The Influence of Rewards and Customer
Heterogeneity on Redemption in Loyalty Programs
2
The influence of rewards and customer heterogeneity on redemption in loyalty programs
Master Thesis
Msc Marketing Management & Intelligence Faculty of Economics and Business
Department of Marketing University of Groningen By:
Rianne Overwijk (S2575515) De Opslach 1
8405BW Luxwoude, The Netherlands +31 6 23166255
r.overwijk.1@student.rug.nl
First supervisor: Prof. Dr. T.H.A. Bijmolt Second supervisor: Msc. A. Minnema Company supervisor: Ms., S. Christiaansen In cooperation with:
Loyalty lab
Van Slingelandtstraat 63
1051 CG Amsterdam, The Netherlands
3
MANAGEMENT SUMMARY
Nowadays many firms in a broad set of industries adopted a loyalty program (LP) to reward frequent customers. It has been proven that LPs deliver several benefits for firms in terms of customer loyalty, additional store spending and increased purchase frequencies. However, the implementation of an LP is a costly marketing tool. Therefore, careful strategic decisions should be made on the implementation and design of the program.
For an LP to be effective, customers should observe the program as valuable and should perceive benefits from participating in it. If there is no value for the customer, marketing investments will be lost. In this line of reasoning, several studies suggest that the program design including the reward offered is of high importance for the success of the LP. Therefore, firms should offer relevant rewards in order to make the program attractive to its members. The attractiveness and success of the program is often measured by the redemption rate. Redemption means that customers hand something in, in this case a voucher, and get something in return. Higher reward redemptions indicate higher perceived value. The redemption of rewards leads to the so called rewarded behavior mechanism; where receiving gifts influences the LP member’s behavior. Redemption creates feelings of satisfaction which in turn leads to higher levels of commitment to the firm, and therefore increased purchase patterns and retention. Understanding which factors influence redemption behavior are of high importance to create value for both the customer and the firm.
Perceived value of an LP is partly determined by the attractiveness of the reward. Therefore this study investigates the effect of four types of rewards on redemption. The rewards can be classified among economic benefits (discount or a gift) or hedonic benefits (personalized service). Within these categories rewards can be either related to the firm or not. This results in four reward types; non monetary non firm related/ monetary firm related/ non monetary non firm related/ or hedonic firm related rewards.
Furthermore, it is unrealistic to assume that value perception or reward redemption is the same for every member. Therefore the effect of customer characteristics has been taken in to account for both redemption behavior and reward preference. In this study customer heterogeneity is included on demographic, psychographic, behavioral and financial customer factors. The study was conducted for an LP in the optics industry, in cooperation with the firm Loyalty lab who provided the data. Data on purchase history and direct mailing was used from five large retailers in the Netherlands, over a period of three years. Customers in the LP receive mailings which they can hand in at the store, in order to receive a reward. Redemption behavior will be investigated by conducting a Logit regression on whether a customer redeemed a reward or not.
4 purchases increases customers are also more likely to redeem. This indicates that retailers might consider to spend more on targeting those customers who made their last purchase several years ago. However, it was also found that long term customers are less likely to redeem than customers who just started a relationship with the firm.
Furthermore, the findings show that there is a difference in reward preference across customers. This indicates that different rewards should be used for different customers or a wider range of rewards should be provided within the program. However, due to instability of the results, specific conclusion cannot be provided and further research is needed. Some of the effects showed stable effects across samples, and therefore guidance for conclusions. Namely, for customers with more time between subsequent purchases, no difference is found among the rewards. Only that offering a reward is more effective than direct marketing without a reward. Next it was found that reward redemption differs per reward across relationship length. It can be concluded that short term customers can be best targeted with a firm related gift or a personalized service, while for long term customers it seems more appropriate to send a relational marketing mail instead of a reward.
These findings provide useful insights for management in such a way that managers are able to more effectively allocate marketing budgets within the LP. Next, insights are gained in the most effective rewards across customers in order to deliver a valuable program and increase the effectiveness of the LP.
