Marketing effects on retention and cross-sell in the banking industry

Marketing effects on retention and cross-sell in

the banking industry

A case study at the Rabobank

Marketing effects on retention and cross-sell in

the banking industry

A case study at the Rabobank

Master Thesis

For completion of the master

Business Administration, specialization: Marketing Intelligence

at Rijksuniversiteit Groningen

Date: 11 July 2013

Author: E.C.R Hamelink

Studentnumber: 1768441

5

Management Summary

In the turbulent financial environment it is important for banks to allocate marketing resources in a more efficient way. By providing insights in the accountability of marketing efforts and the

profitability of customers, banking performance can improve. Different marketing efforts, like increasing product usage by online and mobile banking and the use of direct marketing campaigns to increase customer value are commonly used in the banking industry. Literature about the marketing mix is extensive; however literature about marketing effectiveness in the banking industry is limited. Therefore this study contributes to the scientific literature, by investigating whether these marketing techniques increase the value of the customer database by determining marketing effectiveness on customer retention and cross-sell.

Over the last few years the banking industry has moved from traditional banking towards a multi-channel approach whereby the increasing use of online banking stands central. This study shows that stimulating customer to use online and mobile banking increases retention and cross-sell significantly. Furthermore direct marketing campaigns on natural switching moments in the

customer relationship prevents them from churning. And campaigns whereby customers are

rewarded in form of e.g. cash rebate, gifts, discounts or certain types of preferential treatments, can increase financial results. These effects are influenced by several demographics and levels of

customer commitment. To capture these effect customer segments are created based on

6

Preface

I am proud to present to you my master thesis about marketing effects in the banking industry. I was quite determined to combine my master thesis with an internship to experience how it is to study a real-life management problem. I was advised to approach the marketing and communication manager of Rabobank Stad and Midden Groningen Ido van Veen, since his team was described to me as: a team of fantastic people. And that appeared to be absolutely true. I would like to thank them for sharing their knowledge and providing me with such a nice working environment.

I investigated marketing effects in the student market, since this is an interesting target group for the bank. The bank is as main sponsor of the Keiweek – the introduction week for new students in Groningen – interested in drivers of behavior of this specific customer segment. Even though the process of combining an internship and my master thesis was sometimes quite challenging I am very happy with the results. This thesis contributes to the scientific literature by providing insights at in financial services marketing, next to providing a more practical marketing application for financial services managers.

I would like to say some special thanks to Ido van Veen and Agatha Dooper for giving me this unique chance, their cooperation in defining an interesting research topic and their support. And thanks to thank Marlon Mols, marketer at Rabobank Stad and Midden Groningen, for sharing his insights and knowedge. Furthermore I would like to thank my supervisor prof. dr. J.E. Wieringa for his professional guidance, valuable feedback and insights. I would also like to thank my second

supervisor dr. ir. M.J. Gijsenberg for his additional feedback and sharp comments. I would also like to thank my parents, brother, sister and boyfriend for their ICT-related, grammatical and especially mental support.

7

Index

1.2 Context and relevance ... 10

1.3 Problem statement and research questions ... 10

1.4 Method ... 11

1.5 Outline of the chapters ... 12

2. Theoretical Framework ... 13

2.1 The need for accountability of marketing efforts ... 13

2.2 Dynamics of the banking industry ... 14

2.2.1 Characteristics of financial services ... 14

2.2.2 Consumer behavior ... 15

2.2.3 Technological developments ... 16

2.2.4 Competitive environment ... 17

2.3 Desired outcomes in the banking industry ... 17

2.3.1 Main objectives... 17

2.3.2 Customer retention ... 18

2.3.3 Cross-sell ... 18

2.4 Drivers of retention and cross-sell ... 19

2.4.1 Marketing efforts... 19

2.4.2 Demographic variables ... 21

2.4.3 Customer commitment ... 23

2.5 Conceptual model ... 25

3. Method ... 27

3.1 Choice of data and casus ... 27

3.2 Type of analysis ... 28

3.3 Model formulation ... 30

4. Results ... 33

4.1 Coding scheme ... 33

4.2 Description of the data ... 33

8 4.3.1 Model estimation ... 34 4.3.2 Validation ... 36 4.3.3 Discussion of coefficients ... 37 4.4 Study 2: cross-sell ... 40 4.4.1 Model estimation ... 40 4.4.2 Validation ... 40 4.4.3 Discussion of coefficients ... 42 4.5 Robustness checks ... 47 4.5.1 Heteroskedasticity ... 47

4.5.2 Application of the estimated model on location 8 ... 48

4.6 Latent class analysis ... 48

5. Discussion and implications ... 50

5.1 Main findings and contributions ... 50

5.2 Management implications ... 52

5.3 Research limitations and further research ... 53

9

1. Introduction

1.1 Preface

In the fall of 2008, a financial crisis has turned the economy into a recession. The shortcomings of the financial regulatory system are seen as the principal cause of that crisis. The financial regularity system was fragmented and antiquated, which allowed for some parts of the financial services industry to operate with little or no oversight. By being one of the pillars of economic prosperity banks hold a unique position within the economy. Banks generate opportunities for economic development by providing financial resources for investments and a secure place for savings. The events of 2008 exposed the vulnerability of financial firms whose business depended too heavily on the secured financial markets. Several banks collapsed, causing a drop in customer trust in the whole financial services industry. An article on Nu.nl, published on 26 march 2013, makes this very clear:

‘Consumers do not trust banks’

The biggest Dutch banks have reached an immutable position according to the consulting-agency. ‘When one of the players administers a change, the others react immediately in order to keep up’. This is going to lead to consumer distrust, according to Ronald Berger.

Practically all customers (98%) are currently not planning to switch bank. ‘They neither trust the quality of their own bank nor the competing banks’, says Ronald Berger. In the last year consumer trust in the financial services industry has decreased from 39 to 33 percent; the trust in banks decreased from 44 to 40 percent. The consultant remarks that the limited number of players in the banking sector causes differentiating abilities of banks to be limited. Furthermore the consultant notes the recent ‘pricing war’ around mortgages. ‘This constant competition-struggle, makes the highly needed changes impossible’.

