score (Zaki et al., 2016). However, De Haan et al. (2015) find that NPS is a useful key metric and an effective predictor of customer retention.

(17)

which is also agreed by Keisidou et al. (2013). It is important for service providers to understand the most important aspects of their service encounters that contribute to customers’ satisfaction, in order to increase loyalty (Lai-Ming Tam, 2012). Next to loyalty to a firm, customer satisfaction is also identified as an important factor for channel loyalty (Yang & Peterson, 2004). Yang & Peterson (2004) state that when a customer is satisfied with a service encounters, he or she will use the same channel in the future. Since existing literature argues the direct effect between service encounters and loyalty to both firm and channel, and the indirect effect through customer satisfaction, it is expected that there is a mediation effect between these variables, but no full mediation.

Based on this literature, the following hypotheses are formulated:

H7: Channel experience has a positive effect on customer satisfaction.

H8: Customer satisfaction has a positive effect on both firm and channel loyalty.

(18)

2.4 Conceptual Model

Based on the literature above, the conceptual model for both research questions is created. Part A consists of the elements for the first research question “What drives customers’ channel

selection in the customer support context?” and part B consists of the elements for the second

research question “What is the relation between channel experience and loyalty to both firm

and channel, and how is this mediated by customer satisfaction?”.

Figure 1: Conceptual Model

Personal characteristics Task complexity Channel selection / channel experience Customer satisfaction Loyalty to both firm and channel Part A

(19)

3. Research Design

In this chapter, the data and techniques used in order to answer the research questions is discussed. The following section explains how data is collected, followed by a description of the choice of the variables. Then, the choice of the techniques is explained. At the end of the section, a plan of analysis is provided.

3.1 Data collection

Answering the research questions can be done by testing the hypotheses and drawing conclusions from these results. For this study, data has been collected through a Qualtrics survey among customers of company A. The questionnaires are digitally filled in by 1199 customers of which 1012 could be used. The next chapter explains this in more detail. The survey is sent out between November 21st and December 2nd 2019 by e-mail to 15.628 customers and has a response rate of 6,5%. Company A receives a monthly differing file with e-mail addresses of 100.000 customers. As this file is changing every month, this is lowering the change that customers receive multiple surveys a year. These monthly files are used for ad hoc research and 16.000 e-mail addresses are used to collect data for this study. Some of these e-mail addresses may not be existent anymore or e-mails may be bounced, wherefore 15.628 customers eventually received the survey. In Appendix A, the customer survey can be found.

3.2 Choice of variables

In order to answer the first research question, the following variables are used in this study.

Task complexity is measured on a ratio scale, where 1=very easy and 5=very difficult.

(20)

or living with housemates. Channel selection within customer support is measured as a categorical variable with the outcome options telephone, store, website, app, chat or e-mail.

In order to answer the second research question, the following variables are used. Channel

experience is measured as a binary variable and can be either positive or negative. A positive

channel experience indicates that the customers’ problem has been fixed with the help of company A’s customer support, where a negative channel experience indicates that the customers’ problem has not been fixed with the help of company A’s customer support.

Customer satisfaction, which is the mediator of this model, is measured with the Net Promoter

Score and is derived from the likelihood of recommendation question in the customer survey on an ratio scale. Finally, loyalty to both firm and channel is measured on a categorical scale and accounts for the fact whether one of the following events take place: the next time the customer has a similar problem or question, he or she either uses the same customer support channel (=1), uses a different channel (=2) or is thinking about switching to another provider (=3). An overview of the variables used in this study in order to answer both research questions is listed in table 1 below.

RQ Variables Value labels

1 Task complexity Continuous; ranging from 1 (=very easy) to 5 (= very difficult)

Gender Male / Female

Age Continuous; ranging from 0 to 99

Education LO, VMBO, MAVO, HAVO/VWO, MBO, HBO or WO

Household

Living alone, married and living without children, married and living with children, single and living with children, living with parents, living with housemates

Channel selection Telephone, store, website, app, chat or e-mail

2 Channel experience Positive channel experience (=1) / Negative channel experience (=0) Customer satisfaction Continuous; ranging from 0 (=not at all likely) to 10 (=very likely) Loyalty to both firm and

channel

1=stay and preference for same channel; 2=stay and preference for other channel; 3=churn intention

Table 1: Variables used in this study

3.3 Choice of technique for Channel Selection

(21)

are either unordered or ordered, but always discrete. This study uses an unordered multinomial model. Because in a multinomial logit model the probabilities sum to 1, a base category has to be assigned (Leeflang et al., 2015). All utilities for the other choice option j are defined relative to n and the utility of alternative n is zero. Within the multinomial logit model, the probability for customer i choosing channel j from a choice set with k alternatives is the result of the exponentiated utility of alternative j, divided by the exponentiated sum of utilities of all the other alternatives k (Eggers & Kraus, 2016). The probabilities can be specified in more detail obtaining the following specification:

𝜋_𝑖𝑗 = exp(𝑋𝑖𝑗

′_𝛽)

1+∑𝑛−1_𝑘=1exp(𝑋_𝑖𝑘′ 𝛽) 3.1

Where j = 1, … , n and k ≠ j

For the first research question, where H1, H2, H3, H4 and H5 are incorporated, the following multinomial logit model is created:

