• No results found

The first dataset tests the effects of the customer feedback metrics and the second dataset tests the effects of the customer journey stages

N/A
N/A
Protected

Academic year: 2021

Share "The first dataset tests the effects of the customer feedback metrics and the second dataset tests the effects of the customer journey stages"

Copied!
59
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

The role of customer

experience in churn prediction

An examination of the effect of customer feedback metrics and the customer journey on churn to better understand the construct churn a nd to improve churn

prediction models.

By:

Sandra van de Beld

March 16, 2017

(2)

The role of customer

experience in churn prediction

An examination of the effect of customer feedback metrics and the customer journey on churn to better understand the construct churn and to improve churn

prediction models.

By:

Sandra van de Beld

March 16, 2017

Student number: S2770709

Oosterhamrikkade 33D, 9713 KA Groningen Tel.: +31 6 14682667

Email: S2770709@student.rug.nl

Master Thesis Marketing Management and Marketing Intelligence

University of Groningen Faculty of Economics and Business

Department of Marketing PO Box 800, 9700 AV Groningen

Fist supervisor: Prof. Dr. J. E. Wieringa Second supervisor: F.T. Beke

(3)

Summary

Purpose: This paper aims to better understand the construct churn and to improve churn prediction models by examining the effect of the customer experience and customer journey on churn within the service sector. Thereby, this paper answers the following question: “Which characteristics of customer experience and the customer journey have an impact on the predictability of churn and which

statistical technique performs best, focussing on the service sector?”

Research design: This paper uses existing data that is collected by a Dutch insurance organisation.

They collected the data in 2014 and 2015 by a contact satisfaction survey. This data is used to create two datasets. The first dataset tests the effects of the customer feedback metrics and the second dataset tests the effects of the customer journey stages. By estimating logistic regressions with different variable sets, the effects of the customer feedback metrics and customer journey stages on churn are examined. In addition, different decision trees are estimated. The performance of the logistic

regression and decision tree are compared with the top decile lift to test which statistical technique is most appropriate.

Conclusion: A model that includes the customer feedback metrics, which account for the customer experience, improves the performance of churn prediction models. CSS and NPS are found to be reliable customer feedback metrics. CSS is positive related to churn, the NPS is negative related to churn. A model that includes the customer journey does not improve the performance of churn prediction models. The service request stage is negative related to churn, so that service recovery after a complaint is seen as a way to reduce churn. A model that includes the control variables, customer feedback metrics, customer journey stages and the interaction between the customer feedback metrics and the customer journey yields the best performance with regard to churn prediction. The interaction term between the customer experience stage and CES is negative related to churn. Finally, the decision tree is the best performing statistical technique to predict churn compared to a logistic regression.

Limitations of this paper: This paper uses existing survey data of a single Dutch insurance

organisation. Therefore, it is not possible to account for the effect of some driver of churn, the sample size of dataset two is relatively small and the generalizability of the results are limited. Besides, this paper examines only the performance of the decision tree and logistic regression. However, other statistical techniques are also appropriate to predict churn.

Managerial implications and research directions: The results will help scholars and managers to better understand the concept churn and improve churn prediction models. These results give managers the insights that will help to reduce churn rates and improve the efficiency of marketing budgets, so losses that accompany churn are limited. In addition, the results provide scholars with clarity about drivers of churn, new perspectives to improve churn prediction and directions for further research.

(4)

Acknowledgement

I am proud to present my master thesis, the final touch of my school career. My school career started at Stenden University in Leeuwarden, I graduated and received by bachelor of commerce. To be able to start the master marketing, I had to do a pre-master. I finished my pre-master and decided to do both variations of the master marketing: Management and Intelligence. Because I was used to a different way of teaching by my bachelor degree, I experienced the pre-master and master as a period of hard work. But, now I am here, presenting my master thesis. After a challenging start of my thesis, I am pleased to present this end result. I am thankful for the people that enabled me to present this result.

First of all, I want to thank my first supervisor Prof. Dr. J.E. Wieringa. Jaap Wieringa ensured that, after the delay I had suffered in the beginning, I still could graduate as soon as possible. Due to the delay, my schedule was different from that of my fellow students, but Jaap Wieringa took the time to provide me feedback. Thank you.

Second, I want to thank the marketing research organisation which I wrote my thesis for. This organisation gave me the opportunity to write my thesis after the delay, this organisation provided the data that was required for the analysis and a workspace to work on my thesis. In addition, I want to thank my tutor of the marketing research organisation for the time, flexibility and proper support during the whole period. Also, I want to thank the insurance organisation from which the data belongs to. The insurance organisation provided the additional information necessary to write my thesis, thank you.

Finally, I want to thank my fellow students that were part of the discussion group, best friend, Leone van der Veen, and brother in law, Richard Koopman for their input, support and feedback during the writing of my thesis.

(5)

Table of content

Summary ... 2

Acknowledgement ... 3

1. Introduction ... 1

2. Theoretical Framework ... 4

2.1. The customer experience ... 5

2.1.1. The effect of the customer experience on churn ... 6

2.1.2. Customer Satisfaction ... 7

2.1.3. Contact Satisfaction Score ... 8

2.1.4. Net Promoter Score ... 9

2.1.5. Customer Effort Score ... 9

2.1.6. Relationship commitment and trust ... 10

2.2. The customer journey ... 11

2.2.1. The customer journey in the service sector ... 12

2.2.2. The effect of the customer journey on churn ... 15

2.2.3. Moderation effect of the customer journey and customer feedback metrics on churn ... 15

2.3. Customer, marketing and- environmental characteristics ... 16

2.3.1. Customer characteristics ... 16

2.3.3. Marketing characteristics ... 18

2.3.4. Environmental characteristics ... 18

2.4. Statistical techniques to predict churn ... 19

2.5. Conceptual framework ... 21

3. Research design ... 22

3.1. Data collection ... 22

3.2. Dataset preparation ... 23

3.3. Data cleaning ... 26

3.4.1. Outliers ... 26

3.4.2. Missing values ... 27

3.4. Descriptive statistics ... 27

3.4.1. Descriptive statistics of dataset one ... 27

3.4.2. Descriptive statistics of dataset two... 28

3.4. Methodology ... 29

3.4.1. Logistic regression... 29

3.4.2. Decision tree ... 32

3.4.3. Performance measurement ... 33

(6)