(Source: www.nu.nl)

This article displays the drop in customer trust in the banking industry. This drop in customer trust highlights the urge to regain customer trust in order to survive the increasingly competitive and dynamic environment. Because customers face uncertainty and perceive a high level of risk, banks need to act in order to maintain or deepen customer relationships. They have to do this because in the financial services industry a healthy relationship between customer and supplier is impossible without trust (Wabeke, 2008). Retail banks have to work very hard to regain customer trust (El-Manstrly, Paton, Veloutsou, and Moutinho, 2011). By the effective use of marketing and communication customer relationships can be maintained and deepened (Berger and Nasr, 1998).

10 is needed and the pressure to allocate resources in a more efficient way in order to stay financially healthy is certainly present. The Rabobank’s marketing department wants to know how to make a more effective use of marketing in order to increase customer. Furthermore the bank is interested in cross-sell to increase the profitability of customer relationships. This study investigates the possibilities for the Rabobank by performing a case study at the bank at the accountability of marketing efforts in increasing retention and cross-sell.

1.2 Context and relevance

The accountability of the marketing department has received increased attention since it is a major driver of its influence within the business (Verhoef and Leeflang, 2009). The urge for marketing executives to speak in more financial terms has increased, necessary to gain internal support for marketing initiatives (McAlister, Srinivasan, Kim, 2007).

However measuring marketing accountability at financial services marketing is challenging. Currently a significant gap exists between the knowledge and the ability of financial services suppliers to effectively create and communicate customer expectations. This gap can only be managed if suppliers become more accountable towards customers (O’Loughlin and Szmigin, 2005). Changing from the traditional ‘gut-feeling’ approach in marketing, towards a more analytical approach in order to allocate resources over different marketing and communication instruments is proven to be successful (Wiesel, Pauwels and Arts, 2011). Especially in the financial services industry a strong relationship between marketing effectiveness and the firm’s performance is found (Hinshaw, 2005). This means that companies can increase the value of their customer base by using marketing programs to increase customer retention and cross-sell (Verhoef, van Doorn and Dorotic, 2007). The competitive nature of the marketplace forces banks to adopt a high strategic orientation to ensure that their marketing strategy is effective in generating retention and cross-sell (Appiah-Adu, Fyall, and Singh, 2001). Currently the financial services industry is confronted with a lack of information about brand and marketing performance and an absence of systems and programs to track it (Hinshaw, 2005). This study contributes to the literature by providing insights in the accountability of financial services marketing, in specific in generating retention and cross-sell in the banking industry.

1.3 Problem statement and research questions

This study provides insights in the accountability of marketing efforts in the banking sector based on the following problem statement: Which marketing efforts are most effective in increasing customer retention and cross-sell in the banking industry?

11 mostly based on the market for fast moving consumer goods. Research on the marketing mix is extensive – but research on marketing efforts in the banking industry is limited. These standards are not suitable for the financial services industry characterized by unique dynamics. This research aims to determine which marketing efforts fit these industry specific characteristics. For this reason this performs a case study at the Rabobank as a modest attempt to contribute to the empirical material.

The case study is conducted at the marketing and communication department of the Rabobank Stad en Midden Groningen. This bank is an interesting choice for several reasons. First, the financial service industry is known to collect data continuously on customer behavior in large datasets, making it possible to study actual customer behavior. Second, the financial services industry has been subject to several customer value management studies. Third, the Rabobank is one of the biggest financial services providers in the Netherlands and really customer-driven. Their mission is to be a life time partner of a customer. Lastly, and most important, the bank is open for ideas and constant developing to improve their marketing strategies and customer relationship management.

In this study the following research questions are answered:

1. Why make marketing efforts more accountable in the banking industry? 2. What are the characteristics of the banking industry?

3. What is the effect of marketing on customer retention and cross-sell in the banking industry? 4. What is the effect of demographic variables and customer commitment on marketing

effectiveness on retention and cross-sell in the banking industry?

5. Which marketing efforts are most effective for the Rabobank in generating retention and cross-sell?

6. How do these effects differ by demographics and levels of customer commitment for the Rabobank?

1.4 Method

12 data in order to develop analytical models explaining marketing effects on customer retention and cross-sell.

1.5 Outline of the chapters

A broad scientific literature study with regard to the concepts under investigation – marketing effects on retention and cross-sell in the financial services industry – is provided in the next chapter.

13

2. Theoretical Framework

This theoretical part discusses the accountability of marketing in paragraph 2.1. Characteristics of the financial services industry – in specific the banking industry – are discussed in paragraph 2.2. Paragraph 2.3 discusses the literature on customer retention and cross-sell in the banking industry. Paragraph 2.4 discusses drivers of customer retention and cross-sell. The main findings are summarized in the conceptual model in paragraph 2.5.

2.1 The need for accountability of marketing efforts

The role of the marketing department within a firm has been of debate in the past literature. The general reason the literature provides is the steep declining influence of the marketing department (Verhoef and Leeflang, 2009). Having a strong and influential marketing department contributes to the profitability of firms (Moorman and Rust, 1999). Several authors have found that the marketing department is underappreciated by people from in- or outside the organization (Leeflang, 2009; McAlister, Srinivasan and Kim, 2007; Verhoef et al., 2011; Mintz and Currim, 2012; Shaw and White, 1999; Verhoef and Leeflang, 2009). Some of the authors provide possible reasons for this underappreciating of the marketing department:

- Marketing effects are relative hard to measure (Verhoef and Leeflang, 2009);

- There is little coordination with other departments within the organization (Schultz, 2003); - There is a stronger focus on cost than on result (Schultz, 2003; McAlister, Srinivasan, Kim,

2007);

- The focus lies on short term success instead of long term results (Verhoef and Leeflang, 2009);

- And instruments measuring marketing effects are hardly used (Schultz, 2003 Verhoef, Leeflang e.a., 2011; Wiesel, Pauwels and Arts, 2011).

14 To allocate marketing resources in a more efficient way, the effectiveness of separate marketing efforts needs to be measured. Changing towards a more analytical approach to allocate resources over different marketing and communication concepts is proven to be successful (Wiesel, Pauwels and Arts, 2011). By including marketing mix variables in econometric models, managers can be better informed about marketplace behavior. A model which links firm strategy, metric orientation, type of marketing-mix activity and managerial, firm, environmental characteristics to marketing and financial metric use, is capable of measuring marketing performance. The use of these kinds of metrics is positively related to marketing mix performance (Leeflang, van Heerde and Wittink, 2002). According to Klein et al. (2003) a more analytical approach increases marketing effectiveness in several ways:

- Optimization of results out of marketing investments and defining an optimal marketing budget;

- Measuring the effectiveness of integrated marketing communication compared to the use of all instruments separately;

- Identifying weak points within the organization; - Being able to limit investments per customer/prospect;

- And providing insights in additional earnings for additional investments.