𝜋(𝐶ℎ𝑎𝑛𝑆𝑒𝑙𝑒𝑐𝑡)𝑖𝑗 =

exp(𝛽0𝑗+𝛽1𝑗𝑇𝐶𝑖+𝛽2𝑗𝐴𝐺𝐸𝑖+𝛽3𝑗𝐺𝐸𝑁𝑖+𝛽4𝑗𝐸𝐷𝑈𝑖+𝛽5𝑗𝐻𝐻𝑖)

1+∑𝑛−1_𝑘=1exp(𝛽0𝑘+𝛽1𝑘𝑇𝐶𝑖+𝛽2𝑘𝐴𝐺𝐸𝑖+𝛽3𝑘𝐺𝐸𝑁𝑖+𝛽4𝑘𝐸𝐷𝑈𝑖+𝛽5𝑘𝐻𝐻𝑖) 3.2

Where

𝜋(𝐶ℎ𝑎𝑛𝑆𝑒𝑙𝑒𝑐𝑡)_𝑖𝑗 = probability that customer i prefers channel choice option j;

𝛽0 = intercept;

𝛽₁ = parameter of task complexity variable of either j or the alternatives k; 𝛽₂ = parameter of age variable of either j or the alternatives k;

𝛽₃ = parameter of gender variable of either j or the alternatives k; 𝛽₄ = parameter of education variable of either j or the alternatives k; 𝛽₅ = parameter of household variable of either j or the alternatives k; 𝑇𝐶_𝑖 = task complexity for customer i;

𝐴𝐺𝐸_𝑖 = age of customer i; 𝐺𝐸𝑁_𝑖 = gender of customer i; 𝐸𝐷𝑈𝑖 = education of customer i;

(22)

Estimation and validation of the multinomial logit model is done with odds ratios, which are relative to the base case and are used to interpret the estimated parameters (Leeflang et al., 2015). Multinomial logistic regression is a powerful tool to model choice from a finite set of alternatives, but it suffers from the assumption of Independence of Irrelevant Alternatives (IIA), stating that the odds of choosing one alternative over another is constant regardless of whichever other alternatives are present (Leeflang et al., 2015).

3.4 Mediation effect

The second research question accounts for a mediation effect of customer satisfaction.

To estimate the mediation effect, the mediation model of Hayes (2018) is used. Hayes (2018) states that mediation can be seen as a causal chain where one variable (X) affects a second variable (M) that affects a third variable (Y), where M is the mediator. According to Baron & Kenny (1986), the mediation effect can be derived with three steps: regression the independent variables on the mediator (effect a), testing the relationship of the independent variables on the dependent variable (effect c), and, finally, conducting the relationship between the mediator on the dependent variable (effect b) and again the independent variables on the dependent variable (effect c’).

Graphically, mediation can be depicted in figure 2 below:

Figure 2: Model of Mediation (Hayes, 2018)

The direct effect is explained by the relationship between channel experience and loyalty to

both firm and channel. The mediational path, in which channel experience leads to loyalty to both firm and channel through customer satisfaction, is called the indirect effect.

Channel Experience

Loyalty to both Firm and Channel

Channel Experience

Loyalty to both Firm and Channel

Customer Satisfaction 𝑐′

𝑎 𝑏

(23)

When effect c’ in combination with a and b are significant as well as relation c on its own, this is called partial mediation. If c’ is not significant while a and b are significant, this is called full mediation (Baron & Kenny, 1986).

As in this study the Y is loyalty to both firm and channel and contains three possible outcomes, this needs more attention. Iacobucci (2012) explains that the introduction of categorical variables into mediation is recognized as an important issue. And yet, to date, no single solution has been found (Iacobucci, 2012). For this reason, loyalty to both firm and channel is divided in two different variables. First, the outcomes stay and preference for same support channel and stay and preference for different support channel are merged. Herewith, a binomial variable is created, called firm loyalty from now on, which measures whether a customer wants to stay with company A (0) or intents to churn (1). Next to firm loyalty, a second binomial variable is created, called channel loyalty from now on, and measures whether a customer stays and prefers the same support channel (0) or stays and prefers a different support channel (1). Here, the outcome option churn intention is omitted. This gives the opportunity to follow the steps Iacobucci (2012) explains to check for a mediation effect when Y is a binary categorical variable:

(1) If Y is categorical, a logistic regression model has to be used. This model will estimate the strength of the direct path c.

𝑌̂ = 𝑒𝑥𝑝(𝛽1+𝑐𝑋)

1+𝑒𝑥𝑝(𝛽1+𝑐𝑋) 3.3a

(2) If M is continuous, the following equation has to be fitted via a linear regression model. This model will estimate the strength of path a.

𝑀̂ = 𝛽₂+ 𝑎𝑋 3.4a

Collect the parameter estimate a, and its standard error, 𝑠_𝑎.

(3) If Y is categorical, a logistic regression model has to be used. This model will estimate the strength of bath b as well as c’.