5
PREFACE
Here it is, my master thesis; meaning the end of my studies at the University of Groningen. It means the final project for my Master in Marketing Management and Intelligence. After almost half a year full of stress and hard work, this is the final result. It also means the beginning of a new period in my life and an end of my life as a student. Which I think is on the one hand kind of sad. Personally I enjoy challenge and like to improve myself by obtaining new knowledge and experiencing new things. That was also the mean reason for me to start with this double Master in Marketing. And that is also the reason why I enjoyed it so much as well as writing this thesis. This final project means I need to find new ways to improve myself and I am looking forward to the challenges that come with life as a starting professional. I am proud of my achievements and the final project I deliver to you.
It was an honor to graduate in the field of loyalty programs especially because of my history in the hospitality industry. And second of all, because it allowed me to work with Prof. Dr. T.H.A Bijmolt who was my supervisor during this project. I would like to thank him for sharing his knowledge on this topic and providing me with feedback when needed. Also thanks to him I got in contact with the firm Loyalty Lab. Who provided me the data in order to write this thesis, as without data there would be no research. Therefore I would like to provide some special thanks to the whole team of Loyalty Lab for their help and support. Special thanks I would like to give to Ms. Sandra Christiaansen of Loyalty Lab for her time and cooperation on this project. Hopefully it will provide the company with some relevant insights and do you enjoy reading it as much as I did writing it.
Finally I would like to thank my family and friends who have always supported me during my entire studies. They were always willing to help me, and provided me time to take my mind of things so I was better able to stay focused and finish with good results. Everyone thank you and enjoy reading.
6 TABLE OF CONTENTS MANAGEMENT SUMMARY 3 PREFACE 5 TABLE OF CONTENTS 6 1. INTRODUCTION 9 1.1 Research question 10 1.2 Research context 11
1.3 Contribution and relevance 12
1.4 Outline 13
2. THEORETICAL FRAMEWORK 13
2.1 Reward redemption 14
2.2 Type of Reward 15
2.2.1 Monetary vs. Non-monetary rewards 16
2.2.2 Firm related vs. Non-firm related rewards 16
2.2.3 Hedonic firm related rewards 17
2.2.4 Direct marketing 17 2.3 Customer characteristics 18 2.3.1 Age 19 2.3.2 Gender 19 2.3.3 Behavioral factors 19 2.3.4 Psychographic factors 20
2.3.5 Total store spending 20
2.3.6 Relationship length 21
2.3.7 Customer lifetime cycle 21
7
4.2.1 Main effect model 33
4.2.2 Interaction effect model 35
5. MODEL VALIDATION 39 6. DISCUSSION OF RESULTS 42 6.1 Hypotheses evaluation 42 6.1.1 Type of reward 43 6.1.2 Customer characteristics 43 6.1.3 Interaction effects 45 6.2 Conclusions 46
6.3 Limitations and suggestions for further research 46
6.4 Management implications 47
REFERENCES 49
APPENDICES 54
Appendix A: Model estimates per retailer 54
Appendix B: Model estimates interaction model per reward category 59 Appendix C: Re-estimation of the interaction model; estimation vs. validation results 63
Appendix D: Lift curve 65
8
LIST OF TABLES AND FIGURES
Table 1, Classification of different types of rewards 16
Table 2, Overview of customer characteristics used in the study 22
Table 3, Reward type included in mailing with redemption rates per retailer 26
Table 4, Comparison of redeemers vs. non-redeemers 26
Table 5, Model comparison on validation criteria 31
Table 6, Model results; parameter estimates model 1 equation 2 32
Table 7, Model estimation main effect results 34
Table 8, Output on significance for interaction terms included in the model 35 Table 9, interaction effect model results, with different benchmark categories 36 Table 10, Model estimates comparison estimation vs. validation sample 40
Table 11, Comparison of model fit; estimation vs. validation sample 40
Table 12, Comparison of estimation and validation on 50/50 sample 41
Table 13, Conclusions on hypotheses 43
Table 14, Parameter estimates for retailer 1 54
Table 15, Parameter estimates for retailer 2 55
Table 16, Parameter estimates for retailer 3 56
Table 17, Parameter estimates for retailer 4 57
Table 18, Parameter estimates for retailer 5 58
Table 19, Model results of interaction model (eq.5) 60
Table 20, Model results; significant interactions only (benchmark reward 1) 60 Table 21, Model results; significant results only (benchmark reward 2) 61 Table 22, Model results; significant results only (benchmark reward 3) 61 Table 23, Model results; significant results only (benchmark reward 4) 62 Table 24, Model results; significant results only (benchmark reward 5) 63 Table 25, Interaction effect model results; estimation (80%) vs. validation (20%) 64 Table 26, Interaction effect model results; estimation (50%) vs. validation (50%) 65
Figure 1, Conceptual framework 14
9
1. INTRODUCTION
Loyalty programs (LPs) that reward frequent customers have become one of the most common marketing tools for stimulating product or service usage and retaining customers (Kivetz and Simons, 2002). Companies in several industries nowadays adopted a loyalty program to foster customer relationships. Since the introduction of the first loyalty program in 1981 by American Airlines, the adoption of both firms and customers towards loyalty programs has increased tremendously (Kang, Alejandro and Groza, 2015).