An organization capable of determining the results of separate marketing efforts holds an important advantage for securing improvements (Court, 2005). However many marketing departments have difficulties in becoming more accountable (Verhoef and Leeflang, 2011). Particular in the financial services industry there seems to be a lack of information about brand and marketing performance and an absence of systems and programs to track it (Hinshaw, 2005). Although techniques to evaluate financial return of particular marketing expenditures are available, a high-level model to trade off marketing strategies in the financial services industry is not (Rust, Lemon and Zeithaml, 2004). However a model for measuring marketing effects must be adapted to industry specifics since the financial services industry is characterized by unique dynamics. In the next paragraph the characteristics and dynamics of the financial services industry will be discussed in more detail.

2.2 Dynamics of the banking industry

2.2.1 Characteristics of financial services

15 simultaneously. The provision of relatively complex financial services can hardly be standardized and demands the employment of a large number of specialized personnel (Ennew and Binks, 1999).

Banks have been around quite a long time and are a crucial part of society by making economic traffic possible. What distinguishes banks from other service providers is the strong moral dimension of their services. Banks operate simultaneously in different time frames: immediate, short term, medium term and long term are the time realities of financial services. Customers are not simply judged on their ability to pay, but on their character. Other services providers, like airlines and cinema’s, provide a service regardless of the moral standing of the customer. Banks cannot afford to do this. Furthermore customers expect banks to vouch for their personal integrity to other who may suffer if the character reference is misleading. And governments require banks to support them in dealing with criminals who launder money or act fraudulently. Some customers deal with banks not because they want to but because a bank account might be required to participate in economic traffic (Lynch, 1994).

De Chernatony et al. (2004) developed a scale for measuring brand equity in the financial services industry. The scale includes three dimensions: brand loyalty, customer satisfaction and brand reputation. Brand awareness effects are not include because services are ‘credence goods’ – meaning that it are goods from which quality is hard to measure. Customers look at the reputation and image of the brand instead. They often perceive banks as basically generic in nature and similar to other ‘routine service providers’ like utility – i.e. gas and electricity – providers (O’Loughin and Szmigin, 2005). The challenge for the financial services providers is to identify and communicate unique, meaningful and desired images. The next paragraph discusses which elements drive consumer behavior.

2.2.2 Consumer behavior

16 Devlin (2004) identifies criteria influencing a consumer’s choice process in the banking industry from before the financial crises. His research reveals that the influence of recommendations, the offering of incentives when opening a new account and economic factors – like interest rates paid and fees – influence the customer’s choice. However these criteria need to be applied with carefulness, since banks who act unilaterally by providing high interest rates and fees might hurt their image of reliability leading to a reduction in account openings. Certain criteria have been staying important through time, like the bank’s image and reputation and location factors, such as choosing a bank close to home. However the financial crises might have changed customer drivers since then. Recent research by Chen et al. (2012) shows that fair service has an significant impact on customer satisfaction and plays a role equivalent to service quality in determining customer’s trust. Tailoring services to specific customer demand has become more important. Furthermore, consumer choice criteria develop along with technological developments, like the increasing use of online distribution channels. These days consumers desire a high degree of flexibility regarding the time and place of service consumption (Black et al., 2002). Distribution channels like the internet or the telephone are not merely accepted, but rather explicitly demanded (Walsh, 2002). The next paragraph discusses technological developments in the banking industry.

2.2.3 Technological developments

Customer expectations and their selection process are changing due to the shift from traditional branch banking towards online banking. Some authors highlight the importance of multi-choice channel approaches – incorporating technology as a complement instead of a replacement for personal channels – in the increasing automated environment of financial services (Porter, 2001; Lee, 2002). Hereby financial services suppliers will be able to profitably and successfully match and exceed customer expectations through offering a choice of service and delivery options which can be tailored to specific customer requirements (O’Loughin and Szmigin, 2005).

17 facilitating changing customer needs in an electronic service setting (Lee, 2002). The next paragraph discusses the competitive environment of the banking industry.

2.2.4 Competitive environment

Many banks have failed to create a superior positioning based on differentiation of their competitors (Romaniuk, 2001). The core image is difficult to differentiate from competitors due to the intangibility and lack of patentability (Devlin, 2005). Differentiation in the financial market place requires more balance between the inside-out strategic approach, with the customer as starting point within this balance. By understanding customer’s expectations and values, the company is forced to rethink its culture. Hereby companies would conduct their business in a radically different way leading to differentiation from competitors (Costanzo, Keasey and Short, 2003).

Next to these challenges in the financial services industry, there is a challenge to overcome the high switching barriers – in terms of time and money – perceived by customer (El-Manstrly, Paton, Veloutsou, and Moutinho, 2011). Especially in the financial services industry consumers face high switching barriers, since moving to a new services provider requires the investment of effort, time and money (Colgate and Lang, 2001). On the one hand, customer acquisition will increase when these switching costs are overcome with effective marketing programs. On the other hand, customer retention will increase when switching barriers are created (Gupta et al., 2004)

2.3 Desired outcomes in the banking industry

2.3.1 Main objectives

Rica (2012) names the following main policy objectives of banks: customer acquisition, customer retention, creating long-term relationships between customer and organization, and reducing customers fear related to the variability of supply and the intangibility of products.

18 2.3.2 Customer retention

By maintaining the business relationship between a supplier and a customer, customer retention has often been conceptualized and operationalized as a dimension of the customer loyalty construct (Boulding, Kalra, Stealin and Zeithaml, 1993; Zeithaml, Berry and Parasuraman, 1996). Retention rates can be defined by ‘the chance that the account will remain with the vendor for the next purchase’ (Jackson, 1985). In this study, retention is defined as customers staying with their current bank. As a mean to this end bank accounts should match customer’s personal situation. Focusing on customer retention creates ‘a stable pool of customers for a firm’s products or services’ (Oliver, 1997) and increasing customer retention rates lead to a steep climb of profit (Reichheld, 1993). Reichheld and Sasser (1990) share this vision by showing that retention is valuable, as the costs of winning a new customer is way higher than the cost of retaining a customer.