𝑌̂ = 𝑒𝑥𝑝(𝛽3+𝑐′𝑋+𝑏𝑀)

1+𝑒𝑥𝑝(𝛽3+𝑐′𝑋+𝑏𝑀) 3.5a

(24)

(4) Using the parameter estimates a and b and their standard errors, 𝑠_𝑎 and 𝑠_𝑏, the standardized elements can be computed:

𝑧_𝑎 = 𝑎̂/𝑠̂ _𝑎 𝑧_𝑏 = 𝑏̂/𝑠̂ _𝑏

Their product: 𝑧_𝑎∗𝑏 = 𝑧_𝑎𝑧_𝑏

And their collected standard error: √𝑧_𝑎2_{+ 𝑧} 𝑏2+ 1

(5) Finally, the pièce de résistance has to be computed, the z-test that combines results from OLS and logistic regression, to indicate whether there is a significant mediation effect: 𝑧𝑚𝑒𝑑𝑖𝑎𝑡𝑖𝑜𝑛 =

𝑧_𝑎𝑧_𝑏 √𝑧_𝑎2_{+ 𝑧}

𝑏2+ 1

3.5 Choice of technique for Firm Loyalty and Channel Loyalty

To account for the mediation effect for firm loyalty, the following models are created. First, Eq. (3.3b) is created based on Eq. (3.3a) to test for effect c. Second, Eq. (3.4b) is created based on Eq. (3.4a) to test for effect a. Finally, Eq. (3.5b) is created based on Eq. (3.5a) to test for effects c’ and b. Since the dependent variable firm loyalty is a binary variable, a logistic regression model is used for Eqs. (3.3b) and (3.5b). And since the continuous variable customer

satisfaction is the mediator variable, a linear regression model is used for Eq. (3.4b). In these

models H6, H7, H8 and H9 are incorporated:

Effect c: 𝜋(𝐹𝑖𝑟𝑚𝐿𝑜𝑦)_𝑖 = 𝑒𝑥𝑝(𝛽0+𝛽1𝐶𝐸𝑖) 1+𝑒𝑥𝑝(𝛽0+𝛽1𝐶𝐸𝑖) 3.3b Effect a: 𝑠𝑎𝑡_𝑖 = 𝛼 + 𝛽₁𝐶𝐸_𝑖 + 𝜀 3.4b Effects c’ and b: 𝜋(𝐹𝑖𝑟𝑚𝐿𝑜𝑦)𝑖 = 𝑒𝑥𝑝(𝛽0+𝛽1𝐶𝐸𝑖+𝛽2𝑆𝐴𝑇𝑖) 1+𝑒𝑥𝑝(𝛽0+𝛽1𝐶𝐸𝑖+𝛽2𝑆𝐴𝑇𝑖) 3.5b Where

𝜋(𝐹𝑖𝑟𝑚𝐿𝑜𝑦)_𝑖 = probability that customer i stays loyal to the firm; 𝑠𝑎𝑡_𝑖 = satisfaction (NPS score) of customer i;

(25)

𝛽₀ = intercept;

𝛽1 = parameter of channel experience variable; 𝛽₂ = parameter of customer satisfaction variable; 𝐶𝐸_𝑖 = channel experience of customer i

𝑆𝐴𝑇𝑖 = satisfaction (NPS score) for customer i

𝜀 = error term

To account for the mediation effect for channel loyalty, the following models are created. First, Eq. (3.3c) is created based on Eq. (3.3a) to test for effect c. Second, Eq. (3.4c) is created based on Eq. (3.4a) to test for effect a. Finally, Eq. (3.5c) is created based on Eq. (3.5a) to test for effects c’ and b. Since the dependent variable channel loyalty is a binary variable, a logistic regression model is used for Eqs. (3.3c) and (3.5c). And since the continuous variable customer

satisfaction is the mediator variable, a linear regression model is used for Eq. (3.4b). In these

models H6, H7, H8 and H9 are incorporated:

Effect c: 𝜋(𝐶ℎ𝑎𝑛𝑛𝑒𝑙𝐿𝑜𝑦)𝑖 = 𝑒𝑥𝑝(𝛽0+𝛽1𝐶𝐸𝑖) 1+𝑒𝑥𝑝(𝛽0+𝛽1𝐶𝐸𝑖) 3.3c Effect a: 𝑠𝑎𝑡_𝑖 = 𝛼 + 𝛽₁𝐶𝐸_𝑖 + 𝜀 3.4c Effects c’ and b: 𝜋(𝐶ℎ𝑎𝑛𝑛𝑒𝑙𝐿𝑜𝑦)_𝑖 = 𝑒𝑥𝑝(𝛽0+𝛽1𝐶𝐸𝑖+𝛽2𝑆𝐴𝑇𝑖) 1+𝑒𝑥𝑝(𝛽0+𝛽1𝐶𝐸𝑖+𝛽2𝑆𝐴𝑇𝑖) 3.5c Where

𝜋(𝐶ℎ𝑎𝑛𝑛𝑒𝑙𝐿𝑜𝑦)_𝑖 = probability that customer i prefers the same support channel; 𝑠𝑎𝑡_𝑖 = satisfaction (NPS score) of customer i;

𝛼 = constant

𝛽₀ = intercept;

𝛽1 = parameter of channel experience variable; 𝛽₂ = parameter of customer satisfaction variable; 𝐶𝐸_𝑖 = channel experience of customer i

𝑆𝐴𝑇𝑖 = satisfaction (NPS score) for customer i

(26)

3.6 Plan of analysis

(27)

4. Results

In this chapter, first the dataset is described. Thereafter, channel selection, firm loyalty and

channel loyalty are modelled and results are presented. Findings about the hypothesized effects

are provided at the end of the section.