Chapter 4. Results... 34

4.1. Logistic regressions ... 34

4.1.1. Customer experience ... 35

4.1.2. Customer journey ... 37

4.1.3. Customer journey and customer feedback metrics ... 38

4.1.4. Validation of the logistic regressions ... 39

4.2. Decision trees ... 39

4.3. Summary of the results ... 42

Chapter 5. Discussion ... 43

5.2. Customer experience ... 43

5.2. Customer journey ... 44

5.3. Statistical technique to predict churn ... 46

Chapter 6. Conclusion and recommendation ... 47

6.1. Conclusions ... 47

6.2. Limitations... 48

6.3. Managerial implications ... 48

6.4. Research directions ... 49

Rreferences ... 50

Appendix ... 53

Appendix 1. QUEST decision tree ... 53

(7)

1

1. Introduction

Nowadays, organisations struggle within highly competitive and saturated markets.

Organisations derive revenues from the creation and enhancement of long-term relationships with customers (Coussement, Benoit, & Poel, 2010). But how can organisations sustain long-term

relationships with customers when competition is tough? In these situations, customers are vulnerable to competitive offers and therefore likely to terminate the relationship. Especially within the service sector, due to the intangibility and uncertainty of services (Bond & Stone, 2004; Brockett et al., 2008).

When the competition makes a customer an offer, such as lowest price guarantee, customers are likely to terminate (Bond & Stone, 2004; Brockett et al., 2008). Terminating a relationship is called “churn”

(Risselada, Verhoef, & Bijmolt, 2010). Overcoming churn or retaining existing customers is more profitable than attracting new customers (Reinartz & Kumar, 2002). In general, it is suggested that the costs of attracting new customers can be up to twelve times as high as retaining existing customers (Torkzadeh, Chang, & Hansen, 2006). Predicting churning customers can limit the costs of attracting new customers because marketing resources can be targeted in better ways (Shaffer, Zhang, Shaffer, &

Zhang, 2002). Churn can be predicted by the use of churn prediction models. The performance of such models do highly depend on the chosen set of variables and the type of statistical technique (Neslin et al., 2006; Verbeke, Dejaeger, Martens, Hur, & Baesens, 2012).

To be able to build a model that predicts churn well, the right set of variables should be chosen. To be able to choose the right set of variables, one should understand the construct and drivers of churn. Churn can be understood by managing the entire customer journey and thereby the overall customer experience (Sharma & Patterson, 2000). When an organisation is able to manage the customer experience it can expect enormous rewards: higher Customer Satisfaction, increased revenues, improved employee satisfaction, reduced churn rates (Rawson, Duncan, & Jones, 2013) and long-term customer loyalty (Homburg, Jozi, & Kuehnl, 2015). Losses caused by churn can be limited by using the right marketing strategy (Shaffer et al., 2002).

Customer Experience Management (CEM) is the strategy to manage the customer journey and customer experience and is one of the most promising strategies to meet today’s market challenges, such as high competitive markets in the service sector (Homburg et al., 2015). This is a promising strategy because differentiation is no longer sufficient to survive in high competitive markets (Pine and Gilmore 1998 in Nasution, Sembada, Miliani, Resti, & Prawono, 2014), so that an organisation should strive for unique experiences for customers (Gilmoreb & Pine, 2002), which can be created by CEM.

(8)

2 Customer experiences are created by interactions between touchpoints within the pre-purchase stage, purchase stage and post-purchase stage of the customer journey (Lemon & Verhoef, 2016).

Therefore, designing the customer journey and the associated touchpoints are seen as a valuable tool that improves customer experiences (Zomerdijk & Voss, 2010). Thus the customer experience and the customer journey are interesting concepts to understand and predict churn and thereby limiting concomitant losses.

The above paragraphs suggest that managing the customer experience and designing the customer journey will help to understand churn. When churn is understood, the best set of variables can be chosen, so that more accurate churn prediction models can be build (Verbeke et al., 2012).

Besides the set of variables, the chosen type of statistical technique contributes to the accuracy and performance of the model (Verbeke et al., 2012). To build a good performing churn model, one should understand the positive and negative aspects of the different types of statistical techniques. Because literature does not agree on the best performing type of statistical technique for churn prediction (Risselada et al., 2010; Verbeke et al., 2012), this paper examines which type of statistical technique performs best regarding churn prediction within the service sector. With a good performing churn model, marketing budgets can be deployed at customers that are likely to churn, which will save money, so that organisations become more profitable (Shaffer et al., 2002).

In conclusion, this paper contributes to the improvement of churn models. On the one hand by gaining a better understanding of the construct churn, on the other hand by clarifying the confusion about the best performing type of statistical technique to build a churn model. This paper answers the following research question:

Which characteristics of customer experience and the customer journey have an impact on the predictability of churn and which statistical technique performs best, focussing on the service sector?

This research question is answered by using existing survey data of a Dutch insurance

organisation. Two datasets are created to be able to examine the effect of the customer experience and customer journey. Several logistic regressions with different variable sets are performed. The use of different variable sets obtain insights about the impact of the customer feedback metrics and the customer journey variables on churn prediction. The predictability of churn is assessed by calculating the top decile lift. The logistic regression with the highest top decile lift is the best performing logistic regression model. In addition, different decision trees are performed. The decision tree with the highest top decile lift is the decision tree that performs best. The best performing logistic regression and decision tree are compared. The statistical technique with the highest top decile lift, is the best performing statistical technique.

(9)

3 This research found, that the inclusion of customer feedback metrics improve churn

prediction. CSS and NPS are stable customer feedback metrics. CSS is positive related to churn, NPS is negative related to churn. A model that includes the customer journey does not improve the

performance of churn prediction models. The service request stage is negative related to churn, which suggest service recovery as a way to reduce churn. A model that includes the control variables, customer feedback metrics, customer journey stages and, the interaction between the customer feedback metrics and the customer journey is the logistic regression that is best predicting churn. The interaction term between the customer experience stage and CES is negative related to churn. Finally, the decision tree is better in predicting churn than the logistic regression.