Except for customer retention, loyalty programs offer several other benefits for firms. Leenheer et al., (2007) demonstrated a significant increase in the share of wallet of LP members, while other studies show that loyalty programs positively influence choice of company, purchase frequency and store spending (Bolton, Kannan and Bramlett, 2000; Dorotic et al., 2012; Lewis, 2004; Verhoef, 2003).
However, when implemented incorrectly LPs might become an ineffective, expensive administrative tool instead of a valuable marketing tool (Wansink, 2003). Some studies show mixed support for the effect of an LP on customer share of wallet (Meyer-Waarden and Benavent, 2007; Sharp and Sharp, 1997). One of the reasons might be that customers perceive little benefits in the LP; because only aggregate data is used, and the program does not consider customer heterogeneity (Meyer-Waarden and Benavent, 2009). As it is very costly to implement an LP, careful strategic decisions should be made on the design in order to make an LP effective (Dorotic et al., 2012).
For an LP to be attractive, it must lead to redemption; which demonstrates the benefits for customers (Nunes et al., 2006). If customers perceive little or no benefits from participating in an LP, marketing investments might be lost (Mimouni-Chaabane and Volle, 2010). Reward redemption has an important impact on the behavior of customers, especially before and after redemption (Dorotic et al., 2012). The benefits of an LP might serve as an important factor when choosing between several competitors. Plus, the awarding of points or gifts is a common way to provide customers with a reason to purchase at the firm (Nunes et al., 2006) as the ability to obtain a reward motivates a member to increase expenditures (Dorotic et al., 2012). Dorotic et al., (2012) state that when a customer obtains a reward, he or she might feel encouraged to increase purchases due to emotional attachment or behavioral lessons learned. Redemption is seen as important, as it creates a feeling of satisfaction and a sense of reciprocity which stimulates purchase behavior (Palmatier et al., 2009).
10 1.1 Research question
In order to make LPs more attractive and therefore encourage increased purchases, firms might consider to link rewards towards customers individual needs. This might result in higher LP participation due to the attractiveness of the rewards, more redemptions, and increased spending. This study will investigate the following problem statement:
“How do different rewards and customer characteristics influence redemption behavior in a loyalty program?”
In order to answer this, several research questions have to be investigated;
First it should be understood which different types of rewards exist. Next, as the design of the loyalty program influences whether customers redeem or not (Leenheer et al., 2007) it is relevant to investigate how several types of rewards influence this decision. In the first part of the theoretical framework different types of rewards will be discussed. Rewards can be classified into two broad categories; economic or hedonic rewards. Within these categories rewards can be either related to the firm’s product offering or not. Next a reward can be either monetary in the form of cash or a discount, or non monetary as in a gift. It has to be investigated how these different types of rewards influence the likelihood of redemption. Therefore the first question in this research includes;
RQ1: Which type of reward (monetary firm related/ non monetary firm related/ non monetary non firm related/ hedonic reward) leads to highest redemptions?
Furthermore, previous studies found individual differences in the response to LPs, especially among the impact of demographics and spending levels of members (Dorotic et al., 2012; Leenheer, 2007; Lui, 2007). However these effects are still uncertain (Dorotic et al., 2014) especially in the context of redemption behavior. Therefore, it seems relevant to understand how customer characteristics might possibility influence redemption behavior. In the theoretical framework several customer characteristics will be discussed, which leads to the second research question;
RQ2: How do different customer characteristics (as age, gender, product usage or financial aspects) influence redemption?