4.1 Descriptives of the dataset

The customer survey is sent out to 15.628 customers, of which the process is explained earlier in the previous chapter. 1488 customers started the survey, however only 1199 of them finished the survey. This gives a response rate of 6,5%. The survey was rather short and took approximately less than 5 minutes, however no reward or direct benefit was included for respondents, which might have been a reason for some respondents not to finish the survey. 1194 of the 1199 where still customers of company A of which 1026 contacted the customer support. 607 of them contacted the customer support in the past 6 months and 448 of them more than 6 months ago. The minimal duration of the survey was 59 seconds and the maximum duration 433.156 seconds. This is rather high and might not be reliable, therefore 14 respondents with an unlikely high duration time (e.g., 275.000 seconds = 76 hours, 337.000 seconds = 93 hours) were deleted from the dataset. Eventually, 1012 respondents were used for further analysis.

Figure 3 shows the boxplots of age for each channel, where one can see that respondents using

chat have the lowest age (average of 48,0 years), followed by app (average of 56,0 years) and website (59,3 years). Respondents using store (64,5 years) and e-mail (average of 62,7 years)

(28)

The average age of the respondents in the dataset is rather high (60,4 years). Next to that, there are a lot more men (73%) than women (27%) in the dataset. However, both the high average age and high number of men are common within the customer database of company A, as older men traditionally are the contract holders for the family of the subscriptions with company A. Furthermore, telephone is by far the most used channel (86,8%). Chat is the second most popular channel (5%), followed by store (3,6%), website (2,7%) and e-mail (1,7%). App is not very popular (0,2%).

Although this seems very unbalanced, as Oommen et al. (2011) state, in probabilistic modelling using maximum-likelihood it is important to develop a sample that has the same class distribution as the original population, instead of assuring that the classes are equally sampled. 78% of the respondents says that their problem is fixed after contacting company A’s customer support, however 22% of the respondents still has a problem which has not been fixed after their contact. This leads to 15,7% of the respondents with a churn intention, 3,8% considering to use another support channel the next time they have a similar problem or question, and the majority of 80,5% prefers the usage of the same support channel again.

4.2 Modelling Channel Selection

A multinomial logit model for the dependent variable channel selection is estimated with all explanatory variables (model 1): containing the independent variables task complexity, age,

gender, education and household. A multinomial logistic model’s estimates can be interpreted

in three ways: the Beta coefficient, the odds ratios and the marginal effects. In this study, both the Beta coefficients and the odds ratios are used. The Beta coefficient is log-transformed and cannot be directly interpreted; only the direction of the variable can be interpreted. A significant negative Beta coefficient means that the variable has a significant negative effect on the probability of choosing that choice option relative to the baseline option (intercept). The odds ratio shows a positive relation of the explanatory variable on the probability of choosing that choice option relative to the baseline option (intercept) if the odds ratio is >1 and a negative relation if the odds ratio is <1.

(29)

Hausman-McFadden (1981) test, the IIA assumption is not rejected (p-value = 0,998), so the multinomial logit model can be used.

When conducting the multinomial logit model, no significant parameters are shown for the explanatory variables task complexity, education and gender. Therefore, a model which includes all five explanatory variables is compared with a model which only includes the significant explanatory variables age and household. As Schermelleh-Engel et al. (2003) state, multiple measurements of fit should be looked at, as all the measurements show different aspects of a model. There is no single statistical significance test that identifies a correct model (Schermelleh-Engel et al., 2003). As can be seen in table 2 below, the second model shows a lower AIC than the first model. However, the first model shows a Log-Likelihood value closer to zero. The R2 is very low for both models, which means that for both models, the predictor variables can explain only a low proportion of the variance in the response variable: only 9,2% in the first model and only 5,2% in the second model. The R2 of the first model is slightly higher, which makes sense due to the fact that more explanatory variables are included. Since the R2 for the first model is the highest of the two and because the first model shows a Log-Likelihood value closer to zero than the second model, model 1 is chosen to continue with for further interpretation of the parameter estimates.

Model Log-Likelihood R2 AIC

Model 1 -525,970 0,092 1211,930

Model 2 (without task complexity, gender and education) -549,410 0,052 1168,810

Signif. codes: 0 ‘***’ 0,001 ‘**’ 0,05 ‘.’ 0,1 ‘’ 1

Table 2: Model fit information

The different categories for education are adjusted to new categories from low to high (LO,

LBO, VMBO, MAVO are categorized together as “1”, MBO as “2”, HAVO and VWO as “3”, HBO and WO as “4”), however this did not change the significance of any of the estimates.

(30)

Coefficient

(baseline = telephone) Estimates P-value Odds

Coefficient

(baseline = telephone) Estimates P-value

Chat : Gender (woman) -0,348 0,350

Chat : Household

(married, without children) 0,620 0,197

App : Gender (woman) 0,538 0,709

App : Household

(married, without children) -16,300 0,995 Website : Gender (woman) -0,647 0,212

Website : Household

(married, without children) -0,571 0,269 E-mail : Gender (woman) -0,680 0,301

E-mail : Household

(married, without children) 0,561 0,401

Store : Gender (woman) -0,235 0,558

Store : Household

(married, without children) -0,126 0,754

Chat : Age -0,041 0,000 0,960

Chat : Household

(living with parents) -19,310 0,999

App : Age 0,002 0,977

App : Household

Website : Age 0,005 0,777

Website : Household

E-mail : Age 0,029 0,216

E-mail : Household

Store : Age 0,015 0,296

Store : Household

Chat : Education 0,121 0,366

Chat : Household

(living with housemates) 0,432 0,698

App : Education -0,046 0,938

App : Household

(living with housemates) -16,040 0,999

Website : Education 0,234 0,159

Website : Household

E-mail : Education -0,171 0,389

E-mail : Household

(living with housemates) -16,540 0,999

Store : Education -0,051 0,713

Store : Household

Chat : Household

(married, with children) 1,033 0,036 2,808 Chat : Taskcomplexity -0,067 0,519 App : Household