These results are interesting for scholars because it contributes to the understanding of the construct churn and- the improvement of churn prediction models by taking the customer experience into account. The effect of the customer experience on churn has not extensively been examined before, but leads to better churn prediction as found within this paper. Gaining insight in the direction and strength of the relation between the customer feedback metrics and customer journey stages and- churn contributes to the understanding of the construct churn. On the other hand, the findings of this paper clarifies the confusion about the best performing statistical technique for churn prediction.

Finally, this paper suggest some potential research directions for further exploration.

These results are also interesting for managers, because understanding the construct churn leads to better churn prediction, so that negative effects of churn on the profitability of the

organisation (Reinartz & Kumar, 2002; Torkzadeh et al., 2006) can be reduced. The results contribute to the understanding of the construct churn because the results provide insights in the reliability of customer feedback metrics and the direction of the effect of customer feedback metrics on churn. In addition, the results with regard to the customer journey, suggest proper service recovery as a way to reduce churn. These insights help managers to set the focus on the right variables and build better performing churn models. When churn can be better predicted, customers can be targeted more efficiently, so that marketing budgets are optimally used. Better targeting will reduce churn and reduce high costs related to retention and/or- recruiting programs.

The remainder of this paper is structured as followed. Chapter two consists of the theoretical framework. The theoretical framework discusses the literature regarding churn. In addition, the hypotheses are set and the conceptual model is shown. Chapter three describes the research design.

The research design discusses the data that is used for analysis and the method of analysis. Chapter four presents the results. The results are divided across the statistical techniques: logistic regression and decision tree. Chapter five presents the discussion of the results. Finally, chapter six discusses the conclusion and recommendations of this research.

(10)

4

2. Theoretical Framework

Churn has become a major problem for firms in different service sectors such as: publishing, healthcare, financial services, electronics, telecom, internet and insurances (Neslin et al., 2006) and is a popular subject within the marketing literature (Wieringa & Verhoef, 2007). Churn is a major problem because churn has negative effects for an organisation such as: immediate losses of sales, higher acquisition costs, because new customers should be attracted, (Risselada et al., 2010) and potential risk for negative word of mouth when customers are not satisfied (Homburg et al., 2015;

Verbeke et al., 2012). Especially in the service sector, churn is an issue because these services are often sold on a contractual basis, so that a churned customer is not just one unit loss of sale but a termination of the relationship (Risselada et al., 2010). Within the insurance context, churn is defined as the behaviour of a customer that cancels a policy, either because the need of the insurance is no longer present or the customer switched to another insurance provider (Günther, Tvete, Aas, &

Sandnes, 2014). In literature, churn is often known as “switching” (Wieringa & Verhoef, 2007) or defection (Neslin et al., 2006). In this paper, these concepts are used interchangeably. Concepts that are also related to churn but represent the counterpart, are customer retention, customer loyalty and repurchase behaviour. Because these concepts are highly correlated and are related to the same construct, that is (dis)loyalty, this paper uses all concepts to generate insights and set hypothesis regarding the drivers of churn.

To handle the major problem ‘churn’ many organisations make use of churn prediction models, especially within sectors where customers are bonded to contracts (Verbeke et al., 2012) such as the service sector. Churn prediction models are used to predict the customers with the highest probability to churn (Neslin et al., 2006). With these predictions marketing resources can be used more efficient (Neslin et al., 2006). The performance of a churn prediction model, generally measured with the use of the top decile lift, depends on the model chosen and the variable selection (Neslin et al., 2006; Verbeke et al., 2012). It is important to understand the reasons of churn and the effect of marketing efforts, so the most accurate variables can be included in the model which is necessary for achieving good predictive performance (Verbeke et al., 2012). Previous research has tried to

understand churn by examining the customer experience (Bolton, 1998; Sharma & Patterson, 2000), examining critical incidents that caused churn (Keaveney, 1995) or identified characteristics that differentiate churners from non-churners (Keaveney, 2001). Often examined drivers of churn are:

satisfaction, bad service experiences, customer characteristics such as: age, gender, involvement, income and variety seeking, relationship characteristics such as: relationship length, trust and relationship commitment, marketing characteristics such as: price and quality, direct communication and loyalty programs and, environmental characteristics such as: switching costs and competition.

(11)

5 To be able to build a good performing churn model, the concept of churn should be

understood. Therefore, this chapter starts with describing the crucial concepts and drivers of churn.

Customer experience is found to be a crucial concept with regard to the length of the relationship, if customer experiences are not satisfactory, the relationship length is short (Bolton, 1998). In addition, satisfaction is one of the most examined concepts in the literature of churn and customer experience (De Haan, Verhoef, & Wiesel, 2015; Verhoef, Van Doorn, & Dorotic, 2007). Many studies have examined the direct, indirect, and moderating effects of satisfaction on churn (Gustafsson, Johnson, &

Roos, 2005; Homburg & Giering, 2001; Mittal & Kamakura, 2001; Seiders, Voss, Grewal, & Godfrey, 2005). Besides, Customer Satisfaction is a popular customer feedback metric to measure the customer experience (De Haan et al., 2015). But is satisfaction the best metric to predict churn or are there other metrics that are having a higher predictive power? To answer this question, this chapter starts with a discussion of the existing literature regarding the customer experience and the relation with churn. The discussion is followed by a debate about the customer journey and the role within customer experience and churn. The customer journey is included because it is a crucial concept within customer

experience (Zomerdijk & Voss, 2010) and previous research has found that customer feedback metrics differ in their predictive performance per customer journey stage (Koster, 2016). Therefore, including the customer journey could lead to a better understanding of the construct churn, whereby churn prediction can be improved. The remaining of this chapter is discussing the potential other drivers of churn, that are described in literature. These drivers are divided over three overarching groups namely:

customer characteristics, marketing efforts and environmental characteristics. This chapter ends with a discussion about appropriate statistical techniques to predict churn because the choice of statistical technique does influence the performance of a churn model.

2.1. The customer experience

Customer experience is recognized as a crucial concept for understanding reasons of churn (Sharma & Patterson, 2000). CEM is defined by Homburg et al. (2015) as “the cultural mind-sets toward customer experiences, strategic directions for designing customer experiences, and firm capabilities for continually renewing customer experiences, with the goals of achieving and sustaining long-term customer loyalty”. In addition, other scholars defined the customer experience, the customer experience in the retail setting is defined as followed: “the customer experience construct is holistic in nature and involves the customer's cognitive, affective, emotional, social and physical responses to the retailer”(Verhoef et al., 2009). Lemon and Verhoef (2016) combined several definitions of customer experience which have resulted in the following definition “the customer experience is a

multidimensional construct focusing on a customer’s cognitive, emotional, behavioural, sensorial and social responses to a firm’s offerings during the customer’s entire purchase journey”. This paper uses the definition of Lemon and Verhoef (2016) because this definition is broad and applicable to any kind of sector.