Retails banks aspiring to increase customer retention by means of CRM, need to be familiar with the drivers that help to retain a customer. Two main drivers of customer retention are described by Gustafsson et al. (2005), namely: customer satisfaction and customer commitment. Customer satisfaction is by Fornell et al. (1996) defined as: ‘the overall evaluation by the customer of the performance of an offering to date’ and is supposed to positively influence customer loyalty (Fornell et. al 1996). Commitment can be defined as ‘the desire to maintain a relationship’ (Moorman et al., 1993) and is found to have a direct influence on customer retention (Verhoef, 2003). While satisfaction involves cognitive and backward looking evaluation of the relationship, commitment contains the effective and forward-looking evaluation of the relationship (Bolton, Lemon and Verhoef, 2004). Section 2.4 discusses the drivers of retention in more detail. The next paragraph discusses literate on cross-sell in the financial services industry.

2.3.3 Cross-sell

Since the late eighties, the concept of cross-sell has received much attention within the banking industry. The perspective on cross-sell has moved from selling as much products as possible towards a more customer oriented approach whereby fulfilling customer demand in order to create customer satisfaction is the key goal (Bergendahl, 1995).

19 (2004) define cross-sell in the service industry as: ‘customers can add services to their portfolio which are little, if any, related to the current customer service’. Considering the definitions provided by literature, this study defines cross-sell as: ‘The practice of selling additional product(s) and/or service(s) that are related or unrelated to the product category of the initial product/service’.

Bergendahl (1995) names a number of benefits cross-sell can provide to an organization, namely: increasing customer’s profitability, creating exit barriers, lower acquisition costs of new customers and lower marketing costs. In addition, cross-sell can create enduring relationships between a customer and a firm by signaling quality and educating customers about the scope of products the organization offers and how these products meet their long-term financial needs (Li, Sun and Montgomery, 2011). However, research shows that cross-selling has not shown substantial results over the last twenty years. According to Malms and Schmitz (2011) the reason for failing cross-sell strategies is the fact that the focus is mainly on technology to improve cross-sell, instead of focusing on customer processes. Lymberoploulous et al. (2004) found in their research at cross-sell in the Greek banking industry, that customers are unaware of the products the bank provides. Customer life time value in cross-sell campaigns is usually treated as another segmentation variable to differentiate profitable customers from unprofitable ones. The problem with this approach is that the bank’s intervention changes the customer’s future purchase probabilities and thereby customer’s profitability (Rust and Verhoef, 2005). This requires the marketer to have a long-term view and to generate dynamic offers in accordance with the customer’s evolving financial status and preferences (Li, Sun and Montgomery, 2011). A distinction between drivers of cross-sell is made in terms of internal and external drivers. This study focuses on internal drivers, since firms are not directly able of managing external drivers. The internal drivers are further divided in corporate related and marketing related drivers. This study investigates the marketing related drivers. The lack of studies on the determinants of cross-sell clearly shows that more research is needed (Verhoef, van Doorn and Dorotic, 2007) therefore the next section describes these determinants.

2.4 Drivers of retention and cross-sell

2.4.1 Marketing efforts

Verhoef, van Doorn and Dorotic (2007) state that the effect of marketing instruments on retention and cross-sell is hardly studied. Since there is lack of research at the effectiveness of specific marketing instruments in the financial services industry (Hinshaw, 2005), this paragraph provides some guidelines on marketing and communication concepts instead of recommending specific instruments.

20 Ries, 2003). Marketing and communication programs should engage in meaningful research and dialogue with consumers to identify key financial brand values and appeals (O’Loughin and Szmigin, 2005). The marketing efforts should include positive features of the bank and its services, culminating in an exhortation to pass on the good news – creating word of mouth effects (Devlin and Gerrard, 2004). If customers see that a bank cares about their problems and is showing effort to solve their problems, customers become even more loyal than if the problem has not occurred in the first place (Lewis and Spyrakopoulos, 2001). Furthermore the introduction of programs whereby rewards – in form of e.g. cash rebate, gifts, discounts or certain types of preferential treatments – are granted to existing customers in case of recommending new customers can increase financial results (Devlin and Gerrard, 2004). CRM strategies increase the firm’s customer base by allocating more marketing resources to high-CLV customer while spending less on low-CLV customers (Shah and Kumar, 2008).

A positive relationship between marketing efforts and customer retention is expected for several reasons in this study. Marketing actions influence customer-based brand equity – including differentiation of their image about the company, relevance of the product or service, esteem in and knowledge about the company – which in turn influences customer retention (Stahl, Heitmann, Lehmann and Neslin, 2012). Rust, Lemon and Zeithaml (2004) join this vision by seeing marketing as an investment to improve customer equity. They found that marketing improved customer perceptions, leading to improved customer satisfaction resulting in increased customer acquisition and retention (Rust, Lemon and Zeithaml, 2004). Literature about marketing effects on retention in the banking industry is limited. The finding by Stahl et al. (2012) and Rust et. al (2004) at high-involvement consumer goods might suggest similar effects to be found in the financial services industry. In line with these findings, marketing efforts in the banking industry are expected to have a significant positive impact on the level of customer retention. Therefore, this study hypothesizes: : Marketing efforts are positively related to consumer retention

21 drive cross-sell. In line with these findings, marketing efforts in the banking industry are expected to have a significant positive impact on cross-sell. Therefore, this study hypothesizes:

: Marketing efforts are positively related to cross-sell

The next paragraph discusses the effect of demographics on retention and cross-sell. 2.4.2 Demographic variables

Demographics are commonly used in marketing to discriminate between customers (Zeithaml, 1985). Given that demographics are readily available, cost-effective and demographic based models can quickly be developed and deployed to target customer with customized marketing programs (Leeflang and Wittink, 2000). Retention is influenced by demographics such that customer groups differ in perceiving switching costs (Tesfom and Birch, 2011) and customer’s willingness to purchase additional financial products differ by demographics (Lymberoploulous et al., 2004). Furthermore demographics are found to determine the success of cross-sell strategies within public and private banks (Vyas and Math, 2006).

