HOW TO INFLUENCE CUSTOMER REVIVAL

(1)

HOW TO INFLUENCE CUSTOMER

REVIVAL

By:

Marlies Slütter

(2)

Master thesis

University of Groningen

Faculty of Economics and Businesses

MSc Business Administration Specialization Marketing

Groningen, July 19th, 2011

Name: Marlies Slütter Adress: Vosdijk 17a

7134 RD Vragender Phone: (06) 22348864

E-mail: m.a.slutter@student.rug.nl St. number: 1539353

Supervisors: Regio Achterhoek: Rik Swieringa

(3)

MANAGENT SUMMARY

Increasing evidence exists that well managed customer revival activities can be highly effective and efficient (Griffin & Lowenstein, 2011). Besides, in business practice, it can be observed that companies are increasingly emphasizing activities which aim at winning back lost customers. However, in existing research, this subject is largely neglected. Therefore, customer revival is the topic of this research. More specifically, I researched which strategy a company should follow to recapture lost customers and what role social influence and attitude play in the customer revival process.

This research was applied to the situation of the Regio Achterhoek. The Achterhoek has to deal with a graying population. Companies have difficulties with filling vacancies with highly-educated people. This may be caused by the fact that hardly any HBO or university study could be followed in the Achterhoek. Therefore, a lot of young people leave the Achterhoek to go to college. Subsequently, they do not return because they think that they could not find a job in the region or that no interesting companies are located over there. The Regio Achterhoek is searching for a solution for this problem: what should they do in order to get these graduates back to the Achterhoek?

In order to answer the research question, I conducted interviews and I distributed a questionnaire among students that left the Achterhoek to go to college and that now live outside the region. In total, 153 respondents filled out my questionnaire. However, not all respondents were part of the target group, so the sample consisted of 133 respondents. With the results of the questionnaire, I created an ordinary least squares regression model with ‘probability of revival’ as dependent variable. I used stepwise estimation in order to create the best model. I tested the model for multicollinearity, autocorrelation, heteroscedasticity, and normality and adjusted it based on the results of these tests. This resulted in an equation with 12 independent variables. These independent variables were also used in the latent class regression analysis that I performed subsequently. This analysis was done in order to find out whether different segments could be identified with different parameters.

(4)

that the effect of ‘attitude with respect to the Achterhoek’ is strengthened by this interaction term. Besides, the preference for timing of information ‘a couple of years after graduation’ positively influences the probability of revival and a preference for ‘timing of information: other’ negatively influences this probability. This is probably caused by the fact that most respondents answered here that they were not interested in receiving information at all.

The latent class regression analysis led to a 3-class solution. The first class was characterized by a high probability of revival and a large negative influence of the variable ‘timing of information: other’. Furthermore, for this class, friends and family play an important role in the decision for a residence. For the second class, both the interaction term between ‘value-expressive influence’ and ‘attitude’ and the ‘reason of defection: study’ have a large positive influence on the probability of revival. The probability of revival of the third class is highly positively influenced by ‘general interest in receiving information’ and by the interaction term between ‘general interest in receiving information’ and ‘reason of defection: friends also left the Achterhoek’. These findings show that per class, different independent variables influence the probability of revival.

My research has several limitations that lead to avenues for future research. First, the sample was not representative of the population (in terms of municipality and education); this should be improved in future research. Second, only people that now live outside the Achterhoek filled out the questionnaire. In future research, people that already returned to the Achterhoek may also give some useful input for the Regio Achterhoek. Besides, measuring why people returned and which communication messages were successful will make the results more reliable. Third, large differences exist between marketing a region and marketing a product; future research could test whether the results of this study can be applied in the situation of products.

(5)

PREFACE

This master thesis is part of my Master Marketing (Management & Research) at the Rijksuniversiteit Groningen. I conducted my study at the Regio Achterhoek in Doetinchem, the region where I was born. I hope you all enjoy reading this report.

Using this opportunity, I would like to thank several people for their support during my graduation period. First, I would like to thank Thorsten Wiesel and Sonja Gensler of the Rijksuniversiteit Groningen for their input and their suggestions. Their comments helped me to finish my thesis successfully. Second, I would like to thank Rik Swieringa for supporting me as supervisor at the Regio Achterhoek. He gave me the freedom to research a topic that I personally experienced as very interesting.

Finally, I would like thank my family and friends for their input and support.

Marlies Slütter

(6)

1. INTRODUCTION

1.1 Background of the problem

In many industries, customer defection rates are very high, sometimes adopting values between 20% and 40%. BellSouth Mobility is an example of a company with such a high defection rate. It was losing around 500 of its customers a day. When losing a customer, a company often loses revenues and besides goodwill. Each lost customer is a potential ambassador of bad news. It is often difficult to bring down these rates, although lots of customer retention activities are performed. Hence, winning back lost customers has become, next to acquiring and maintaining customers, a major challenge in companies’ customer relationship management. BellSouth Mobility used focus groups to get more insight into the defection process and its lost customers’ motives. Subsequently, they created a customized direct-mail program with follow-up phone calls to lost customers. This way, they were able to achieve a 10 percent reconnect rate (Griffin & Lowenstein, 2001).

A lost customer is one who has established a relationship with a company but now has terminated the relationship (Homburg, Hoyer & Stock, 2007). Customers can either cease to purchase or explicitly terminate the relationship. A customer-regaining or customer revival strategy aims at rebuilding the relationship with lost customers (Stauss & Friege, 1999). Stauss and Friege (1999) found in a case study that the net return on investment from a new customer obtained from an external list is 23% compared with a 214% return on investment from a customer that has been reacquired. Besides, a study by Marketing Metrics found that an average firm has a 60 – 70 percent probability of successfully selling again to active buyers, a 20 – 40 percent probability of successfully selling to lost customers, and only 5 – 20 percent probability of making a successful sale to a new prospect (Griffin & Lowenstein, 2001). Next to these incremental sales and higher probability of selling, win-back programs have other benefits as well. Acquisition costs are lower compared to new prospect recruitment; service improvement opportunities can be identified by tracking the defected customers’ reasons for leaving; an increased capability arises to detect at-risk customers by learning from defected customers; and it creates the ability to limit negative word of mouth from switchers and increase positive word of mouth through those that are reacquired. As a result of these benefits, customer reacquisition initiatives are becoming more popular (Griffin & Lowenstein, 2001; Tokman, Davis & Lemon, 2007; and Stauss & Friege, 1999).