(12)

6 There have been a few scholars who tried to conceptualize customer experience management.

The ‘Customer Experience Framework’ created by Nasution et al. (2014) is one of them, this framework is mainly focussing on the customer experience and the customer journey for the service sector. The Customer Experience Framework (CEF) is used to understand the process of customer experience management and examine the success of the CEM strategy (Nasution et al., 2014). The most basic element of the Customer Experience Framework are the values, wants and needs of the customer (Nasution et al., 2014). Products and services should be in line with the needs of customers to be able to create customer experiences (Nasution et al., 2014). To provide unique emotional and physical customer experiences an organisation should make use of the right set of experiential strategies (Nasution et al., 2014). These strategies will attract customers to interact and be involved with the product and service throughout different stages of the purchasing process (Nasution et al., 2014). The interactions and involvements result in many customer experiences which lead to an accumulated customer experience (Nasution et al., 2014). A positive accumulated customer experience leads to an emotional relationship with the organisation: which creates trust, satisfaction and loyalty (Nasution et al., 2014). This conceptualization of customer experience infers that customer experience management is a strategy that enhances customer experiences which ensures loyalty behaviour so that CEM can be seen as a driver of churn reduction.

2.1.1. The effect of the customer experience on churn

To be able to examine the impact of the customer experience on churn, the customer experience should be made measurable. When measuring customer experience one has to take into account the multidimensionality of the construct and the different customer response types over the entire customer journey, as mentioned within the definition of customer experience (Lemon &

Verhoef, 2016). An ideal metric would measure the overall customer experience, but such a metric is still in development (Lemon & Verhoef, 2016). Several organisations have attempted to create such metrics, an example is the Customer Experience Index (CIX). The CIX capitalizes on the

multidimensionality of the customer experience by measuring the emotional and- rational customer loyalty. The emotional and- rational loyalty of customers is measured with a sequence of questions.

These questions are found to be drivers of Customer Satisfaction. CIX can be seen as a metric to measure the Customer Satisfaction, which can be linked to the financial performance of the

organisation (Store support, 2017). This metric is still in development, most organisations measure the customer experience with the customer feedback metrics (Lemon & Verhoef, 2016). Customer

feedback metrics do often measure certain aspects of the customer experience, such as: overarching customer perception, perception for a single transaction or the experience in a specific point in time (Lemon & Verhoef, 2016). Customer feedback metrics that are often debated in literature and/ or used in organisations are: Customer Satisfaction, Top 2 box Customer Satisfaction, Contact Satisfaction Score, the Net Promoter Score and the Customer Effort Score (De Haan et al., 2015). These metrics

(13)

7 are discussed within the following paragraphs. In addition, relationship commitment and trust are seen as customer metrics, these metrics are less often examined in literature or used in management systems (Gupta & Zeithaml, 2006).

Previous research is not conclusive about the effect of the different customer feedback metrics and the relation with customer retention. De Haan et al. (2015) researched the effect of a selection of customer feedback metrics on customer retention in different sectors, so this research accounts for the potential differences in effects in sectors. Customer Satisfaction, Top-2 box Customer Satisfaction and Net Promoter Score are customer feedback metrics that are significantly related to customer retention in almost all sectors, but the predictive power of the metrics differ per sector (De Haan et al., 2015).

Previous research found that a combination of different metrics improves predictions (De Haan et al., 2015; Morgan & Rego, 2006), because different dimensions of customer experience are combined.

Because CEM is recognized as a crucial concept for understanding churn (Sharma & Patterson, 2000), customer feedback metrics are used to measure customer experience (Lemon & Verhoef, 2016) and most customer feedback metrics are significantly related to customer retention in different sectors (De Haan et al., 2015), the following hypothesis is formulated:

Hypothesis 1: Including the following customer feedback metrics: Customer Satisfaction, the Contact Satisfaction Score, the Net Promoter Score and the Customer Effort Score, improves churn prediction

compared to a model without these metrics in terms of the top decile lift 2.1.2. Customer Satisfaction

Customer Satisfaction is the dominant customer feedback metric within marketing (De Haan et al., 2015). Customer Satisfaction is a summary of the experienced psychological state created by the customer during confirming or disconfirming expectations regarding a specific service transaction or experience (Palmer, 2010). Customers are satisfied when actual performance exceeds the expectation of the performance (Szymanski & Henard, 2001). Customers are dissatisfied when a customer expects a higher performance than the customer is actually experiencing (Szymanski & Henard, 2001).

Existing literature is not conclusive about the relation between satisfaction and churn (De Haan et al., 2015; Gustafsson et al., 2005; Verhoef, 2003; Verhoef, Van Doorn, et al., 2007). Verhoef (2003) summarises the studies that have investigated the relation between satisfaction and customer retention, three of the four studies found a significant negative relation. Also Gustafsson et al. (2005) and Koster (2016) found a negative significant relation between satisfaction and churn: the more satisfied customers are, the less likely they churn. In contrast, Verhoef (2003) found that Customer Satisfaction does not have an effect on customer retention. The contradicting findings could be caused by a variety of reasons: the metric used to measure Customer Satisfaction, the type of sector that is researched, the inclusions or absence of other variables within the model, and the inclusion and/or exclusion of moderators.

(14)

8 Satisfaction is measured with different metrics, the most familiar ones are the top 2 box satisfaction and overall Customer Satisfaction (De Haan et al., 2015). In addition, the Contact

Satisfaction Score is used by organisations but not recognized in literature. The top 2 box satisfaction and overall Customer Satisfaction are metrics that measure the satisfaction level of the customer in general. In contrast, the Contact Satisfaction Score is a metric that measures the satisfaction level of a customer at a specific moment in time, this metric is discussed in the next paragraph. So, previous research did not use the same kind of metric to measure the Customer Satisfaction, the different effects found between Customer Satisfaction and churn can be the result of the difference in metrics used to measure the Customer Satisfaction. In addition, the contradicting effects could have to do with the backward looking aspect of the Customer Satisfaction measurement (Verhoef, Van Doorn, et al., 2007). Meaning that, with a backward looking metric, it is possible that between the time of measurement and churn, certain events can occur that cause the churn, so that these events are not captured within the backward looking metric and the effect between the backward looking metric and churn becomes unclear (Verhoef, Van Doorn, et al., 2007).