Specifically studies including the demographics age and gender show significant effects on consumer behavior (Zeithaml, 1985; de Moura Engracia Giraldj and Ikeda, 2009; Carter, 2010; Naseri and Elliott, 2011). Age is a simple, yet critical demographic variable since customer perceptions and behavior differ per age category. For example younger adults use credit cards significantly more (Mathur and Moschis, 1994) and are less concerned with honesty and reliability than older customers (Stafford, 1996). Gupta and Basal (2011) state that younger customers perceive banks as secure and efficient, while older customers perceive banks as reliable and responsible. Tesfom and Birch (2011) conclude that the younger the bank customer, the less s/he perceives relational benefits, switching costs and efforts of service recovery as switching barriers and the more likely s/he is to perceive the availability and attractiveness of alternatives. Thus, younger customers are more likely than older customer to end their relationship with their present bank, regardless of the time frame. These findings are in line with the argument of Colgate and Lang (2011) that those customers who have seriously considered moving to another bank tend to be younger than those who did not seriously considered moving to another bank. In line with these findings, customer retention is expected to be higher for older than for younger customers. Therefore, this study hypothesizes: : The demographic variable age is positively related to customer retention

Furthermore levels of cross-sell might differ by age categories. Since young customers are expected to evaluate more alternatives, this study expects that cross-sell to younger customer is more challenging than to older customer. Therefore, this study hypothesizes:

22 Since younger customers are expected to be less sensitive to switching barriers, it is important for bank marketers to be sensitive to different age groups of customers when developing marketing and advertising strategies (Stafford, 1996). Since younger customers perceive less relationship benefits, we expect that marketing efforts are less effective on retention and cross-sell for younger customers. Therefore this study hypothesizes:

: The effect of marketing efforts on consumer retention increases as age increases

And,

: The effect of marketing efforts on cross-selling increases as age increases

Despite the need for extra research at the effect of the demographic variable gender in the banking industry, customizing marketing programs based on gender might be profitable. Recently, the female’s involvement in financial decisions in households has increased. Overall, service quality seems more important for females than for males. Research shows that females respond positively to advertising appeals that emphasize reliability and honesty. If these aspects are included in marketing efforts strong attitudes towards the bank can be created (Stafford, 1996). Furthermore females rate bank performance higher than males on all dimensions except reliability where both men and females have given equal performance ratings. Overall the performance perceptions of banks are higher perceived by women (Gupta and Bansal, 2011). These finding suggest that females are more emotionally involved with their present bank. This study therefore expects women to be more loyal towards their bank than men. In line with these findings this study hypothesizes:

: Retention is higher for females than for males.

Furthermore a variety of availability of services is more meaningful for females than for males (Stafford, 1996). This study expects that females might be more tended to cross-sell than males. Therefore this study hypothesizes:

: Cross-sell is higher for females than for males

Women perceive a higher risk in buying financial products than men and respond more positively on advertising and marketing campaigns creating customer trust and commitment (Stafford, 1996). This study therefore expects that females might are more sensitively towards marketing than men. This study hypothesizes:

: The effect of marketing efforts on consumer retention is higher for females than for

males And,

: The effect of marketing efforts on cross-sell is higher for females than for males

23 process. Furthermore research by Hong Jea, Hong Won and Kwak Yoon (2010) reveals that the network centrality and the closeness centrality have a significant impact on the location of banks. Nevertheless there is no available research about the discrepancies between local and non-local customers in the banking industry. This study expects that the level of retention and cross-sell is higher for non-local than local customers, since people who live further away but are still clients have made and kept a more deliberate choice instead of simply choosing the closest bank. They show more commitment and are hence expected to be more loyal and more ‘active clients’. Therefore this study hypothesizes:

: Retention is higher for non-local customers than for local customers

And,

: Local customers cross-sell more than non-local customers

Since non-local customers are expected to be more involved with the bank this study expects them customers to be more sensitive to marketing efforts of the bank then local customers. This study hypothesizes:

: The effect of marketing efforts on consumer retention is higher for non-local than for

local customers And,

: The effect of marketing efforts on cross-sell is higher for non-local than for local

customers

The next paragraph discusses the effect of customer commitment in the banking industry. 2.4.3 Customer commitment

An increasing number of organizations determine the value of their customer database based on CLV in order to develop retention strategies. By increasing loyalty a strong brand demands less marketing support over time. Through measuring past customer loyalty CLV can be determined (Bijmolt et al., 2010). Only a limited number of studies have considered commitment as an antecedent of customer retention (Verhoef, van Doorn and Dorotic, 2007).

24 (Ganesh, Arnold and Reynolds, 2000; Dick and Basu 1994; Guiltinan, 1989). Hence, stayers who lack experiential knowledge are more likely to remain loyal to the service provider, even under conditions of dissatisfaction (Oliva, Oliver and MacMillan, 1992). Kamakura et al. (2003) find a positive association between the number of products owned by the customer and retention.

These finding suggest that when customers have proven to be loyal and face high switching barriers they are more likely to stay and/or become profitable to the organization. These switching costs are measured by the number of products the customer is committed to at the organization and thus more (active) loyal. This study assumes that the higher the number of products the customer purchases at the organization the more likely s/he is to retain as a customer and to cross-sell. Therefore this study hypothesizes:

: The number of products the customer purchases at the organization is positively related

to customer retention And,

: The number of products the customer purchases at the organization is positively related

to cross-sell

Since research shows that strong brands demand less marketing support as customer loyalty and commitment increases (Bijmolt et al., 2010), customer are expected to become more sensitive towards marketing efforts as they are becoming more loyal. Therefore this study assumes that the effect of marketing is stronger for customers who purchase a higher number of products customer at the organization. Therefore this study hypothesizes:

: The effect of marketing efforts on consumer retention increases as the number of

products the customers purchases at the organization increases And,

: The effect of marketing efforts on cross-selling increases as the number of

products the customers purchases at the organization increases

Furthermore the length of the customer relationship might indicate (passive) customer loyalty and commitment (Bijmolt et al., 2010). Customer commitment can be defined as ‘the desire to maintain a relationship’ (Moorman et al., 1993). Commitment is found to have a direct influence on customer retention (Verhoef, 2003). In line with these findings this study assumes that the longer the customer’s relationship with the organization the more likely the customer is to retain as a customer and to cross-sell. Therefore this study hypothesizes:

: The length of the customer’s relationship with the organization is positively related to

25 : The length of the customer’s relationship with the organization is positively related

to cross-sell

Furthermore, this study expects that the longer the customer relationship with the organization the more sensitive the customer is to marketing efforts:

: The effect of marketing efforts on consumer retention increases as the length of the

customer relationship with the organization increases And,

: The effect of marketing efforts on cross-selling increases as the length of the

customer relationship with the organization increases

In the next paragraph the main findings will be summarized by providing the conceptual model as a foundation for the empirical part of the research.

2.5 Conceptual model

This section summarizes the main findings of the theoretical framework by providing the conceptual model. The conceptual model displayed in Figure 1 includes the expected relationships based on the discussed literature. These relationships will be tested in the empirical part of this study.