The effect of various types of rewards on redemption behavior might differ across customers as well, as the perceived value of rewards differs across customer (Meyer-Waarden, 2013). Therefore a loyalty program should be appealing to individual customer needs. This might mean that certain rewards might be more appealing for one customer than for the other. The last part of the literature review focuses on the moderation effect of customer characteristics. However little to no research exists on how the effect of the different rewards differs across customers. Therefore the current research will also investigate;
11 1.2 Research context
In order to examine the problem, the research will be conducted in cooperation with a loyalty marketing company in the Netherlands, named LoyaltyLab. LoyaltyLab is an independent company that creates and supervises effective loyalty programs for retailers (loyaltylab.nl). For this study an LP in the optics industry will be used, the so called “Uneyeted Loyalty Program”. There are some specific features within the program that distinguish it from other loyalty programs. Uneyeted is an LP that has been developed for private opticians in Europe. Opticians can join the LP and LoyaltyLab will then get access to their customer databases in order to provide the opticians with detailed information about their customers. As a result, LoyaltyLab assists opticians to set up marketing campaigns in order to increase purchase behavior and establish relationships. This is done by the use of incentives communicated through direct marketing mails. As soon as the optician gets a new customer, he or she will automatically be enrolled in the LP due to the registration of information after a purchase. From this point on the customer is a target to different mailings which include a reward that can be collected in the store.
Mailings are done based on the phase of the customer lifecycle in which the customer is situated. LoyaltyLab identified four different phases; After sales, Strengthen of relation, Shorten repurchases, and Re-activation. Customers in the “after sales” phase are customers who recently made a purchase (3 to 9 months ago). The next stage (“strengthen relationship”) includes customers who made their last purchase one or two years ago. The mailings in this stage aim at additional purchases by including a reward that can be collected when making another purchase. In the “shorten repurchase” stage, customers have made their last purchase with the company three years ago. Here the optician wants to try to sell another pair of glasses. In order to do so he will send a coupon with an incentive that can be collected at the store. The last stage of the customer lifecycle is the one of “re-activation”. These are the customers who made their last purchase four to five years ago. In this phase the retailer tries to reactivate the relationship and wants to sell a new pair of glasses to make the customer active in the cycle again. After each stage yields; if a purchase has been accomplished, the cycle starts over again. Within the stages different mailing moments exist as well as regular direct marketing mails are send without a reward.
12 1.3 Contribution and relevance
The main purpose of this study is to investigate the influence of different rewards and customer characteristics on redemption. Despite the large body of research done on LP rewards, there is little known about the differences in perceived benefits and redemption across customers. This research contributes to existing studies in four different ways.
First of all, a large number of studies exist in the field of LPs. Many of these studies investigated the effectiveness of LPs or how several variables influence LP adoption (DeWulf, Odekerken-Schroder and Lacobucci, 2001; Dorotic et al., 2012; Leenheer et al., 2007; McCall and Voorhees, 2010). Relatively less research exists on the effect of redemption on increased purchase patterns (DeWulf et al., 2001; Dorotic et al., 2012; Meyer-Waarden, 2013) or on factors influencing redemption behavior. Some studies investigate the impact on direct mail response (Arora and Stoner, 1992; Bawa, 1996; Gázquez-Abad et al., 2011), however these do not include rewards. This study is one of the few that uses reward redemption as an indicator of loyalty program effectiveness. This is relevant in such a way, that redemptions reflect the success of the program and indirectly the attractiveness. Furthermore reward redemptions cause post reward effects and are therefore an important measure for LP success.
Next to that, much research has been conducted on the impact of LP design or reward type and how they influence loyalty or purchase behavior (Daryanto et al., 2010; Dorotic et al., 2012; Keh and Lee, 2006; Leenheer, 2004; Leenheer et al., 2007; Meyer-Waarden et al., 2006). Most of these studies however investigate the effect of economic versus hedonic benefits (Dorotic et al., 2012; Furinto and Balqiah, 2009; Leenheer et al., 2007; Melancon et al., 2011; Mimouni-Chabane and Volle, 2010) or the preference between direct versus indirect rewards (Keh and Lee, 2006; Nunes and Dréze, 2006; Nunes and Park, 2003; Rothschild and Gaidis, 1981). None of these studies takes a more detailed view on reward type within the different categories and measures the direct effect on likelihood of redemption across all possible reward categories at once. This research will investigate the preference between different types of rewards within the category of economic and hedonic benefits as these are both seen as important for LP members (Dorotic et al., 2012). Through adding different types of rewards, a broader understanding is gained on reward attractiveness which might help firms decide on the type of reward to offer.