The difference in findings can also be the result of the sector that is studied. De Haan et al.

(2015) studied the performance of customer feedback metrics per sector, including the overall

Customer Satisfaction metric and the top 2 box Customer Satisfaction metric. The top 2 box Customer Satisfaction had the highest impact on customer retention across industries (De Haan et al., 2015). But, both Customer Satisfaction measures had an insignificant relation with customer retention within the insurance sector (De Haan et al., 2015). This indicates that the measurement used and the sector of research matters for the relation between satisfaction and churn.

Other authors suggest a potential non-linear relation between satisfaction and churn (Mittal &

Kamakura, 2001) and potential other variables that moderate the relation between satisfaction and churn (Verhoef, Van Doorn, et al., 2007). Moderators examined in literature are customer

characteristics (Mittal & Kamakura, 2001), and environmental variables (Verhoef, Van Doorn, et al., 2007). The potential moderation effect of customer characteristics and environmental factors on the relation between satisfaction and churn is respectively discussed in paragraph 2.3.1. and 2.3.4. Overall, the literature suggests that there is a relation between satisfaction and churn, based on that the

following hypotheses is formulated:

Hypothesis 1A: Customer Satisfaction is negative related to churn 2.1.3. Contact Satisfaction Score

Contact Satisfaction Score (CSS) is a customer feedback metric that is used by organisations but is not recognized in literature. The CSS is measured by the following question: How satisfied are you with the way the contact between you and the organisation has expired? Indicate your satisfaction with a number between one and ten, where a one stands for "very dissatisfied" and ten for "very

(15)

9 satisfied". The validation and predictive power of this metric is not examined in previous research, although some research has recognized the importance of contact moments with an organisation regarding churn (Keaveney, 1995). Because contact moments, also called touchpoints (Norton & Pine II, 2013), are important regarding churn and overall Customer Satisfaction is expect to have a

significant negative relation with churn, the following hypothesis is formulated:

Hypothesis 1B: the Customer Satisfaction Score is negative related to churn 2.1.4. Net Promoter Score

The Net Promoter Score (NPS) is a customer feedback metric that was introduced by

Reichheld (2003). This metric measures the Customer Satisfaction with a single question: How likely is it that you would recommend this company to a friend (Reichheld, 2003)? This question can be answered with a value between one and ten. The values one until six are the respondents that are the detractors, the values seven and eight are the passive respondents and the values nine and ten are the promotors. The promotors are the respondents that would recommend the company to a friend. The number of promotors are suggested to be a prediction of the future business performance (Reichheld, 2003). Within the years, different NPS metrics are developed such as: the NPS official and the NPS value (De Haan et al., 2015). But the opinions about the relation between the NPS and future company growth are divided (Morgan & Rego, 2006). Morgan and Rego (2006) found that there is no clear relation between the NPS and the business performance. Research that examined more specifically the relation between the NPS and customer retention, found that the NPS is one of the best performing customer feedback metrics regarding the impact on customer retention within organisations across industries (De Haan et al., 2015). For the insurance sector, the official NPS performs best compared to other customer feedback metrics (De Haan et al., 2015). The contradicting findings could have to do with the uncertainty of the metric, because the metric is asking for the recommendation intentions.

When someone intends to recommend, it is not certain he or she is actually going to do it, making the relation with business performance or customer retention unclear. In addition, the contradicting findings could have to do with the use of different NPS metrics and the different sectors that are researched. Because De Haan et al. (2015) find a significant negative relation between the NPS and churn within the service sector, the following hypothesis is formulated:

Hypothesis 1C: the Net Promoter Score is negative related to churn 2.1.5. Customer Effort Score

The Customer Effort Score (CES) is a relative new customer feedback metric that has been introduced by Dixon, Freeman and Thomas (2010). This metric measures the perception of the service performance of a specific past experience by a single question: How much effort did you have to put forth to handle your request (De Haan et al., 2015)? A value of one indicates a low amount of effort, a value of five indicates a high amount of effort. De Haan et al. (2015) is the only research that has

(16)

10 examined the performance of the CES in relation to customer retention, as far as known. De Haan et al. (2015) researched the effects between different customer feedback metrics and customer retention in different sectors. This research found a significant effect between CES and customer retention in one sector, namely banks. In the other 17 sectors, no significant effect is found. But, this research also found that CES has incremental power: when the CES and customer satisfaction metric are combined the predictive performance of customer retention is improved in terms of the top decile lift. The authors suggest that this result has to do with the specific and past focus of the metric, which makes the metric not appropriate for measuring the overall experience (De Haan et al., 2015) because customer experience is a multidimensional construct. But, combining different customer feedback metrics, which have different types of focus, will improve the prediction of customer retention because in this way multiple dimensions of the experience are measured. Because CES is introduced with the thought of having an impact on business performance and it seems logical that customers who have had exert a lot of effort to get their request handled are more likely to churn, the following hypothesis is formulated:

Hypothesis 1D: the Customer Effort Score is positive related to churn 2.1.6. Relationship commitment and trust

The type of relationship commitment and trust are also seen as customer metrics, but these metrics are less often used for management measure systems and academic research (Gupta &

Zeithaml, 2006). In addition relationship commitment is suggested to have a mediating effect on loyalty and a moderating effect on the relation between Customer Satisfaction and- trust on churn. The effects of Relationship commitment and- trust are therefore briefly discussed within the following paragraphs.