(9)

will present to the lost customer and how this is presented (Stauss & Friege, 1999). The customer revival process from the point of view of the customer is the topic of this thesis.

1.2 Research questions

The main research question that I will try to answer is: Which strategy should a company follow to recapture a lost customer and what role do social influence and attitude play in the customer revival process?

Before the main research question can be answered, several sub questions have to be answered:

1) Do the reasons of defection influence the relationship between customer revival strategy and the probability of revival?

2) Which strategy should a company follow to increase lost customers’ probability of revival?

3) To what extent does lost customers’ attitude with respect to the brand influence the probability of revival?

4) To what extent does social influence affect the relationship between the attitude with respect to the brand and the probability of revival?

1.3 Theoretical relevance

Several researchers have already addressed the topic of customer revival. However, the number of empirical studies on this topic is limited. None of these empirical studies addresses the effectiveness of different customer revival strategies. However, some of them did mention this topic as an important area for future research.

As Stauss and Friege (1999) argue, it is important to explain the differences in the reactions of lost customers to various ways of communication and various regain offers. Besides, Pick (2010) discusses existing research findings on the topic of customer revival. She also presents the process of win-back as an important aspect for further research. Furthermore, according to Polonsky, Gupta, Beldona and Hyman (2010) successful customer revival depends on a range of issues, including the specific partners and the factors that lead to previous inactivity. These issues should be considered from multiple perspectives, rather than solely from the perspective of the firm seeking to reactivate the relationship. According to these different authors, examining the process of customer revival in more depth is important. Therefore, this will be the topic of this research.

(10)

relationship (length, time since ending, customer satisfaction, social capital and reasons for defection) and characteristics of the revival offer (price, social benefits and interaction). These relationships are shown in figure 3 in Appendix I.

First, characteristics of lost customers are considered. Homburg, Hoyer and Stock (2007) found that age and involvement positively affect revival performance in the telecommunication industry. In addition, they found that the variety seeking motive has a negative impact on customer revival, so customers that have a tendency to seek diversity in their choices of services and goods are less likely to be reacquired.

Second, characteristics of the first relationship are topic of existing research. Research of Homburg, Hoyer and Stock (2007) showed that customer satisfaction with the revival activities and overall customer satisfaction positively influence lost customers’ behaviour. Besides, Thomas, Blattberg and Fox (2004) showed that the duration of the first relationship and lapse duration also influence the likelihood of reacquisition. The probability of reacquiring a customer is higher if the lapse duration is shorter and if the first tenure is longer. Tokman, Davis and Lemon (2007) researched what the influence of social capital (the accumulation of special treatments and favors) on the probability of relationship revival was. They found that social capital has a significant impact on the relationship between service benefits and win-back offer worth. Researchers do not agree whether reasons of defection influence the probability of revival, therefore, the relationship between the reasons of defection and the probability of revival will be examined in this research.

Third, characteristics of the revival offer may influence the probability of revival. The models that Thomas, Blattberg and Fox (2004) estimated showed that the offer price has the largest effect on reacquisition likelihood. Customers are more likely to be reacquired if the reacquisition price is lower. The likelihood of a customer being reacquired decreases with the difference between the reacquisition price and the last price offered in the prior relationship. Their research also showed that higher retention prices lead to longer relationship durations. Besides, Tokman, Davis and Lemon (2007) showed that a win-back offer presenting a low-price option was much more effective if the original switching reason was price related. In addition, Homburg, Hoyer and Stock (2007) concluded that the customer’s perception of what he gets depends largely on how he feels treated by the firm’s revival activities.

(11)

I will examine how interest in various customer revival strategies (in terms of moment and frequency of contact, content and communication channel) relates to the probability of revival and whether this relationship is moderated by the reasons of defection. Besides, in the customer revival process, I will distinguish lost customers’ attitude with respect to the brand and switch-back intentions. An attitude represents an association between an object and an evaluation (Gounaris & Stakhakopoulos, 2004). Ajzen (1991) claimed that behavioural intentions were influenced by the attitude toward the behaviour, subjective norms and perceived behavioural control. I will examine what role the attitude with respect to a company or brand and subjective norms play in influencing behavioural intentions, in this case switch-back intentions.

1.4 Region marketing

The main research question is answered by examining a very specific marketing situation: the marketing of a region. Economic and cultural globalisation are important trends that cities, regions and countries all over the world are faced with. One of the effects of globalisation is increased competition among places. Competition for resources, business relocation, foreign investments, visits and even residents is evident in today’s world (Kotler et al, 1999 cited in Karavatzis, 2005). This competition leads to the need for places to market themselves. In general there are four important groups that cities are competing for: citizens, companies, visitors, and creative talents and students (Hospers, 2009).

According to Ashworth and Voogd (1987), the concepts of ‘city marketing’ and ‘region marketing’ are used to describe processes that are applied to improve the position of the city or region within the market. City marketing can be defined as ‘the long-term process and policy instrument consisting of different, interrelated activities aimed at attracting and maintaining specific target groups for a certain city’ (Lombarts, 2008 cited in Hospers, 2009). City marketing can be divided into two types: efforts aimed at residents are called ‘warm city marketing’; those efforts that are aimed at attracting new residents are part of ‘cold city marketing’.

(12)

because of labor reasons. People that move over a longer distance are often highly educated and between 25 and 44 years old (Feijten & Visser, 2005). The older they get, the less people move over a long distance (Hospers, 2009). Young people often move because they want to live independent of their parents (28%), because of cohabitation or marriage (16%), or because of labour or study related reasons (27%) (Feijten & Visser, 2005).

1.5 Practical relevance

Currently, regions within the Netherlands have to deal with ‘graying’ residents (Hospers, 2009). The generation of baby boomers reaches the age of 65 and they are getting retired. Therefore, young people are needed to ensure that the area remains vital. Especially in regions in the South (Limburg), North (Groningen) and East of the Netherlands, young people are moving away after high school and often they do not return. This can be seen as defection: the customers (the young residents) of the region defect (they move) and they do not come back to the company (they do not return to the region). Therefore, the region has to try to regain the residents: how to get back these ‘lost customers’?