(married, with children) 0,510 0,753 App : Taskcomplexity -0,447 0,271

Website : Household

(married, with children) 0,315 0,583 Website : Taskcomplexity 0,015 0,097

E-mail : Household

(married, with children) 0,803 0,376 E-mail : Taskcomplexity 1,502 0,316

Store : Household

(married, with children) -0,682 0,328 Store : Taskcomplexity -1,585 0,242

Signif. codes: 0 ‘***’ 0,001 ‘**’ 0,05 ‘.’ 0,1 ‘’ 1

Table 3: Results of the multinomial logit model

As table 3 shows, the multinomial logit model for channel selection does not show a significant effect for any of the channels for the explanatory variables gender, education and task

complexity.

For the explanatory variable age, the model with reference level ‘telephone’ shows a significant effect for the channel chat (p-value = 0,000). With ‘telephone’ as the reference level, age shows a significant negative effect for chat relative to telephone with an estimate of -0,041 and factor 0,960. This means that the odds of choosing chat over (the base case) telephone changes by a factor 0,960 and thus decreases when age increases with one year.

(31)

0,036). With ‘telephone’ as the reference level, household category married and living with

children shows a significant positive effect for chat relative to telephone with an estimate of

1,033 and factor 2,808. This means that the odds of choosing chat over (the base case)

telephone changes by a factor 2,808 and thus increases when household category married and living with children increases with one unit relative to the baseline category household category living alone.

4.3 Modelling Mediation effect of Customer Satisfaction on

Firm Loyalty and Channel Loyalty

As described in chapter 3, for both binary dependent variables firm loyalty and channel loyalty, three different models will be estimated. The first logistic regression model (3.4) estimates effect c, the linear regression model (3.5) estimates effect a and the second logistic regression model (3.6) estimates effects b as well as c’. The results for the different effects for the first binary dependent variable firm loyalty can be found in figure 4 and table 4 below.

Figure 4: Mediation model of Firm Loyalty

Effect Beta Std Error Z/T-value P-value

c 2,372 0,192 12,348 0,000

a 3,021 0,111 -27,230 0,000

b 0,584 0,050 -11,588 0,000

c' 1,122 0,239 4,687 0,000

Table 4: Results mediation analysis for Firm Loyalty

Given this output, one can follow the steps provided by Iacobucci (2012) which are explained Channel

Experience Firm Loyalty

(32)

whether a mediation effect is present (Iacobucci, 2012). One can see that 𝑧_𝑎 = 𝑎̂/𝑠̂ = 27,230 _𝑎 and 𝑧_𝑏 = 𝑏̂/𝑠̂ = 11,588, wherefore their product 𝑧_𝑏 _𝑎𝑧_𝑏= 315,510 and √𝑧_𝑎2_{+ 𝑧}

𝑏2+ 1= 29,610. By calculating the 𝑧_{𝑚𝑒𝑑𝑖𝑎𝑡𝑖𝑜𝑛} = 𝑧𝑎𝑧𝑏

√𝑧𝑎2+𝑧𝑏2+1

, one can find that this number is 10,656. As 10,656

exceeds |1,96|, one can conclude that this is significant at the 𝛼 = 0,05 level (Iacobucci, 2012). Besides that, as c’ is significant, one can conclude that a partial mediation effect is present. Knowing this, the parameter estimates can now be interpreted. The estimate for effect c is 2,372 which means that a positive relationship is found between channel experience and firm loyalty. For every one unit increase in channel experience (in other words: when a positive channel experience becomes more likely), the log odds of firm loyalty increases by 2,372. The estimate for effect a is 3,021 which means that a positive relationship is found between channel

experience and customer satisfaction. When channel experience increases with one unit, the

customer satisfaction increases with 3,021. The estimate for effect b is 0,584 which means that a positive relationship between customer satisfaction and firm loyalty is found. For every one unit increase in customer satisfaction, the log odds of firm loyalty increases by 0,584. The estimate for effect c’ is 1,122 which means that a positive relationship is found between channel

experience and firm loyalty when the mediating effect customer satisfaction is present. For

every one unit change in channel experience when customer satisfaction is included in the model, the log odds of firm loyalty increases by 1,122.

Figure 5 and table 5 below show the results for the different effects for the second binary dependent variable channel loyalty.