Relationship commitment is the viscosity that keeps customers loyal to an organisation even when satisfaction may be low (Gustafsson et al., 2005). This infers a direct relation between

relationship commitment and loyalty. Lemon and Verhoef (2016) suggest that commitment mediates the effect between customer experience and loyalty, commitment is the consequence of customer experience and influences loyalty behaviour of customer. Two overarching causes for viscosity are defined: affective commitment and calculative commitment (Gustafsson et al., 2005). Affective commitment is the emotional attachment to the organisation, such as liking the brand (Gustafsson et al., 2005). Calculative commitment is the commitment that comes from switching costs, dependence and lack of choice (Gustafsson et al., 2005). Switching costs are an example of negative consequences of switching, which create an exit barrier for customers to switch and therefore can be seen as a reason to remain the relationship (Verhoef, Van Doorn, et al., 2007). Switching costs are found to have an impact on the nature of the relation between trust and satisfaction on the length of the relationship (Sharma & Patterson, 2000). The relation between satisfaction and relationship length is weaker when

(17)

11 switching costs are high (Sharma & Patterson, 2000) which infers that calculative commitment

moderates the effect between satisfaction and relationship length. Fullerton (2003) found that churn intentions are highest when affective commitment and calculative commitment are low, and churn intentions are lowest when affective commitment is high, but calculative commitment is low. These results suggest that customers have a stronger commitment to an organisation when the tie is voluntarily created by emotions and feelings regarding an organisation and that this emotional tie results in loyal behaviour, so churn reduction. Verhoef (2003) confirms that affective commitment is positively related to customer retention, but this research recommends to create committed customers with the use of loyalty programs that are economic incentive driven because economic driven

programs leads to greater customer retention according to their finding. The role of loyalty programs within churn is discussed in paragraph 2.3.4. In summary, all research found a positive relation

between commitment and customer retention, but the type and strength of commitment depends on the underlying motives: emotional attachment with organisation or psychological and- economic costs (Bolton, 1998).

The concepts of commitment and trust are often named together, because they measure the same construct. As mentioned in the paragraph above, commitment is the customer’s connection to an organisation. Trust is the overall assessment of the benevolence and reliability of an organisation (Lemon & Verhoef, 2016). Trust could be built by customer experiences (Lemon & Verhoef, 2016). It is found that just as commitment, trust is significant positively related to retention (Verhoef, Langerak,

& Donkers, 2007). So that, the amount of trust can cause the different outcomes of the relation between Customer Satisfaction and churn.

2.2. The customer journey

Within the introduction, designing the customer journey and the associated touchpoints are introduced as a way to improve customer experience (Zomerdijk & Voss, 2010). This suggest that the customer journey and associated touchpoints play important roles within the relation between

customer experience and churn. The customer journey is an order of events that the customer goes through to learn about the organisation and- his offerings, purchase products or services and interact with the organisation's offerings (Norton & Pine II, 2013). In general the customer journey consist of the previous experience stage, the pre-purchase stage, the purchase stage, the post-purchase stage and the future experience stages (Lemon & Verhoef, 2016). Each customer stage can have multiple touchpoints. Touchpoints are moments when the customer “touches” the organisation (Norton & Pine II, 2013). It is hard to standardize the customer journey and the associated touchpoints because this is a dynamic and- ongoing process (Zomerdijk & Voss, 2010) and specific per sector, organisation and customer. Because this paper is focussing on the service sector, the next paragraph discusses the customer journey and associated touchpoints within the service sector.

(18)

12 2.2.1. The customer journey in the service sector

Generally, the customer journey in the service sector exist of the following stages: pre- purchase stage, purchase stage, customer stage, service experience stage, service request stage and the termination stage (Store support, 2015). The customer journey stages and the associated touchpoints are described below. Also the expected direct relation between customer journey and churn is discussed. Besides the direct effect, a moderation effect is expected because previous research has found that customer feedback metrics do differ in performance per customer journey stage (Koster, 2016). A predictive model that includes an interaction between the best performing customer feedback metric and the customer journey stages is found to have the highest predictive power regarding predictions of the change in number of policies (Koster, 2016). Therefore, the direct effect of the customer journey stages on churn and the interaction effect of the customer journey stages and customer feedback metric on churn are discussed below. As stated in the definition of churn, someone can only churn when someone is a consumer (Günther et al., 2014). Someone is a consumer when a purchase has taken place. Therefore, the customer journey stages: purchase stage, customer stage, service experience stage and service request stage are the stages where churn can take place.

Therefore, these stages receive more attention in the discussion below.

Stage 1. Pre-purchase: this stage involves all interactions with the organisation and offerings before the purchase transaction has taken place (Lemon & Verhoef, 2016). This starts with a

recognition of the need, searching for the possibilities and consideration of the purchase (Lemon &

Verhoef, 2016). Associated touchpoints are: brand awareness, social media, sponsoring, mailings, presence in the region, commercials and attention in the media (Store support, 2015). Information with regard to this customer journey stage gives insights in the reason that someone became a customer of an organization. Someone that is assigned to this stage cannot churn because he is not a customer of an organization yet, but having information with regard to this stage could be valuable to explain churn.

The best performing customer metric for predicting the change in the number of policies within this stage is the NPS (Koster, 2016).

Stage 2. Purchase: this stage involves all interactions with the organisation during a purchase (Lemon & Verhoef, 2016). The service sector is a sector that works on contractual basis, most of the times. Therefore, the purchase stage involves mostly exchanging personal data and arranging the contract. Associated touchpoints are: orientation across platforms such as the internet, personal advice of an organisation and/or- recommendation (Store support, 2015). It is assumed that consumers within this stage are involved with the organisation because consumers in this stage interact with the product and service. It is found that involvement is positive related to loyalty (Hui, Ting, & Xiaoyi, 2017), so that customers that are more involved are more loyal. The involvement that comes from the interaction with a product or service is called purchase involvement (Beatty, Homer, & Kahle, 1998). Besides purchase involvement, Beatty, Homer and Kahle (1998) defined ego involvement which is created by

(19)

13 the inherent interest, needs and values with the purchase category (Beatty et al., 1998; Seiders et al., 2005). Whether the customer is ego involved, depends on the personality of the customer. It is shown that involvement moderates the effect of satisfaction on sales and the intention to repurchase

(Homburg & Giering, 2001). In general, customers that are higher involved with a product or service know more about the product or service category and therefore are needing less information during the buying process (Homburg & Giering, 2001). So that, a higher involved person emphasis less on the consulting and advisory contact moments during the purchase process, which leads to a weaker effect between satisfaction and repurchase behaviour for higher involved persons (Homburg & Giering, 2001). The best performing customer metric to predict the amount of policies within this stage is the CSS (Koster, 2016). This paper suggest that the strength of the relation between the purchase stage and the CSS depends on the ego involvement of the customer.