The Achterhoek – a region situated in the east of Holland – is an example of a region that does not have a university or an HBO-institution. Therefore, many young residents are moving to other places in the Netherlands for their education. These students are an important target group for the region, because it is easier to attract people that already have a bond with the region (Hospers, 2009).

(13)

Figure 1: Regio Achterhoek (in green) (Nederlandse Genealogische Vereniging, 2011)

After their graduation, half of the students that are originally from the Achterhoek prefer a job in the Achterhoek. If an interesting job if offered to them, 2/3 of the students prefers working in the Achterhoek, 30% possibly will work in the Achterhoek and only 4% definitely not. Reasons not to work in the Achterhoek are similar to its less attractive characteristics: its accessibility by public transport, lack of career opportunities or employment, no vivid cities or towns and its isolation (Leijenhorst et al., 2010).

The Regio Achterhoek – a partnership between eight municipalities in the Achterhoek as shown in figure 1 - tries to show its residents in its marketing campaign, using the slogan ECHT Achterhoek, that the Achterhoek is an attractive region to live and to work. The ‘warm’ region marketing campaign mainly focuses on the current residents and in particular the students that used to live in the region, but moved away to study at an HBO-institution or university. The Regio Achterhoek is now figuring out how it can reach these students so they go back to the region after their graduation. In this thesis, it will be examined what strategy the Regio Achterhoek should follow to increase the probability of return.

1.6 Structure of this thesis

(14)

Probability of customer revival Attitude Social influence RD: Study General interest in receiving information H1i-p H1a H3 H4 Age Involvement Variety seeking Length of relationship Time since ending Information channel

Information timing Information frequency

Customer revival strategy

Relationship characteristics H5 H6 H7 H8 H9 H2a H2b H2c H2d Customer characteristics

RD: Live on your own

RD: Work

RD: Student life

RD: Friend also left Achterhoek RD: Live in a differerent area RD: Being up to something new RD: Relationship H1b H1c H1d H1e H1f H1g H1h Reasons of defection

2. LITERATURE

As mentioned in chapter 1, topic of this research will be the process of customer revival. In this chapter, I will present my conceptual framework and the hypotheses that will be tested. In figure 2, this conceptual framework is shown.

Figure 2: Framework of the study

(15)

relationship characteristics (age of relationship and time since ending) that have been found to influence the probability of revival in existing research also influence the probability of revival in this specific situation.

2.1 Reasons of defection

Researchers do not agree whether reasons of defection influence the degree of success of revival activities, therefore I will include this into this research. Homburg, Hoyer and Stock (2007) did not find a significant effect of reasons of defection on revival performance. Contrary, Stauss and Friege (1999) and Helfert, Hermann and Zellner (2003) argued in their conceptual framework that companies should analyze reasons of defection and adjust the retention strategy to these reasons.

Bogomolova (2010) also showed in a business-to-business setting that customers that defect because of price reasons show a slightly higher propensity to associate the company with having high fees and being too profit-oriented. Furthermore, customers that defected because they received a better offer from a competitor had a more positive overall brand evaluation of their former brand compared to customers that defected because of price or service reasons. In addition, this group was more likely to consider their former brand. However, Bogomolova (2010) did not consider the relationship between reasons of defection and lost customers’ probability of revival.

As discussed in the introduction, the research question of regaining lost customer will be applied on a specific marketing situation: how to attract citizens back to the region where they were born. A lot of research has been done on the question why people move. According to Rossi (1980), moving is ‘adapting a living situation as much as possible to the needs of the habitant by changing the place to live’. In existing literature, three factors are presented that lead to the decision to move. First, the family life cycle may play a role (Rossi, 1980). In this case, people often move over short distances (Feijten & Visser, 2005). Second, changes in the labour situation of people may lead to moving to another place. The third reason consists of wishes with respect to improving the situation of living. When people want to live in a larger or nicer house, they also often move over a short distance.

Feijten and Visser (2005) researched why people move within the Netherlands. They used data of the Woon Onderzoek Nederland and the Gemeentelijke Basisadministratie Persoonsgegevens (GBA) as basis of their research. As mentioned in the introduction, young people mainly move because they want to live independent of their parents, because of cohabitation or marriage, or because of labour or study related reasons.

(16)

other reasons. The people that move because of their study often do not deliberately choose to move out of the region, they maybe would have chosen to stay in the region if it would have been possible to follow their study in the Achterhoek. Therefore, they are probably more likely to return. This reasoning also holds for the people that left because of their ‘work’. These people maybe would have stayed within the region if they could find a job in the Achterhoek. Therefore, they are more likely to return – if they could find a job in the Achterhoek.

Contrary, the people that moved because of other reasons are less likely to return. The group that wanted to live in a different area is less likely to return, since they moved because they did not like the area. The people that left the region because they wanted to live on their own are also less likely to return, because in the past they wanted to live independent of their family. In the future this may not change. Furthermore, it is not very probable that the people that left the region because they were up to something new will return. They wanted to experience new things and in the future they are not likely to return to their roots. The people that left the Achterhoek because of their relationship probably have a partner that is not originally from the Achterhoek. They decided to live together outside the Achterhoek. Therefore, they are also not very likely to return. Some people left the Achterhoek because they wanted to experience a student life; which is not possible within the region. These people want to enjoy studying in a large city. After their graduation, it is not very likely that they will return to the Achterhoek, since they are not looking for a quiet area to live. Finally, the last reason to leave the Achterhoek is that friends also left the region. When their friends return to the Achterhoek, these people may also return. However, it also likely that they together created a new community in their new student city or that they met new people. Therefore, they may not want to return. I have composed the following hypotheses:

H1a: The reason of defection ‘study’ positively influences the probability of revival.

H1b: The reason of defection ‘wanting to live in a different area’ negatively influences the probability of revival.

H1c: The reason of defection ‘wanting to live on your own’ negatively influences the probability of revival.

H1d: The reason of defection ‘being up to something new’ negatively influences the probability of revival.

H1e: The reason of defection ‘relationship’ negatively influences the probability of revival.

(17)

H1g: The reason of defection ‘student life’ negatively influences the probability of revival.