Furthermore, little conclusions have been made on how redemption behavior differs across customers. Some conclusions have been made on the effectiveness of LPs between light, moderate or heavy buyers (Dorotic et al., 2012; Meyer-Waarden and Benavent, 2006). Others included customer characteristics in studies on behavioral loyalty (Dorotic er al., 2012; Leenheer et al, 2007). But none of these studies included different customer characteristics or investigated how they influence redemption. This study uses customer characteristics other than purchase frequency to investigate redemption behavior which will result in a better understanding of which members are most susceptive for reward mailings. Therefore this research will contribute to a better understanding on the effectiveness of LPs across different customers.
13 when it comes to loyalty or LP adoption (Dorotic et al., 2012). However, very little research exists on customer differences in reward preferences (Kopalle et al., 2012). This research provides relevant contributions to loyalty program design suggestions in the retail industry by considering customer heterogeneity and how the redemption differs when looking at different types of rewards.
From a managerial perspective, the research is relevant in such a way that it might help managers better understand reward redemption. This is relevant as higher redemption indicates a relevant and valuable LP for customers, which in turn results in increased purchase behavior. By understanding how different rewards influence redemption managers can make the program more attractive and effective. Next to that, the research contributes to a better understanding of how customers vary in redemption behavior. This is important for a higher effectiveness of scarce resource allocation across customers. Furthermore, this research will provide a better understanding on which type of reward to use in order to attract a certain customer to the program, or increase the effectiveness of the program and therefore additional purchases.
Thanks to this research Loyaltylab gets insights in which type of reward leads to higher redemption rates. As well, the company gets insights in which reward is preferred by which type of customer. Through the inclusion of customer characteristics, retailers can more effectively adjust rewards towards their customers in order to trigger redemption and therefore additional store spending. By knowing how different customers are influenced by the reward structure, marketing budgets can be more effectively distributed and profits due to higher purchase rates of customers might compensate the high costs of LPs.
1.4 Outline
The remaining of this paper is structured as following; first a conceptual model is presented which is used as a starting point for this research. The concepts presented will be explained in the theoretical framework, and hypotheses will be derived. Next a data description and research methodology is provided including explanations on the variables used, followed by the results and main findings of the study. Finally a discussion of the findings is given together with a general conclusion including managerial implications and limitations of the research.
2. THEORETICAL FRAMEWORK
14 2.1 Reward redemption
An LP can be divined as an integrative incentive program offered by a retailer to stimulate additional spending and ongoing relationships with customers (Berman, 2006; Demoulin and Zidda, 2008; Dorotic et al., 2012; Leenheer and Bijmolt, 2008; Nunes and Dréze, 2006). The main purpose of an LP is to foster customer loyalty by the use of incentives (Dorotic, et al., 2012; Meyer-Waarden and Benavent, 2009). Rewards are seen as important in encouraging behavioral and attitudinal loyalty (Smith and Sparks, 2009). Behavioral loyalty includes increased purchase frequency, volume and retention (Dorotic et al., 2012). While attitudinal loyalty means higher level of commitment and positive attitudes towards the firm (Dorotic et al., 2012). The redemption of loyalty program rewards extensively influences LP members’ behavior (Dorotic et al., 2014). After a reward is received, a customer develops positive feelings towards the firm which results into higher purchase patterns (Taylor and Neslin, 2005), this is called the rewarded behavior mechanism. Palmatier et al., (2009) state that this occurs due to reciprocal feelings like gratitude. Another explanation might be due to behavioral learning; repurchases lead to rewards (Dorotic et al., 2012; Rothschild and Gaidis, 1981).
Furthermore, redemption behavior reflects the benefits for the customers; high redemption rates indicate high perceived value (Yi and Jeon, 2003). It can be said that redemption rates measure both the success and the failure of an LP (Smith and Spark, 2009).