Stage 3. Customer: this stage involves all interactions with the organisation when the customer is in possession of a contract but did not use the service yet. Associated touchpoints are:

welcoming the customer, extending the contract, gaining insights into the contract or service, receiving proactive information or offers, consultation session and payment of the fee (Store support, 2015). It is assumed that consumers within this stage are less involved with the service organisation than followed up stages because consumers in this stage are customers that did not used the service yet, so that it is likely that these customers did not had contact with the organisation for a period of time. Koster (2016) did not find a significant interaction effect between the customer stage and a customer feedback metric when predicting the change in number of policies. This could be related to the absence of any customer service experience within the customer stage. Customers within the customer stage are likely to have had no contact with the organisation, so that it is logical that customer feedback metrics that are measuring the customer experience, are not significantly interacting with this stage.

Stage 4. Service experience: this stage involves all interactions with the organisation during the usage of the service: referring to the post-purchase stage mentioned earlier in this section. The post-purchase stage is suggested to be a trigger that either could lead to customer loyalty or starting the pre-purchase stage again: searching and considering alternatives (Lemon & Verhoef, 2016).

Associated touchpoints are: declaration, declaration procession, contact with emergency centre, contact with experts, use emergency service, refinisher, damage reparation and payment of declaration (Store support, 2015). Customers that are in this stage are customers that make use of the service. For some sectors this means that the customer aims for reimbursing a declaration from an organisation.

Consumers in this stage are likely to be involved with the organisation because they interact with the service, which refers to purchase involvement (Beatty et al., 1998). High involvement is positive related to loyalty (Hui et al., 2017), so that the high involvement state of the customer within the service experience stage infers a direct negative effect between this stage and churn. Also, the

principle of reciprocity infers a direct effect. The principle of reciprocity states that people do to others

(20)

14 as they do to us (Fennis & Stroebe, 2016), which means that when the insurance organisation did a favour to the customer, the customer will return the favour. Based on the reciprocity theory, customers who made use of the service and reimbursed a declaration, are not likely to churn. In addition to the direct effect, a moderation effect is expected, Homberg and Giering (2001) found some evidence of an interaction effect between involvement and satisfaction when modelling repurchase behaviour. This suggests that including an interaction effect between service experience stage and satisfaction could improve churn prediction. In addition, Koster (2016) found that satisfaction is the best performing customer feedback metric within the service experience stage and that this interaction does significantly improves prediction regarding the change in number of policies.

Stage 5. Service request: this stage involves all interactions with the organisation after the usage of the service. Associated touchpoints are: complaints, questions and change in contracts. The largest part of the customers in this stage have complaints, expressing these complaints ensures that the customers end up in this stage (Store support, 2015). Because expressing complains, means interacting with the organisation, the customers in this stage are involved with the organisation.

Literature states that complaining behaviour will give insights in Customer Satisfaction and business performance (Morgan & Rego, 2006), which suggest an interaction effect between complaining behaviour and Customer Satisfaction in regard to churn and a direct effect between complaining behaviour and churn. Previous research found the following reasons for churn: core service failure, service encounter failure, employees responses to service failure and inconvenience (Keaveney, 1995).

Based on this, a direct effect between service request stage and churn is suggested. Keaveney (1995) found that 45% of the respondents in his research relate their churn to a single factor, the top three reasons of churn where: core service failure, pricing problems and service encounter failures. 55% of the respondents relate their churn to a combination of the previous mentioned factors. When an organisation is able to recover or prevent a second failure, customers can still be satisfied and churn can actually be reduced. Service recovery can take place in the service request stage, so that this stage does not exclusively consists of customers that are dissatisfied. To improve churn prediction, it is suggested to include an interaction effect between a customer feedback metric that evaluates the satisfaction level and, the service request stage so more accurate predictions can be done. Previous research conclude that the NPS is the best performing customer feedback metric within the service request stage and the interaction between the NPS and service request stage improves prediction regarding the change in the number (Koster, 2016).

Stage 6. Termination: this stage involves all interactions with the organisation that relate to terminating a contract with the organisation. This is the stage in which a customer churns, either because the need is no longer present or the customer switched to another organisation (Günther et al., 2014). Associated touchpoints are: termination of all contracts, termination of a part of the contracts, retention and win back (Store support, 2015). This is the stage of interest for this research.

(21)

15 2.2.2. The effect of the customer journey on churn

Managing the entire customer journey can have enormous rewards for the organisation such as reduced churn rates (Rawson et al., 2013). Therefore, it is interesting to see what the impact is of customer journey stages on churn. As described within this section, churn can only take place from the customer stage onwards, meaning that customers within: the purchase stage, the customer stage, service experience stage and service request stage can churn, others not. The suggestions that the involvement level and complaining behaviour differs across customers per stage infers that the churn rates differ per customer journey stage. Therefore, it is expected that including the customer journey stages in a churn model will improve prediction. This leads to the following hypothesis:

Hypothesis 2: The purchase stage, customer stage, service experience stage and service request stage differ in churn rate, so that including these stages improves churn prediction in terms of the top decile

lift

2.2.3. Moderation effect of the customer journey and customer feedback metrics on churn

Inferences between the interaction of customer journey stages and customer feedback metrics are made earlier this section. The service experience stage and services request stage are stages that consist of customers that experienced the service. This service experience can be perceived as positive or negative. Customer feedback metrics is a measurement tool that can be used to evaluate the service experience. So that, it is likely that the evaluation of the customer feedback metric moderates the effect between the customer journey stages and churn prediction. Koster (2016) found that the interaction between customer feedback metrics and customer journey stages do improve predictive power of a model that predicts the number of policies (Koster, 2016). Therefore, the following hypothesis is formulated:

Hypothesis 3: customer feedback metrics moderate the strength of the relation between the customer journey stages and churn, so that including these interactions improves churn prediction in

terms of the top decile lift

More specifically, the NPS is found to be the best performing customer feedback metric within the pre-purchase stage and the service request stage with regard to predicting the change in number of policies (Koster, 2016). Because customers within the service request stage complain often and complaining behaviour gives insights into the relation between Customer Satisfaction and churn it is expected that churn prediction can be improved by the inclusion of an interaction effect between the NPS and the service request stage. The CSS is the best performing customer feedback metric in the purchase stage and Customer Satisfaction is the best performing customer feedback metric in the termination stage with regard to predicting the change in number of policies (Koster, 2016). So that, it is expected that churn can be improved by taken into account the interaction effect of CSS and the purchase stage. In addition, it is expected that an interaction between satisfaction and service

(22)

16 experience stage improves churn prediction, because this stage exists of high involved customers and involvement moderates the effect between satisfaction and repurchase intention (Homburg & Giering, 2001).