H1h: The reason of defection ‘friends also left the Achterhoek’ negatively influences the probability of revival.

In line with the framework of Stauss and Friege (1999), I expect that lost residents with different reasons of defection respond differently to customer revival strategies. I propose that the people that left the region because of study- or work-related reasons are more likely to respond to customer revival strategies. These people did not leave the region because they did not like the region, but because they could not follow the study they wanted or they could not find a job they wanted. When these circumstances change, they are likely to return. Therefore, their interest in receiving information may lead faster to a return to the Achterhoek. Hence, I expect that these reasons of defection play a positive moderating role in the relationship between the variables ‘general interest in receiving information’ and ‘probability of revival’.

The people that left the region because of one of the other reasons are probably less likely to respond to customer revival strategies. These people chose deliberately to leave the Achterhoek and interest in receiving information may not directly lead to a return to the Achterhoek, because these circumstances (for example lack of large cities in the region, a relationship outside the Achterhoek or friends that also left the region) probably will not change. Thus, I expect that in these cases, the reasons of defection play a negative moderating role in the relationship between ‘general interest in receiving information’ and ‘probability of revival’.

Therefore, I have composed the following hypotheses:

H1i: The relationship between general interest in receiving information and the probability of revival becomes stronger as the reason of defection ‘study’ plays a more important role.

H1j: The relationship between general interest in receiving information and the probability of revival becomes weaker as the reason of defection ‘wanting to live in a different area’ plays a more important role.

H1k: The relationship between general interest in receiving information and the probability of revival becomes weaker as the reason of defection ‘wanting to live on your own’ plays a more important role.

(18)

H1m: The relationship between general interest in receiving information and the probability of revival becomes weaker as the reason of defection ‘relationship’ plays a more important role.

H1n: The relationship between general interest in receiving information and the probability of revival becomes stronger as the reason of defection ‘work’ plays a more important role.

H1o: The relationship between general interest in receiving information and the probability of revival becomes weaker as the reason of defection ‘student life’ plays a more important role.

H1p: The relationship between general interest in receiving information and the probability of revival becomes weaker as the reason of defection ‘friends also left the Achterhoek’ plays a more important role.

2.2 Customer revival strategy

In this research, I will distinguish four characteristics of a customer revival strategy: content, channel, moment, and frequency of contact.

Marketers can use personal communication channels in which they communicate directly with lost customers, or nonpersonal communication channels; media that carry messages without personal contact or feedback (Kotler & Armstrong, 2008). Emerging empirical evidence on the topic of customer retention indicates that marketing contact trough direct mail, telesales, and sales people (personal communication channels) are critical for influencing retention. The relative effectiveness of highly interpersonal sales contacts is greater than that of less interpersonal modes (Verhoef et al, 2009). According to Stauss and Friege (1999), lost customers should be approached personally. This allows for an immediate categorization of the customer based on the causes of termination. I also propose that a more personal mode of communication is more effective in influencing lost customers’ decision to come back. I suggest that lost customers that are interested in receiving information via a personal channel have higher switch back intentions than customers that are less interested in receiving information via a more personal channel.

H2a: Preference for a more personal communication channel positively influences the probability of revival.

(19)

are more likely to switch back. Hence, I propose that customers that are interested in having more contact will also be more likely to switch back. Furthermore, I propose that customers that are interested in receiving information faster after defection have a higher probability of revival. This is in line with the research finding of Thomas, Blattberg and Fox (2004) who found that the probability of reacquiring a customer is higher if the lapse duration is shorter.

H2b: Preference for more frequent contact positively influences the probability of revival.

H2c: Preference for receiving information faster after defection positively influences the probability of revival.

Finally, I propose that general interest in receiving information about the Achterhoek positively influences the probability of revival. When people are interested in receiving information about the region, they still have a connection with it and therefore are probably also more likely to return.

H2d: General interest in receiving information positively influences the probability of revival.

2.3 Attitude and probability of revival

An attitude is an overall evaluation that expresses how much we like or dislike an object, issue, person, action or in this case, a region. Attitudes are important because they guide our thoughts, influence our feelings and affect our behaviour. Attitudes may be based on cognitions or beliefs, or on emotions (Hoyer & MacInnis, 2008).

Dick and Basu (1993) argue that attitudinal loyalty will lead to behavioural loyalty. This claim was researched by Bandyopadhyay and Martell (2007). Their results showed that single users of a brand (they use only one brand) demonstrate a stronger attitude than multiple users (they use more than one brand). Furthermore, they found that multiple users had a much higher attribute score than non-users. In line with these findings, I expect that lost customers that have a positive attitude with respect to their former brand have a higher likelihood to switch back than customers that have a negative attitude.

H3: Lost customers’ attitude positively influences the probability of revival.

2.4 Social influence

(20)

a certain product by the social group the individual refers to. By adapting their attitudes and behaviour, consumers fulfil their aspirations and at the same time reduce the perceived risk of making a decision. Ajzen (1991) also claimed that behavioural intentions were influenced by the attitude toward the behaviour, subjective norms and perceived behavioural control. I will not take this last variable into account, but I will include the role of social norms in this research.

Reference groups or other influence sources can exert normative influence or informational influence. Normative influence is social pressure designed to encourage conformity to the expectations of others (Hoyer & MacInnis, 2008). Perceived behavioral norms or role requirements, if contrary to an attitude, can lead to behavior that is not consistent with the attitude. For example, a teenager may have a high relative attitude toward a fashion boutique but may feel reluctant to buy clothes at the store due to his perception that his parents disapprove of the high price level of the store (Dick & Basu, 1994). Hoyer and MacInnis (2008) predict that reference groups are likely to have considerable influence on the brand that is bought when the product is publicly consumed but not when it is privately consumed. Furthermore, according to Hoyer and MacInnis (2008) reference groups particularly might exert influence when people buy a luxury item.

Next to normative influence, informational influence may also play a role. Reference groups and other influence sources can exert informational influence by offering information to help a person make decisions. Consumers tend to be susceptible to informational influence when considering complex products; when they perceive product purchase or usage to be risky’ and when they themselves cannot tell the difference between brands (Hoyer and MacInnis (2008).