Figure 5: Mediation model of Channel Loyalty Channel

Experience Channel Loyalty

(33)

Effect Beta Std Error Z/T-value P-value

c 2,903 0,344 8,445 0,000

a 1,725 0,107 -16,190 0,000

b 0,321 0,080 -4,013 0,000

c' 2,414 0,362 6,665 0,000

Table 5: Results mediation analysis for Channel Loyalty

Here, the same holds as for the previous mediation test. Given this output, one can follow the steps provided by Iacobucci (2012) which are explained in chapter 3. With these steps, the 𝑧_{𝑚𝑒𝑑𝑖𝑎𝑡𝑖𝑜𝑛} can be found, which can be used to conclude whether a mediation effect is present (Iacobucci, 2012). One can see that 𝑧_𝑎 = 𝑎̂/𝑠̂ = 16,190 and 𝑧_𝑎 _𝑏 = 𝑏̂/𝑠̂ = 4,013, wherefore _𝑏 their product 𝑧_𝑎𝑧_𝑏= 64,970 and √𝑧_𝑎2_{+ 𝑧}

𝑏2 + 1= 16,710. By calculating the 𝑧𝑚𝑒𝑑𝑖𝑎𝑡𝑖𝑜𝑛 = 𝑧𝑎𝑧𝑏

√𝑧𝑎2+𝑧𝑏2+1

, one can find that this number is 3,889. As 3,889 exceeds |1,96|, one can conclude

that this is significant at the 𝛼 = 0,05 level (Iacobucci, 2012). Besides that, as c’ is significant, one can conclude that a partial mediation effect is present. Knowing this, the parameter estimates can now be interpreted. The estimate for effect c is 2,903 which means that a positive relationship is found between channel experience and channel loyalty. For every one unit increase in channel experience (in other words: when a positive channel experience becomes more likely), the log odds of channel loyalty increases with 2,903. The estimate for effect a is 1,725 which means that a positive relationship is found between channel experience and

customer satisfaction. When channel experience increases with one, the customer satisfaction

increases with 1,725. The estimate for effect b is 0,321 which means that a positive relationship between customer satisfaction and channel loyalty is found. For every one unit increases in customer satisfaction, the log odds of channel loyalty increases with 0,321. The estimate for effect c’ is 2,414 which means that a positive relationship is found between channel experience and channel loyalty when the mediating effect customer satisfaction is present. For every one unit change in channel experience when customer satisfaction is included in the model, the log odds of firm loyalty increases by 2,414.

As figure 4 clearly shows, the customer satisfaction differs for every type of loyalty. The preference for same support channel shows an average customer satisfaction of 8. The preference for other support channel shows an average customer satisfaction of 6, and the

(34)

Figure 6: Boxplot of NPS for each type of Loyalty

4.4 Hypothesis testing

(35)

on customer satisfaction. This is supported with 𝛽 = 3,021 and p = 0,000 for the firm loyalty model and 𝛽 = 1,725 and p = 0,000 for the channel loyalty model. For H8 to be accepted,

customer satisfaction should have a significant and positive effect on both firm loyalty and channel loyalty. This is supported with 𝛽 = 0,584 and p = 0,000 for the firm loyalty model and 𝛽 = 0,321 and p = 0,000 for the channel loyalty model. H9 is supported when there is a partial mediation effect of customer satisfaction on the relationship between channel experience and both firm loyalty and channel loyalty. This is supported with 𝑧𝑚𝑒𝑑𝑖𝑎𝑡𝑖𝑜𝑛 = 10,656 for firm

loyalty and 𝑧_{𝑚𝑒𝑑𝑖𝑎𝑡𝑖𝑜𝑛} = 3,889 for channel loyalty. An overview of the hypotheses and their support status to this research is presented in table 6 below.

Hypothesis Description Expectation Finding

1

Task complexity has a negative effect on the probability that

an online support channel is preferred. - Not supported 2

Age has a negative effect on the probability that an online

support channel is preferred. + Partially supported

3

Men are more likely to use online support channels than

women. - Not supported

4

Education has a positive effect on the probability that an

online support channel is preferred. + Not supported

5

Household with children has a positive effect on the

probability that an online support channel is preferred. + Partially supported 6

Channel experience has a positive effect on both firm and

channel loyalty. + Supported

7

Channel experience has a positive effect on customer

satisfaction. + Supported

8

Customer satisfaction has a positive effect on both firm and

channel loyalty. + Supported

9

Customer satisfaction partially mediates the positive effect

of channel experience on loyalty. + Supported

(36)

5. Discussion

The aim of this study was to clarify which drivers influence channel choice within customer support context, together with the relation between channel experience within customer support and customer loyalty, both to the firm and the channel, mediated by customer satisfaction. This final section discusses the results of the hypotheses and the research questions in more detail. When hypothesis are not supported, possible explanations for the non-significant results are provided. At the end of this section, managerial implications are provided together with limitations of this study and suggestions for further research.

5.1 Conclusions

Prior research has discussed drivers of channel selection in the pre-purchase and the purchase stage. However, not much research has appeared yet about the drivers of channel selection in the post-purchase stage, although a lot is changing within customer support and the channels through which this is offered. This study tried to fill this gap by examining possible drivers for channel selection in the post-purchase stage. Next to that, the relationship between channel experience within customer support and customer loyalty to both the firm and the channel is examined, together with customer satisfaction as a possible mediating effect.

The first research question is formulated as follows:

What drives customers’ channel selection in the customer support context?

(37)

between parental uses of Internet and their children’s use of Internet. This might be an explanation for households living with children preferring chat as a customer support channel. Besides that, the study of Harvey & Mukhopadhyay (2006) delineates the time deficit of families with children. This may be another reason for these customers to use chat as a customer support channel, as chat delivers high contact speed (Pieterson & Van Dijk, 2007) and is a quick and effortless channel (Chaparro-Peláez et al., 2016).

Next to above mentioned personal characteristics, no significant results are found for gender,

education and task complexity as drivers for channel selection.