As explained in paragraph 2.4.1 marketing efforts are expected to positively influence customer retention and cross-sell, displayed by respectively and in Figure 1. Furthermore paragraph 2.4.2 states that demographic variables influence these effects. First, the demographic age is expected to be positively related to retention ( ) and cross-sell ( ) and the effect of marketing

efforts on either retention and cross-sell is expected to increase as age increases (respectively

and ). Second, the demographic gender is expected to influence these relationships as well:

females are expected to have a higher retention ( ) and cross-sell ( ) rate than males, and are

expected to be more sensitive to marketing efforts ( and ). Third, discrepancies between local

and non-local customers are expected as: Non-local customers are expected to have a higher retention ( ) and cross-sell ( ) levels and to be more sensitive to marketing efforts than local

customers ( and ). Next to the demographics, levels of customer commitment are expected

to influence retention and cross-sell, as discussed in paragraph 2.8. The number of products the customer purchases at the bank is expected to be positively related to retention and cross-sell (

and ) and the effectiveness of marketing efforts is expected to increase as the number of

products the customer purchases at the organization increases, displayed by and in Figure

1. In addition the length of the customer relation is expected to be positively related to retention ( ) and cross-sell ( ). And the effect of marketing efforts is expected to increase as the length of

27

3. Method

The empirical study at Rabobank Stad en Midden Groningen provides insights in research question ‘Which marketing instruments are most effective in increasing customer retention and cross-sell in the banking industry? ‘.The choice of data and the casus are discussed paragraph 3.1. The type of analysis is explained in section 3.2. Paragraph 3.3 provides the mathematical model.

3.1 Choice of data and casus

The main benefit of using a firm’s database for doing customer research is avoiding the main problem of traditional market research: a non-representative sample. By using the customer database of a firm, information about the complete customer population is available. Another benefit is the availability of cross-sectional data with a longitudinal character, making it possible to research evolution over time. Analytical techniques use these data in customer relationship management for cross-sell strategies like market basket analyses or retention strategies like survival distribution functions (de Pelsmacker and van Kenhove, 2006). This research is based on customer data provided by Rabobank Stad en Midden Groningen.

The Rabobank is a Dutch bank consisting of 139 autonomous corporations. The bank is part of the Rabobank Group and is the biggest financial services provider in the Netherlands. They want to be at the heart of society, nearby and trendsetting in their service offering – and aim to be a driving and innovating force, contributing to the sustainable development of prosperity and well-being. Its goal is to help people and communities achieve their present and future ambitions, by strengthening mutual collaboration and supplying the best possible financial solutions (www.rabobank.nl). They want to be a lifelong, personal and financial partner of their clients. To build and maintain lifelong customer relationships the bank needs to offer financial services matching customer demand over time.

28 someone who has become a customer before his eighteenth is seventeen percent more profitable than someone who has become customer at a higher age.

Figure 2 Why investing in students pays off. Source: Onderzoeksrapport Jonge klanten; investeren omdat het loont. Marktmanagement, M&S Particulieren, Rabobank Nederland

Because of the high potential of this customer segment this research uses data of customers who are indicated as students (n=16.720) in the customer database of the Rabobank Stad and Midden Groningen at the time of March 2013. An analytical model will be developed to answer Rabobank’s management problem: ‘Which marketing efforts are most effective in increasing customer retention and cross-sell at the Rabobank’? .

3.2 Type of analysis

29 Logistic regression is undoubtedly a widely used bivariate dependence technique and highly suitable for retention and cross-sell analysis. Logistic regression analysis does not face the strict assumptions of multivariate normality and equal variance (covariance matrices across groups) assumptions on which discriminant analysis relies. Furthermore it is similar to multiple regressions, having straightforward statistical tests making estimation and interpretation of results less complicated (Hair, Black, Babin and Anderson, 2010).

Retention is traditionally seen as a binary issue aimed to predict whether the customer is likely to retain or defect during a pre-given time period (Bijmolt et al., 2010). Several authors provided an overview of the binary choice models that were used by academics and practitioners in modeling retention (Neslin et al., 2006; Verhoef et al., 2007; Bijmolt et al., 2010). These models include logistic regression analysis, decision trees and discriminant analysis (See Kamakura et al., 2005 for an overview). For example Verhoef, Langerak and Donkers (2007) considered the effect of brand equity, prior ties and switching costs on retention by applying logistic regression analysis. Furthermore Donkers, Verhoef and de Jong (2007) apply logistic regression analysis in a parametric binary prediction model to explain customer retention. Next to the use of analytical models on retention, cross-sell models can contribute to the allocation of marketing resources.

Li, Sun and Wilcox (2005) build on the idea that customers have predictable life cycles and, as a result, buy certain products at certain points in time. In order to predict cross-sell in the banking industry, they developed a multivariate profit model. In this approach cross-sell is seen as a binary issue, aimed to predict whether the customer reacts positively (by buying additional products and/or services) on a marketing campaign (Bolton, Lemon and Verhoef, 2007; Ngobo, 2005). Related to this Knott, Hayes, and Neslin (2002) apply a logistic regression model for predicting which product costumers are going to buy next. Bolton, Lemon and Verhoef (2007) counter-intuitively show that poor service quality may actually positively affect upgrading. Ngobo (2005) also applied logistic regression in order to investigate antecedents of service upgrading.

30 and their sensitivity to marketing efforts, marketing resources can be allocated more efficiently and marketing outcomes can improve.

3.3 Model formulation

The hypotheses discussed in chapter two – summarized in Table 1 – are tested by logistic regression. The model includes the independent variables: marketing efforts, age, gender, locality, number of products and length of relationship. The independent variable length of relationship displays the length of the customer-business relationship from the moment the customer is segmented as student in the customer database, making it possible to compare profitability between new and current customers. The independent variable age displays the age in years of customers from the point of becoming a customer of the Rabobank. The independent variable age is not showing much fluctuation in the data, since all customers in the population are of similar age. Therefore it might not be possible to measure age effects in an adequate way. The variable locality is measured by the city the customer is registered. The independent variable city is displayed by a dummy variable indicating whether or not the customer lives in the area the bank operates (the places Groningen, Hoogezand, Haren, Veendam and Siddeburen). Another variable included is the amount of income deposited on the bank account, e.g. student loans, as an indicator for the level of use and commitment to the bank. Furthermore control variables on the channel used to distribute marketing (paper mailing, telephone and email) are added to the model to investigate which distribution channels are most effective in creating retention and cross-sell.