I follow the reasoning of Dick and Basu (1994) and propose that social influence (both normative as well as informative influence) weakens the relationship between lost customers’ attitude towards the brand and the probability of revival. In this case, a decision where to live can be seen as ‘public’, people can observe where we live. Therefore, normative influence may play a role here. Furthermore, it is a complex decision and it may be difficult to forecast whether someone will be happy living at a certain place. Therefore, informational influence may play a role as well. I expect that social influence negatively moderates the relationship between attitude and the probability of revival. So, this relationship is weakened when lost customers are highly influenced by others.

(21)

2.5 Customer characteristics

Homburg, Hoyer and Stock (2007) use the definition of Kahn (1995) of variety seeking: “…the tendency of individuals to seek diversity in their choices of services and goods”. The strength of the variety-seeking motive varies across customers. Customers with a strong variety-seeking motive are more open to new brands and relationships. As a consequence, revival of a relationship is less likely for these customers with a strong variety-seeking motive as opposed to customers with weak variety-seeking motives. Homburg, Hoyer and Stock (2007) found evidence for this hypothesis. They found that the variety seeking motive has a negative impact on revival performance. I hypothesise this is also the case in this specific situation. People that have a strong variety-seeking motive are less likely to return to the Achterhoek, because they prefer living in a ‘new’, different area.

H5: Variety seeking negatively influences the probability of revival.

Involvement can be defined as “personal relevance or the extent to which a product has a direct bearing or consequence on a customer’s life”. Highly involved customers are more likely to have a strong motivation to process cognitive or affective information related to a specific product. Besides, they are more interested and search more intensively for product related information (Homburg, Hoyer & Stock, 2007). Homburg, Hoyer and Stock (2007) found that involvement positively affects revival performance. I propose that this finding also holds in the situation of returning to a region: people that are highly involved with the Achterhoek are more likely to return.

H6: Involvement positively influences the probability of revival.

Existing research has found that older individuals are more willing to engage into the continuance of an existing relationship than younger people. Age was shown to have a particularly strong positive effect on revival performance (Homburg, Hoyer & Stock, 2007). I propose that this is also the case in the situation of returning to the Achterhoek. After their graduation, students often want to live in a larger city in the Randstad. When they start thinking of children, they are more likely to return to the quiet area of the Achterhoek (at an age of 30 – 35 years old). After that, the probability of return probably decreases. Since only people till 35 years old will be part of the sample, I propose:

H7: Age positively influences the probability of revival.

2.6 Relationship characteristics

(22)

Stock: 2007). Homburg, Hoyer and Stock (2007) therefore argue that the probability to gain back customers who have had a long relationship with a company is higher opposed to those who have only had a short relationship. However, they do not find empirical evidence for this hypothesis. Thomas, Blattberg and Fox (2004) did find evidence for this hypothesis. Their reacquisition model showed that the probability of a firm reacquiring a lost customer is higher if the first tenure is longer. Hence, I propose:

H8: Length of the first relationship positively influences the probability of revival.

In addition, Thomas, Blattberg and Fox (2004) found that the probability of a firm reacquiring a lost customer is higher if the lapse duration is shorter. The longer the time since the last purchase, the more likely a lost customer is to have engaged a new service or to have developed new behaviors. However, this may not be the case for this specific region marketing situation. As mentioned before, people are more likely to return when they decide to ‘settle’; therefore it may be more likely that they will return after a longer period of absence. No evidence for this way of reasoning has been found yet. Therefore, I propose in line with the research finding of Thomas, Blattberg and Fox (2004):

(23)

3. RESEARCH METHOD

In this chapter, the research method that is used to answer the research questions will be discussed. I will perform a descriptive study. The major objective is to describe the customer revival process in more detail and identify variables that influence the probability of revival. A cross-sectional design will be used; information will be collected from a sample out of the population only once (Malhotra, 2007).

Since the Regio Achterhoek does not have any data available on (former) students that returned to the Achterhoek, primary data has to be collected. In order to collect these data, two types of research will be performed. First, interviews will be conducted. The answers on these interviews are used to formulate questions for a survey that will be hold.

3.1 Interviews

With respect to the concepts of ‘reason of defection’ in the situation of living and ‘customer revival strategy’ few existing literature is available. Therefore, I am not able to base the scales of these concepts only on existing literature. I decided to conduct qualitative, semi-structured interviews in order to create scales for these questions.

I will interview 15 (former) students that left the Achterhoek to study somewhere else. I tried to create a sample that is quite representative for the population of students and graduates that are originally from the Achterhoek and that are now studying or working in various large cities in the Netherlands. However, since judgmental sampling is used and since the sample is very small, the results probably are not generalizable to the population of students in general.

(24)

The data will be analyzed by coding and grouping. First, the interviews will be transcribed. Second, I will create several answering categories based on the answers of the interviewees. Thereby, I will mainly pay attention to those pieces of information that are important for the survey (especially reasons of defection, and various characteristics of the customer revival strategy). Subsequently, I will subdivide the various answers into the categories. After that, I will calculate how often every category of answers is mentioned and I will draw conclusions. The analysis of the answers is shown in Appendix IV. I will discuss the most important findings in paragraph 3.4 where the measurement scales are presented.

3.2 Survey

A survey will be used to analyze the different relationships as presented in chapter 2. The data obtained from a survey are reliable, since the responses are limited to the alternatives stated. Furthermore, coding, analysis and interpretation of data obtained through a survey is relatively simple. A disadvantage of using a survey is that respondents may be unable or unwilling to provide the desired information. They may not be aware of for example the reasons why they left the region or they have not thought about where to live in the future. Besides, the validity of certain constructs may be lower, because questions are structured and response alternatives fixed (Malhotra, 2007).

3.3 Sample

The target population of the survey consists of students and former students that are originally from the Achterhoek and currently live somewhere else. After people get married and have children, they are less likely to move over a large distance (Hospers, 2009). Therefore, I decided to define the target population as people who originally come from the Achterhoek; who left the region and moved to a place with a university or HBO-institution; and who are younger than 35 years old.

I expect that the existing attitude with respect to returning to the Achterhoek may differ as a result of a number of demographics: residence (current and former), level of education (HBO versus university), study, and year of study. Judgmental sampling will be used in order to include students living in different cities, doing different studies and being in different years of study within the sample.