2.3. Customer, marketing and- environmental characteristics

In the above paragraphs the drivers of churn regarding the customer experience and customer journey were discussed. Previous research has highlighted some potential other drivers of churn:

gender (Homburg & Giering, 2001; Mittal & Kamakura, 2001), age, income (Gustafsson et al., 2005;

Homburg & Giering, 2001; Seiders et al., 2005), education (Mittal & Kamakura, 2001), situational triggers (Gustafsson et al., 2005), variety seeking (Homburg & Giering, 2001), price and quality perception, communication (Venkatesan & Kumar, 2004), loyalty programs (Seiders et al., 2005;

Verhoef, 2003; Verhoef, Van Doorn, et al., 2007), competition, (Sharma & Patterson, 2000; Verhoef, Van Doorn, et al., 2007) and sector specific conditions (Duijmelinck, Mosca, & Ven, 2015). These drivers are divided across overarching terms namely: customer characteristics, marketing

characteristics and environmental characteristics.

2.3.1. Customer characteristics

Customer characteristics seem to be important drivers of psychological and behavioural constructs (Homburg & Giering, 2001). Customer characteristics are often easy to measure and- incorporate and, are found to have a moderating effect within the relation satisfaction and loyalty behaviour, so this is worthwhile to investigate the effects of these factors on churn (Homburg &

Giering, 2001). The following customer characteristics are examined to have a direct, indirect and/or moderating effect on churn: gender, age, income, education, situational triggers and variety seeking.

These customer characteristics are discussed below.

Gender is proposed to have an effect on churn because men and women seem to value different aspects of products and services when they shop, so that it is expected that they show

different loyalty behaviour (Homburg & Giering, 2001). Men tend to base the repurchase intentions on the satisfaction of the functionality of the product itself in contrast to women. Women base the

repurchase intentions on the satisfaction about the personal interaction (Homburg & Giering, 2001).

Mittal and Kamakura (2001) found that the overall probability of a repurchase is higher for women than for men, where satisfaction does not seem to mediate the effect. This research suggest that overall women are more loyal, independent of satisfaction.

Age is proposed to have an effect on churn due to the difference in information processing between young and elderly customers (Homburg & Giering, 2001). Age is shown to have a positive moderating effect between satisfaction and loyalty (Homburg & Giering, 2001). Younger customers in contrast to elderly customers, rely less on satisfaction with the product, they base their decisions on the information provided by sales personnel (Homburg & Giering, 2001). Mittal and Kamakura (2001)

(23)

17 confirms that elderly customers do have more stable preferences compared to younger customers, therefore the change is higher that they will purchase again.

Income is proposed to have an effect on churn because difference in income can cause different perceptions of the financial risk associated with the purchase of a product or service (Homburg & Giering, 2001). But results regarding this relationship differ across previous research.

Seiders et al. (2005) conclude that there is no direct effect between income and repurchase intentions and that income does not moderate the effect between satisfaction and repurchase intentions. In contrast, Homburg and Giering (2001) found that income moderates the relation between product satisfaction and loyalty. The relation between product satisfaction and loyalty is weaker for customers with higher incomes in contrast to customer with lower incomes (Homburg & Giering, 2001).

The findings of Homburg and Giering (2001) are explained by the potential correlation between income and education. People with higher incomes are expected to have a higher education level so that these customers are more engaged in information processing and base their purchase decision on information instead of satisfaction (Homburg & Giering, 2001). Mittal and Kamakura (2001) confirm that higher educated customers are having lower levels of retention because they are able to process information more sufficient and are more cognizant of alternatives in the market. But Mittal and Kamakura (2001) do not confirm the role of satisfaction within the relation between education and retention.

Besides the direct and indirect effects of income on churn, it is also suggested that the change in living, working or family conditions can be a trigger of churn or can be a situational trigger that influences the effect of satisfaction on churn, among other this includes change in income (Gustafsson et al., 2005). This relation seems logical, because a change in income can ensure that a price of a product is perceived higher or lower than before so that, a customer may be triggered to search for a better deal and is therefore more likely to churn. But, Gustafsson et al. 2005 did not found a direct or indirect effect of situational triggers on churn.

Under certain conditions everybody has the need to seek for variety (Homburg & Giering, 2001). Seeking for variety means that the wants and needs of the customer cannot be fulfilled by a single product, so that a customer seeks variety to avoid feelings of boredom and monotony (Homburg

& Giering, 2001). Homburg and Giering (2001) found that variety seeking is negatively moderating the effect between satisfaction with the product and loyalty. Meaning that the impact of satisfaction on loyalty is weaker for customers who have a strong need to seek for variety.

Referenties

GERELATEERDE DOCUMENTEN

A consumer is closer to the conversion square when visiting the focus brand’s website, than an information/comparison website or app, a generic search or a competitor’s

The final objective of this thesis, was to study the possible positive moderating effect of using a small screened device (versus a large screen) while visiting inspiration

Keywords: customer journey, touchpoints, segmentation, profit drivers, online, discounted expected transactions, negative binomial regression, latent class cluster

The gap in the literature on segments based on demographical variables and the rise in use of (multiple) devices in the customer journey do make research on this crucial. In

Clickstream data is the electronic record or internet usage (Bucklin and Sismeiro, 2009). However, literature lacks insights into a holistic view of actual web behavior with both

Checking if this is indeed true a Z-test must be performed according to Clogg, Petkova and Haritou (1995).. 20 dependent variable. Both conditions are satisfied for the comparison

• Provides insights into the effect of customer satisfaction, measured through online product reviews, on repurchase behavior!. • Adresses the question whether the reasons for

Besides investigating the overall effect of the five different customer experience dimensions (cognitive, emotional, sensorial, social, and behavioural) on customer loyalty, I