Although prior literature states that men complete purchases through online channels more frequently than women (Burkolter & Kluge, 2011), non-significant difference between men and woman about the level of attitude towards the Internet have also been reported in the past (Havelka, 2003; Howard & Smith, 1986; Igbaria, 1993). Furthermore, research by e.g. Atan (2002), Morris et al. (2005) and Venkatesh & Morris (2000) show that women are equally likely to use information technologies when compared with their male counterparts. Furthermore, Venkatesh & Agarwal (2006) expect the weights of technology adoption to be comparable across men and women, andBeilock & Dimitrova (2003) state that more and more women have joined the online community. Therefore, this may be an explanation for the fact that no significant results are found in this study for men being more likely than women to use online support channels.

(38)

society (Beilock & Dimitrova, 2003), usage of online channels might have been spread among all customers with the same amount, regardless their education. This may explain the non-significant results of education as driver of online channel selection within the context of customer support.

Task complexity was hypothesized as having a negative effect on the probability that an online

support channel is preferred. However, no significant results are found. Prior literature states that more personal channels are preferred by people when they experience complex problems (Pieterson & Van Dijk, 2017). However, it might be that the respondents do not perceive the offline support channels as more personal than the online support channels, as personalization within online support channels is increasing (Riemer & Totz, 2003). For example, chat, one of the online support channels, is seen as a channel where personalization is perfectly possible due to the social interaction (Elmorshidy, 2013). Also, Byström & Järverlin (1995) state that the need for problem solving information increases when task complexity increases, and online support channels are often associated with a quick, costless and effortless way of gathering information (Chaparro-Peláez et al., 2016). This may be an explanation for task complexity not having a significant negative effect on the probability that customer choose online support channels.

The second research question is formulated as follows:

What is the relation between channel experience within customer support and loyalty to both firm and channel, and how is this mediated by customer satisfaction?

In this study, the dependent variable loyalty is split into firm loyalty and channel loyalty. The results show a partial mediation effect of customer satisfaction on the relationship between

channel experience and firm loyalty. A positive relationship is present between channel

experience and firm loyalty, meaning that a positive channel experience results in higher firm loyalty and a negative channel experience in lower firm loyalty, and this relationship is further explained by the mediating variable of customer satisfaction. The results also show a partial mediation effect of customer satisfaction on the relationship between channel experience and

channel loyalty. A positive relationship is present between channel experience and channel

(39)

explained by the mediating variable of customer satisfaction. This partial mediation effect corresponds with prior literature, as positive experiences within customer support increases customers’ satisfaction and loyalty (Caruana, 2002; Setó-Pamies, 2012; Parasuraman et al., 2005), to both the channel (Ansari et al., 2008) and the firm (Mort, 2019). Besides that, a high customer satisfaction results in increased loyalty (Fornell, 1992; Anton, 1996), again, to both the channel (Yang & Peterson, 2004) and the firm (Lai-Ming Tam, 2012). As shown in figure 4, the customer satisfaction differs for every type of loyalty. The preference for the same support channel shows the highest customer satisfaction with an average of 8. Churn intention shows the lowest customer satisfaction with an average of 4 and the preference for another support channel shows a customer satisfaction with an average of 6. Altogether, customer

satisfaction explains the relationship between channel experience and both firm loyalty and channel loyalty.

5.2 Managerial Implications

From the perspective of a firm, the channel through which customer support is distributed can have a strong impact on the perceived quality of this support by the customer (Goffin, 1999). For this reason, the distribution channel of customer support is of high importance (Goffin, 1999).

Customer support is a field that modernizes rapidly under the influence of new technology (Clements, 2019), and within the online channels the interest in chat as customer support service has grown significantly in recent years (Elmorshidy, 2013). Chat becomes a more popular and natural form of contact between companies and customers (Toister, 2019). Today, customers want their questions or problems to be solved instantly on the spot, instead of waiting to receive a reply (Elmorshidy, 2013). And “this is exactly what customer support via chat is all about” (Elmorshidy, 2013, p. 590).

(40)

information in a quick, costless and effortless way (Chaparro-Peláez et al., 2016). And quick and effective resolutions, that is exactly what customers nowadays expect from service encounters (Zendesk, 2019). Furthermore, chat is a channel through personalization, social interactions and immediate response can be ensured for low cost (Elmorshidy, 2013).

Next to that, delivering a satisfactory channel experience should be a priority for businesses, as this makes customers stay with the same support channel, but even more importantly, with the company. A positive channel experience results in a higher probability of customer loyalty to both the channel and the firm, and this relationship is explained by the customers’ satisfaction. Managers should emphasize the importance of the quality of customer support as it can provide invaluable insights (Lotz et al., 2018). As mentioned earlier, “customer care should be closely involved in designing customer journeys, as no other function handles a broader range of customer touchpoints than customer care” (Lotz et al., 2018, Customer care’s vital role in the transformation section, para. 2). Furthermore, “thanks to customer care’s responsibilities, its frontline agents enable the function to hear ‘the voice of the customer’ on a daily basis, monitor trends and overall sentiment, and identify pain points, improvement levers, and success factors” (Lotz et al., 2018, Customer care’s vital role in the transformation section, para. 2).

5.3 Limitations and Future Research

Next to conclusions and managerial implications, several limitations of this study should be noted. First, this study is exclusively focused on company A. The usage of customer support channels may differ across different industries, and even across different companies within this industry. This points out that future research should be carried out on a broader variety of companies or industries.