(25)

The survey will be distributed via different channels: via e-mail, via the Achterhoek Netwerk on LinkedIn, via the website and Twitter-account of ECHT Achterhoek, and within the train between Winterswijk and Arnhem and between Winterswijk and Zutphen on Friday the 18th of March in the afternoon. A lot of students go to their parents’ home during the weekend, so this way, a lot of students can be reached. In this manner, I try to create a representative sample. However, I expect that mainly students that already have a positive attitude with respect to the Achterhoek are more likely to return to the Achterhoek in the weekends, and to visit the Twitter page and website of ECHT Achterhoek. In chapter 5, I will discuss this problem of generalizability in more detail.

3.4 Measure development and assessment

Multiple item scales will be used to measure each construct. Most items will be based on existing research and are slightly modified to fit the context of this study. Since few research is available in which scales are defined of ‘reasons of defection’ – especially in the case of leaving a region - and ‘customer revival strategies’, I base the scales on the interviews that I have conducted and that are discussed in paragraph 3.1.

Within the questionnaire, first questions about demographics will be asked: age, gender, level of education, current residence, study, year of graduation, and former residence. This way, it can be determined whether the sample is representative of the population. Furthermore, these demographics are used to describe the different classes of the latent class regression analysis.

(26)

Table 1: Constructs, scales and literature support

Construct Scale Literature support

Reasons of defection _- I wanted to live on my own - Because of my relationship

- Because I wanted to live in a different area - Because of my (future) work

- Because of my study

- I wanted to experience a real student life - Because I was up to something new - My friends also left the Achterhoek

The measurement method of Homburg, Hoyer and Stock (2007) is used. Respondents are asked to distribute 100 points according to the impact of the different criteria. These reasons are based on Feijten and Visser (2005) and on the results of the interviews.

Revival activities Interest in various characteristics of revival activities is measured:

• Frequency of contact (never, once a year, four times a year, once a month, once a week, the information should be always available): respondents should choose one answer.

• Content of contact (job vacancies, housing, sports clubs, companies, cultural events, internships, accessibility): asking for interest for every item using a 7-point Likert scale. • Channel of contact (news paper, flyer,

website, e-mail, personal letter, during events, personal conversation): asking for interest for every channel using a 7-point Likert scale.

• Timing of contact (high school, college, year of graduation, just after graduation, a couple of years after graduation): respondents can choose more than one answer.

• General interest in receiving information: asking for interest using a 5-point Likert scale.

The questions are based on the results of the interviews. In all cases, the option ‘something different’ will be added.

Attitude with respect to the brand

Four statements (7-point Likert scale): - I’m proud that I’m originally from the

Achterhoek.

- The overall lifestyle in the Achterhoek is better than in the rest of the Netherlands.

- I would prefer living in the Achterhoek than any other place after my graduation.

- The Achterhoek has a good reputation among students that are originally from the

Achterhoek.

The measurement scale of attitude with respect to the living area of Merrilees et al. (2009) has been adjusted to measure brand attitude of the Achterhoek. Originally, the measurement scale was developed to measure the attitude of the current living area. I adjusted it to make it possible to measure attitude with respect to the Achterhoek as future residence.

Social influence Eight statements (7-point Likert scale):

- It is important that others like the place where I will live

- When I will look for a place to live, generally this will be a place that I think others will approve of.

- I like to know what place makes a good impression on others.

- I achieve a sense of belonging by living at the same place where my relatives live.

- I identify with other people by living in the same area they live.

The scale of Bearden, Netemeyer and Teel (1989) of normative and

(27)

- To make sure I live in a nice place, I often observe relatives where they live.

- If I have little experience with a certain area, I would ask my relatives about it.

- I will gather a lot of information from friends or family before I will decide where to live. Probability of

customer revival

Customers have to indicate the likelihood that they will return (7-point Likert scale):

1. Unlikely – Likely 2. Improbable – Probable 3. Impossible – Possible 4. Uncertain – Certain 5. Definitely not – Definitely

The scale of Tokman, Davis and Lemon (2007) has been used to measure the likelihood of return.

Variety seeking Two statements (7-point Likert scale): - I really like to try new things - I am always searching for changes

The measurement scale Homburg, Hoyer and Stock (2007) of variety seeking has been adjusted. Involvement Two statements (7-point Likert scale):

- The environment where I live is important to me

- I am well informed about the latest developments with respect to living and working in different areas.

The measurement scale of

involvement of Homburg, Hoyer and Stock (2007) will be used.

Age Objective measurement of age. Length of first

relationship

Objective measurement of length of first relationship.

Time since ending Objective measurement of time since ending.

3.4.1 Reasons of defection

As discussed in the introduction, the research question of regaining lost customer will be applied on how to attract citizens back to the region where they were born. The construct of reasons of defection is based on research findings of Feijten and Visser (2005) and Rossi (1980) and on the results of the interviews.

In the interviews, I asked why the respondents moved. ‘Study’, ‘living in your own’, ‘possibilities to develop your self’, ‘experiencing a student life’, and ‘starting a new adventure’ were the reasons that were most often mentioned. Most of the reasons that are mentioned are similar to the categories as presented by Feijten and Visser (2005). However, I will add the categories of ‘experiencing a student life’ and ‘being up to something new’ to the questionnaire.

(28)

3.4.2 Revival strategy

I would have preferred to test the effectiveness of different revival strategies; however, lost customers have not been approached yet by the Regio Achterhoek. Thus, it is not possible to test which revival strategy and which channel of communication have been most effective in influencing customer revival. Therefore, I decided to test the effectiveness of these strategies by asking lost customers to what extent they would like to receive information. I base the different scales on the results of the interviews as discussed in 3.1.

I distinguish different strategies based on the content, channel, moment, and frequency of contact. First, I will ask in what type of information the respondents are interested (question 15). The following types of information were mentioned in the interviews:

- Information about jobs / traineeships (47%)

- Information about housing (27%)

- Offers of sports clubs, clubs for young people (7%)

- Information about companies (33%)

- Information about culture / cultural events (20%)

- No information (27%)

I will add ‘information about internships’ and ‘information about accessibility’ (transport) to this list. Transport plays a mayor role in the decision to leave the Achterhoek and may also play a role in the decision to return. Furthermore, students maybe want to do their final internships in the Achterhoek and subsequently stay in the area. I will use a seven-point Likert scale in order to determine which type of information the respondents would like to receive. This way, possibilities of analysis are increased in comparison to using a yes/no scale.