Table 1 Hypothesis

independent/dependent Retention cross-sell marketing-> retention marketing-> cross-sell marketing efforts + + NA NA age + + + + Gender - - - - Local - - - - # of products + + + + length relationship + + + +

31 predominantly affected by company behavior on the customer relationship (Bolton, Lemon and Verhoef, 2004).

First, the logistic model on the dependent retention is formulated as: ( ) [∑ ] [∑ ] [∑ ] [∑ ] [∑ ] [∑ ] Whereby: ( ) , ( ), ( ) ( ) ( ) ( )

32 Whereby: ( ), , ( ), ( ) ( ) ( ) ( ) ( )

The logistic regression analysis generates a general model, wherein the significance of parameters shows the existence, strength and direction of effects (Hair, Black, Babin and Anderson, 2010). In order to validate these findings, problems with multicollineairity – the extent to which a variable can be explained by other variables in the analysis (Hair, Black, Babin and Anderson, 2010) – are checked and solved if needed. The different models will be compared and validated by the use of e.g. top decile lifts and BIC. This main validation measurement corrects for the number of variables included in the model, making it possible find a model which explains as much variance in the data while maintaining a model as simple as possible.

33

4. Results

This section discusses the results of the case study at Rabobank Stad and Midden Groningen. The coding scheme is discussed in section 4.1. A description of the data is provided in section 4.2. Study 1 in section 4.3. discusses the parameter estimates of the logistic model on retention. And study 2 in section 4.4, discusses the parameters estimates of the cross-sell models on respectively saving, paying and consumptive loan. Section 4.5 performs some robustness checks to test how well the model performs under changing circumstances. Section 4.6 performs a latent class analysis to identify homogeneous customer groups.

4.1 Coding scheme

The dependent variable retention is displayed by a dummy variable indicating whether the customer retained (=1) or churned (=0). The cross-sell dependents, are displayed by dummy variables which indicate whether the customer did (=1) or did not (=0) have the product. The dataset includes five independent direct marketing variables: the uses of online and mobile banking by the customer id displayed by M1 and M2 respectively, the direct marketing campaigns – ‘Becoming a Student’ and ‘Graduating’ – are displayed by M3 and M4 respectively and M5 indicates whether the customer subscribed at the Keiweek and in return received the tuition fee for the introduction week. Furthermore dummy variables indicate gender (male=1 and female=0), locality by whether the customer is registered in the area the bank operates (=1) or in another city (=0) and the control variables channel – paper mailing, telephone and email – displaying whether the channel is (=1) or is not (=0) used to distribute marketing and communication efforts. The independent age displays the age of the customer at the moment of starting the customer-business relationship with Rabobank Stad and Midden Groningen. And the independent length relationship displays the length of the relationship at the moment the customers is indicated as ’student’ in the customer database.

4.2 Description of the data

Logistic regression analysis does not assume a linear relationship between the dependent and independent variables, the dependent variables do not need to be normally distributed. There is no homogeneity of variance assumption, in other words, the variances do not have to be the same within categories, normally distributed error terms are not assumed and the independent variables do not have to be interval or unbounded (Blattberg, Kim, Neslin, 2008). However, exploring the data provides some insights in the nature of the data and its underlying relationships.

34 database at the age of seventeen, whereby the average length of the customer relationship is ten years. About half (50.9%) of the respondents live in a city in the operating area of the bank. And almost all (99.2%) of the respondents live in the Netherlands. International students are not of key interest to the bank since the retention rate is expected to be very low since they most likely return to their homeland when finishing their study in the Netherlands. Since the percentage of people living abroad is so small the variable city will be used to distinguish between local and non-local customers.

The direct marketing campaigns range over the last two years. Respectively 1.5% and 26.6% of the customers are the database is approached by the direct marketing campaigns ‘becoming a student’ and ‘graduating’. The respondents indicated their preference way of contact, positively answered for the channels by: 73.1% on paper mailing, 72.1% on telephone and 70.3% on email, whereby more answering options were possible. The retention rate among the customers in the database is 62.6%, thus 37.4% of the customers churned.

When looking at the cross-sell products, 47.9% of the respondents have the service ‘paying’, 45.9% ‘savings’, 1.7% ‘shares’, 22.3% ‘consumptive loan’, 0.8% ‘mortgage’, 17% ‘damage insurance’ and 0.6% ‘life insurance’. The products purchased by more than 20% of the respondents – paying, saving and consumptive loan– are modeled. The variable ‘number of products’ includes products purchased by at least 10% of the customers.

4.3 Study 1: retention

4.3.1 Model estimation

This section constructs the retention model by estimating the parameters of the earlier mentioned independents and interactions. The logistic regression applied by generalized linear models on dependent retention (binary: 0=exit, 1=stayed) is analyzed for all locations of the bank on customer level.

35 p<.005), M1xlocal (expβ=2.594, p<.000), length relation (expβ=7.47E33, p<.000), m1xnroducts (expβ=2.879, p<.000) and m2xnproducts (expβ=1.472, p<.05).

The locations of the bank – twenty-four locations in the operating area – highly differ in the number of customers. Applying one model with the same parameters estimates for all of the locations might bias the outcomes. Location 9 is by far the main location of the bank containing 63.9% of all the observations and thus customers. Bank location 8 also contains 11% of the customers, while many of the other location are small service offices containing less than 2% of all customers in the database – as displayed in Figure 3.

The model is estimated for location 9 which is by far the main location, preventing results from becoming biased by type of banking. At the end of our analysis it will be checked whether this model is also applicable for location 8 in the section 4.8 robustness checks.