Second, omnichannel use of customer support channels is not taken into account. It might be that a customer looked at company A’s website or used chat to find an answer before he or she used the telephone to call company A’s customer support desk. In this study, only the latter customer support channel contact is focused on.

(41)

customer to either use another channel or wait until the next day. There is no distinction made between the moments of contact during the day within this study.

Fourth, the respondents in the customer survey where asked if they had an intention to churn, which is only an indication for churn. This would not specifically mean that these customers are actually going to churn when their subscription with company A ends.

(42)

References

Alton, L. (2019). Customer Service vs Customer Experience: What’s the Difference? Smarter CX, presented by Oracle, viewed on 01-11-2019, <https://smartercx.com/customer-service-vs-customer-experience-whats-the-difference/>

Ansari, A., Mela, C.F., & Neslin, S.A. (2008). Customer channel migration. Journal of

Marketing Research, 45(1), 60-76.

Anton, J. (1996). Customer Relationship Management: Making Hard Decisions with Soft Numbers. Upper Saddle River: Prentice-Hall.

Ascarza, E., Iyengar, R., & Schleicher, M. (2016). The perils of proactive churn prevention using plan recommendations: Evidence from a field experiment. Journal of Marketing

Research, 53(1), 46-60.

Atan, H., Azli, N., Rahman, Z., and Idrus, R. (2002). Computers in distance education: Gender differences in self-perceived computer competencies. Journal of Educational Media, 27(3), 123-135.

Athanassopoulos, A.D. (2000). Customer satisfaction cues to support market segmentation and explain switching behavior. Journal of Business Research, 47(3), 191-207.

Baron, R.M., & Kenny, D.A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of

personality and social psychology, 51(6), 1173.

Beilock, R., & Dimitrova, D. V. (2003). An exploratory model of inter-country Internet diffusion. Telecommunications policy, 27(3-4), 237-252.

Bellman, S., Lohse, G.L., & Johnson, E.J. (1999). Predictors of Online Buying Behavior.

Communications of the ACM, 42(12): 32-38.

Benson A.R., Kumar, R., & Tomkins, A. (2016). On the relevance of irrelevant alternatives. In

Proceedings of the 25th International Conference on World Wide Web, 963-973. International

World Wide Web Conferences Steering Committee.

Bhatnagar, A. and Ghose, S. (2004). Online information search termination patterns across product categories and consumer demographics. Journal of Retailing, 80(3), 221-228.

Bhattacharya, C. (1998). When customers are members: Customer retention in paid membership contexts. Journal of the Academy of Marketing Science, 26(1), 31-44.

Bilgicer, T., Jedidi, K., Lehmann, D., & Neslin, S. (2015). Social Contagion and Customer Adoption of New Sales Channels. Journal of Retailing, 91(2), 254-271.

(43)

Bitner, M.J., & Hubbert, A.R. (1994). Encounter Satisfaction Versus Overall Satisfaction Versus Quality: The Customer’s Voice. In Service Quality: New Directions in Theory and

Practice. Eds. Roland T. Rust and Richard L. Oliver. Thousand Oaks, CA: Sage, 72-94.

Bitner, M.J., & Wang, H.S. (2014). Service encounters in service marketing research.

Handbook of service marketing research, 11, 221.

Bitner, M.J., Booms, B.H., & Tetreault, M.S. (1990). The service encounter: Diagnosing favorable and unfavorable incidents. The Journal of Marketing, 71-84.

Bitner, M.J., Brown, S.W., & Meuter, M.L. (2000). Technology infusion in service encounters.

Journal of the Academy of Marketing Science, 28(1), 138-149.

Bolton, R.N., & Drew, J.H. (1992). Mitigating the effect of service encounters. Marketing

Letters, 3(1), 57-70.

Bolton, R.N., & Lemon, K.N. (1999). A dynamic model of customers’ usage of services: Usage as an antecedent and consequence of satisfaction. Journal of Marketing Research, 36(2), 171-186.

Boulding, W., Kalra, A., & Staelin, R. (1999). The quality double whammy. Marketing

science, 18(4), 463-484.

Brickfield, C. F. (1984). Attitudes and perceptions of older people toward technology. In Aging

and technological advances, (pp. 31-38). Springer, Boston, MA.

Burke, R.R. (2002), Technology and the Customer Interface: What Consumers Want in the Physical and Virtual Store. Journal of the Academy of Marketing Science, 30, 411-32.

Burkolter, D. and Kluge, A. (2011), Online consumer behavior and its relationship with sociodemographics, shopping orientations, need for emotion, and fashion leadership, Journal

of Business and Media Psychology, 2(2), 20-28.

Byström, K., & Järverlin, K. (1995). Task Complexity Affects Information Seeking and Use.

Information Processing and Management, 31(2), 191-213.

Campbell, D. (1988). Task Complexity: A Review and Analysis. The Academy of Management

Review, 13(1), 40- 52.

Cao, L., & Li, L. (2015). The Impact of Cross-Channel Integration on Retailers’ Sales Growth.

Journal of Retailing, 91(2), 198-216.

Caruana, A. (2002). Service loyalty: the effects of service quality and the mediating role of customer satisfaction. European journal of marketing, 36(7/8), 811-828.

Thesis MSc Marketing Intelligence