(29)

Fourth, I will ask how often and when the respondents would like to receive or read information. For both items, a closed question is used. The answering possibilities of how often the respondents like to receive information are based on the interviews: never / once a year / four times a year / once a month / once a week / this information should be always available. The respondents can tick one answer. I also based the answering possibilities of the question when the respondents want to receive information on the interviews: only the year of graduation / during career guidance on high school / a couple of years after graduation. I add two possibilities to this list: during college (so from the first year until the last year) and just after graduation. Some students may be looking for an internship and are therefore interested in receiving information about career opportunities before they are graduating. Besides, some students may start looking for a job just after their graduation and therefore would like to receive information about career opportunities at that moment. Respondents could choose more than one answer.

3.4.3 Interpersonal influence

Bearden, Netemeyer and Teel (1989) present a scale for measuring consumer susceptibility to interpersonal influence. They distinguish two types of interpersonal influence: normative influence and informative influence. They present an 8-item scale to measure normative influence and a 4-item scale to measure informative influence.

I have adjusted their measurement scale to the situation of moving to a place. In this situation, not all statements are suitable to replicate. I selected 8 items that were appropriate in the situation of moving to a place (3 items to measure informative influence and 5 items to measure normative influence). I will use a seven-point Likert scale so the respondent could indicate whether they disagree or agree with the statements.

(30)

outcome of the factor analysis – to replace the original set of variables, I will compute factor scores based on the average scores of the variables that are part of the factor.

3.4.4 Attitude

Traditionally, attitude of a consumer towards a brand is operationalized by measuring consumer perceptions of the ‘overall rating’ of the brand. I will use the adjusted brand attitude scale of Merrilees, Miller and Herington (2009) to measure brand attitude with respect to the region. They present a four item scale to measure brand attitude with respect to Gold Coast, consisting of the following items: proud to live in GC; overall lifestyle is good; rather live here than any other place; and good reputation among residents. The respondents were asked to what extent they agreed to the statements. I have adjusted the scale of Merrillees, Miller and Herrington (2009) to the situation of the Achterhoek. I will use a principal components analysis with varimax rotation in order to confirm the unidimensionality of the construct. Besides, I will make use of a Cronbach’s alpha measure to assess the consistency of the scale. In order to create only one attitude variable to replace the original set of variables, I will compute factor scores based on the loadings of all variables on the factor.

3.4.5 Probability of revival

Tokman, Davis and Lemon (2007) created a model with probability of revival as dependent variable. They used a six item 5-point semantic differential scale which was modified from Notani’s (1997) repurchase intentions scale. Tokman, Davis and Lemon (2007) used the following scale-items to measure the likelihood of return: unlikely – likely; non existent – existent; improbable – probable; impossible – possible; uncertain – certain; and definitely not – definitely. I will use the same scale to measure switch-back intentions. I will put the different items in a principal components analysis with varimax rotation in order to confirm the unidimensionality of the construct (Tokman, Davis and Lemon, 2007). In order to create only one variable to replace the original set of variables, I will compute factor scores based on the loadings of all variables on the factor.

(31)

area; it gives a feeling of being home; because of their job; because living is cheap in the Achterhoek; and because there are no traffic jams within the area. Reasons not to return are that there are no jobs; the respondents had a new social life and they preferred living in a larger city. I combined these answers with the findings of Leijenhorst et al. (2010). As mentioned in the introduction they found that students do not want to work in the Achterhoek because of its bad accessibility by public transport, lack of career opportunities or employment, no vivid cities or towns and its isolation. On the other hand, the Achterhoek is found to be attractive because of its peacefulness, space and nature.

3.4.6 Customer characteristics

Measurement scales of involvement and variety seeking are included in my questionnaire. These scales are based on existing research of Homburg, Hoyer and Stock (2007). I will use a principal components analysis with varimax rotation in order to confirm the unidimensionality of the constructs. I will use a Cronbach’s alpha measure to assess the consistency of the scales. In order to create only two variables (‘involvement’ and ‘variety seeking’) to replace the original set of variables, I will compute factor scores based on the average scores of the variables that are part of the factor. Besides, ‘age’ has been found to positively influence the probability of revival (Homburg, Hoyer & Stock, 2007). This finding will be tested by adding an objective measure of age to the questionnaire.

3.4.7 Characteristics of the first relationship

Length of the first relationship and time since relationship ending have also been found to have an influence on switch-back intentions. Therefore, I also include these two variables as objective measures in my questionnaire.

3.5 Pre-test

Prior to conducting the main survey, the questionnaire will be pre-tested and the questions will be adjusted. Ten people will answer all questions. Subsequently they are asked whether all questions and answering possibilities are clear and whether their interpretation is right. As a result of this pre-test several questions will be adjusted and some answering possibilities could be deleted.

3.6 Plan of analysis

3.6.1 Regression model

(32)

information timing, and information frequency; interaction variables of ‘reason of leaving the region * general interest in information’ and ‘attitude with respect to the region * social influence’; involvement; variety seeking; age; length of first relationship and time since ending are used as predictors. It will be tested whether these variables significantly influence the probability of revival.

I will test all hypotheses with this model. H1a-h will be tested by determining whether there is a significant positive or negative influence of the various reasons of defection on the probability of revival. H1i-p will be tested by analyzing whether there is an interaction effect between the reasons of defection and general interest in receiving information that positively or negatively affects the probability of revival. I will test H2a by comparing whether a more personal channel of communication (personal conversation, personal letter, e-mail) has a significant more positive effect on probability of revival than a less personal channel of communication (newspaper, flyer, website). H2b will be confirmed when the desired frequency of contact has a significant, positive effect on the probability of revival (so a higher frequency should more positively influence the probability of revival than less frequent contact). I will test H2c by analyzing whether a significant influence of the moment of contact on the probability of revival exists. The influence of a moment that is soon after defection should be larger than a moment that is not soon after defection. H2d will be tested by analyzing whether ‘general interest in information’ significantly positively influences the probability of revival. Besides, I test H3 by analyzing whether the attitude towards the brand significantly affects the probability of revival. I test H4 by including interaction terms between the different types of social influence and attitude in the regression model. H5, H6 and H7 are tested by analyzing whether the factors of involvement and variety seeking and whether age significantly affect the probability of revival. I test H8 and H9 by considering whether the length of the first relationship and the time since ending have a significant effect on the probability of revival. In order to test the different hypotheses, the t-values of the predictors will be considered.