Multicollineairity is expected to bias the estimates since the model showing high R-squares only finds a very limited number of significant parameter estimates. High correlations among explanatory variables causing multicollineairity is a common problem when estimating linear or generalized linear models, including logistic regression, leading to unreliable and unstable estimates of regression coefficients (Blattberg, Kim, Neslin, 2008). A linear regression analysis applied to detect multicollineairity –VIF-values >7 and strong multicollineairity at VIF-values>10 – shows many high VIF-values. The control variables for channel – paper mailing (VIF=7.47), telephone (VIF= 6.70) and email (VIF=3.16) – showing medium high VIF-values are of no concern since the VIF-values are not critical and the strength of effect is not the main interest of this study. The multicollineairity of the interaction terms –many VIF-values >200 – is caused by the fact that x and z are very likely to correlate with their product x*z, causing p-values to be unrealistic high. In the retention model including many binary independent variables displayed by either 1 or 0, information is lost by computing the interaction effects since many of the effects are multiplied by zero. The multicollineairity between the interactions is tried to be solved by recoding the dummy variables by 1 for positive outcomes and -1 for negative outcomes. However the recoding did not solved multicollineairity of interaction effects. To find reliable parameter estimates for the main interest of

0 5000 10000 15000 0 1 3 4 5 6 7 8 9 10 11 12 13 15 17 19 20 21 22 23 24 25 28 31 99 N u m b e r o f c u sto m e rs

Location of the bank

36 this study, the marketing effects, the interaction terms are excluded from the logistic regression analysis. The variable age is excluded since it correlates highly with length relationship – respectively VIF=34.34 and VIF=34.35 –because of the way there are computed. The variable age is hard to measure since age barely fluctuates between students. The marketing variable M1 (online banking) correlates highly – VIF= 9.58 – with M2 (mobile banking) since customers can only use mobile banking when having online banking as well. The variables are recoded as: one variable for only online banking – M1recoded – and one for having both – equals M2. Hereby the VIF-value of M2 (VIF=8.59) slightly increases but not critical and the VIF-value of M1 decreases to a non-critical value (VIF=4.97).

The interaction effects are investigated by estimating separate models for each of the moderating variables. The overlap in confidence intervals is used to check whether the coefficients per interaction effect are statistical significantly different. The intervals can actually overlap by as much as 25% and still be statistically significant, so by this method one can only confirm statistically difference if the confidence levels do not overlap. If the intervals do not overlap then one can be at least 95% confident there is a difference between the coefficients of the models. If there is a large overlap, then the difference is not significant (at the p <.05 level). The moderation effect of gender is tested by separating the model between females and males, of locality by separating the model by people customers live in the operating are of the bank and people who do not, of number of

products by separating the model for customers who have none and customers who have 1 or more of the main products and the moderating effect of length relationship by separating the model between new customers and customers who have been longer with the bank at the moment of becoming student. The next section discusses the validation of the computed model.

4.3.2 Validation

A logistic regression analysis was conducted to explain retention for the model as estimated in the previous section. To test whether all variables add value to the model they are stepwise included in blocks to compare model fit – as displayed in Table 2.

Table 2 Change in model fit

block of variables -2 Log

Likelihood

Cox & Snell R-square Nagelkerke R-square Marketing variables 7.600.740 .439 .606 Demographics 7.600.633 .439 .606 Relationship variables 7.449.788 .447 .617 Control variables channel 5.827.248 .525 .724

37 complete and the interaction models. A test of the full model against a constant only model was statistically significant for the complete model, indicating that the explanatory variables as a set reliably distinguish between churners and retained customer (chi square= 7940.946, p<.ooo with df=13). The BIC (3452.477) shows that the model fit has improved compared to the starting model of paragraph 4.3.1.

The lift curve in Figure 4 illustrates whether the model is able to identify groups of customers with higher probabilities of retention than other groups (Neslin et. al, 2006). A random model would lead to equally high probabilities of retention in each group. A good model would contain much higher probabilities in top ranked groups, leading to ‘lifts’ higher than 1 in those groups. When summing up these lift ratios, the so-called cumulative lift curve shows that the model estimates probabilities better than a random model since it lies above the base line, though is not very steep yet.

The top decile lift can be calculated as the percentage of retained customer in the top decile divided by the percentage of retained customers in the entire set: 99.9%/65.3%=1.53, which is the maximum decile. This means that when soliciting to the top 10% of the file based on the model, one can expect 1.53 times the total number of responders found by random soliciting 10%-of-file. The larger the cumulative lift the better the accuracy for a given depth of file will be. 4.3.3 Discussion of coefficients

Table 3 presents the estimates of the retention model, linking ea. marketing actions to retention. The Wald criterion demonstrates that many of the coefficients made a significant contribution to explain variance in the data. The effects of the marketing variables, covariates, control variables and interaction effects are compared to the hypotheses stated by literature.

Marketing variables

In support of online (M1) and mobile (M2) banking are positively related to retention. Assuming

all other variables to stay constant, a customer who has online banking is 7.52 times more likely and a customer who has mobile banking is 8.61 times more likely to stay with the bank than a customer who does not have online or mobile banking. The direct marketing campaigns ‘becoming a student’ (M3) and ‘graduating’ (M4) are negatively related to retention, caused by the fact that these

0 50 100 0 20 40 60 80 100 Per ce n tage o r re sp o n d e n ts Percentage of customers

Figure 4 Cumulative Lift curve

retention

38 Table 3 Estimation results retention

Complete Gender Local #products

Length relationship female male non-local local 0 ≥1 0 ≥1

Independent Expβ Expβ Expβ Expβ Expβ Expβ Expβ Expβ Expβ

39 campaigns are directed to customers with a higher probability of churning because they are on a natural switching point in their customer relationship. Sadly it is not possible to compare this churn rate with a churn rate before the campaigns where used since the campaigns have been used for so many years. Making it impossible to say whether the campaigns actually decreased the churn rate and are successful in customer retention, as in support of . Unlike expected the Keiweek (M5) is

not significantly influencing retention, as hypothesizes in .

Covariates

Since age effects were not possible to measure, cannot be confirmed. Gender and locality are

not found to significantly influence retention, as stated in and . The number of products is

negatively related to retention, as in contrast with . This unexpected sign can be caused by the

fact that many of the current customers, having a short relationship with the bank, have on average a lower number of products than customers who have had a long relationship with the bank and might have churned by now. The continuance of this study neglects this unexpected sign, because the main interest of this study is the effect of marketing efforts. Furthermore, cannot be confirmed since

the length of relationship is not found to significantly influence retention. Control variables

The control variables for channel are found to influence retention. Assuming all other variables to stay constant the use of paper mailing, telephone and email increase retention by a factor of 2.305, 5.501 and 1.319 respectively.

Interaction effects

The underlined coefficients in bold in Table 3 display the coefficients of the interaction models without overlapping the confidence being significantly different at p <.05. The effect of the number of products is negatively related to retention for both non-local and local customers. This effect is significantly stronger for local customer, for each product the customer purchases the probability of retention decreases with 88% against a decrease of 65% for non-local customers.

Furthermore, the use of mobile banking (M2) increases the probability of retention with 553.8% for customers having one or more of the products, in contract with . The effect of locality