3.6.2 Testing the assumptions in multiple regression analysis

Missing variables and outliers

Missing variables are coded as missing and are not included in the estimation of the regression model. Furthermore, extreme values are detected by analyzing the box plots and the residuals. When these outliers highly influence the results, they are deleted.

(33)

Several assumptions have to be made about the relationship between the dependent and independent variables that affect the statistical procedure (least squares) used for multiple regression. These assumptions are:

• Constant variance of the error terms: - the Levene’s test for homogeneity of variance is used to test for the presence of unequal variances – heteroscedasticity (Hair et al., 2010). Thereby, the residuals of the observations are randomly divided into two groups and the equality of variance for these two groups is tested.

• Independence of the error terms / autocorrelation: in order to test the independence of the error terms, a Durbin Watson test is used. When the error terms are independent, we expect them to be close to 2. Small values indicate positive autocorrelation, large values negative autocorrelation (Leeflang, Wittink, Wedel & Naert, 2000).

• Normality of the error term distribution: in order to test the normality of the error term distribution, a Kolmogorov-Smirnov test will be performed (Hair et al., 2010).

• Multicollinearity: multicollinearity can be detected by testing the correlations between all independent variables, and by checking the VIF-scores and Tolerance levels. When correlations are higher than 0.70, multicollinearity may be a problem. Furthermore, when VIF-scores are higher than 10 or Tolerance levels are lower than 0.10, multicollinearity problems are almost certain.

Model fit

The overall goodness of fit of the model will be measured by considering the adjusted coefficient of determination or adjusted R2, which measures the proportion of total variance in the dependent variable explained by the model (Leeflang et al., 2000). Furthermore, I will test the significance of the model as a whole by performing an ANOVA-test.

(34)

3.6.3 Latent class regression analysis

In order to get more insight into the variables that influence the probability of revival, a latent class regression analysis will be performed. This regression analysis will be based on the independent variables that were part of the ordinary least squares regression analysis. These variables are shown in table 2. The dependent variable is ‘probability of revival’. The complete list of inactive covariates is shown in table 26 in Appendix VI.

Table 2: Predictor variables latent class regression Value-expressive influence x Attitude

Value-expressive influence Attitude

General Interest in information RD: Work

RD: Study

RD: Friends also left the Achterhoek TI: A couple of years after graduation TI: Something different

Interaction General interest information x RD: Work Interaction General interest information x RD: Study

Interaction General interest information x RD: Friends also left the Achterhoek

The advantages of performing a latent class regression analysis over a traditional regression model (as discussed in 3.6.2) are that the traditional assumption that the same holds for all cases is relaxed; diagnostic statistics are available to determine the value of R2; and covariates can be included to improve classification of each case into the most likely segment. A potential drawback of a latent class regression model is that there is no guarantee that the solution will be the maximum likelihood solution (Magidson & Vermunt, 2011).

(35)

Criteria: AIC, AIC3, CAIC, BIC and LL. When comparing models, the lower the Information Criteria, the better the model. Andrews and Currim (2003) researched which of the information criteria has the highest segment retention success rate. They suggest to use the Akaike’s Information Criterion with a penalty factor of three (AIC3). Other researchers suggest to use CAIC (or BIC), especially with large sample sizes (Gensler, 2009).

Furthermore, the number of parameters that has to be estimated has to be taken into account. When too many parameters have to be estimated, overestimation may be the case. Besides, I will take a look at the cluster sizes; they should be substantial. I will compare one- to four-class models, since with a larger number of classes, the number of parameters that has to be estimated becomes too large and the cluster sizes become too small.

All these factors will be considered in the determination of the right number of classes. After a decision about the number of classes has been made, the final solution will be analyzed in more detail. It will be considered to what extent the regression model differs based on the independent variables. Thereby, the Z-scores will be considered. The Wald (=) statistic will be considered in order to determine whether the classes significantly differ based on the various dependent variables. When this is not the case, the model will be re-estimated and the effects will be restricted to be class independent. Besides, I will consider the R2 and the Classification Error that are measures of model fit.

(36)

4. RESULTS

In this chapter, the results of the research will be discussed. First, in paragraph 4.1 I will discuss the background of the sample of respondents. Subsequently, the general results will be presented in 4.2. In 4.3, I will discuss which transformations are performed so the results of the questionnaire can be used for further analyses. Subject of 4.4 are the results of the regression analysis and in paragraph 4.5 I will present the results of the latent class regression analysis.

4.1 Sample

A total number of 153 respondents completed the questionnaire. However, only 133 respondents were part of the target group (people who are originally from the Achterhoek, who now live outside the Achterhoek and who are younger than 35 years old). Another 16 people did not answer all questions, so the answers of 117 respondents could be used in all analyses. The questions that the respondents did not answer were marked as ‘missing’.

In Appendix VII, several descriptives of the sample are shown. On average, the respondents leave the Achterhoek when they are 18 years old, just after high-school. Most respondents graduated in 2010, so they already have made the decision where to live after their graduation.

The average distance to the Achterhoek (Doetinchem) is 97.52 kilometres. The distance of all cities to Doetinchem was measured using Google Maps (Google, 2011). Places abroad were not included in this calculation (four respondents lived abroad). The standard deviation is quite large of ‘distance to Doetinchem’ (57.59), so the distances vary a lot.

41.4% of the respondents is male, 58.6% is female. 1.5% of the respondents followed a study on MBO-level, 35.3 % on HBO-level, 61.7% on WO-level, and 1.5% a different study. This is probably not in line with the level of education of the complete population, since the percentage that follows a study on WO-level is very high.

Table 3: Municipality

Municipality Percentage questionnaire

HOW TO INFLUENCE CUSTOMER REVIVAL