15 This indicates that reward redemption is an important measurement for the success of an LP. Reward redemption means that customers turn something in (e.g. coupon or points), in order to get something in return. In the current research customers receive mailings or vouchers which they can hand in at the store in order to receive a certain reward. The percentage of customers that collect the reward compared to the total number of customers that received the voucher is called the redemption rate. Redemption rates vary significantly across industries and LPs (Smith and Sparks, 2009). Firms should always strive for higher redemption rates as they lead to the rewarded behavior mechanism (Blattberg, Kim and Neslin, 2008; Dorotic et al., 2012; 2014, Taylor et al., 2005).
Redemption behavior differs across customers, and the effect of the rewarded behavior mechanism depends on the loyalty program design (Blattberg et al., 2008; Dorotic et al., 2012; Ferguson and Hlavinka, 2008; Wirtz, Mattila and Lwin, 2007). Part of the design includes reward type, which has an important impact on whether customers redeem or not (Leenheer et al., 2007). Therefore, the effect of different rewards and customer characteristics on redemption will be investigated.
2.2 Type of Reward
The attractiveness of the reward offered by the firm is of high importance for the success of the LP (Lui and Brock, 2009). Wirtz et al., (2007) found that the more attractive the reward is perceived, the greater the perception of the LP. Higher perceived value in turn leads to an increase in the share of wallet (Floh et al., 2014), as members are highly motivated to obtain the reward. Based on the theory of self-determination, Meyer-Waarden (2013) found that various rewards have different impact on motivation to engage in a certain behavior like loyalty or redemption. The type of reward should reflect the desired benefits of the customers, because this reflects the customer perception of value and therefore behavior (Dorotic et al., 2012).
In general, a distinction can be made between utilitarian and hedonic benefits. The former refers to economic benefits as discounts or gifts. The latter refers to non-economic benefits as personalized treatment (Dorotic et al., 2012; Leenheer et al., 2007; Melancon, Noble and Noble, 2011). The current study will mainly focus on economic benefits as these are seen as most important for customers when participating in an LP (Dorotic et al., 2012; Furinto, Pawitra and Balqiah, 2009; Mimouni-Chaabane et al., 2010). Next to that, they are tangible and easily observed. However the LP studied here might also send direct mails with a free service or a personal message regarding information on glasses or inspections. These rewards will be classified among hedonic benefits or direct mails with no reward. Table 1 provides an overview of all different types of rewards. Next the categories used in this study will be explained.
16 2.2.1 Monetary vs. Non-monetary rewards
Within the category of economic benefits, a distinction can be made between monetary and non-monetary rewards; monetary rewards are rewards that relate to cash. They provide the customer with savings through discounts or cash back (Dorotic et al., 2012). Non-monetary rewards are compensations that do not involve cash but provide the customer with a material reward (Businessdictonary.com) Often, non-monetary rewards lead to emotional benefits. Gifts can generate positive feelings and memories towards the firm. Those associations might strengthen the rewarded behavior mechanism (Roehm, Pullins and Roehm, 2002). This finding is also confirmed by Nunes and Dréze (2006) who, in addition, found that incentives that promise pleasure (gifts) are more effective than pure cash rewards. This indicates that offering something for free is relatively more effective than a cash discount (Ha and Stoel, 2014; Nunes and Park, 2003).
2.2.2 Firm related vs. Non-firm related rewards
Rewards can be related to the firms’ offerings (Dorotic et al, 2012), which are described as firm related rewards, as they directly support the value proposition of the product or service offered to the customer (Dowling and Uncles, 1997). But they can also be unrelated, which indicates that received rewards are not related to the product or service offered. Existing research mainly indicates that firm related rewards are preferred over non-firm related rewards (Dorotic et al., 2012; McCall and Voorhees, 2010). Keh and Lee (2006) demonstrate that direct (vs. indirect) rewards, as they relate to the firm, are more easily processed and therefore generate higher valuation of the service. Furthermore, the customers’ involvement plays an important role. For many low involvement products, the reward becomes more important than the product. While for high involvement products the product is the primary reward and not the incentive (Dowling and Uncles, 1997; Yi and Jeon, 2003). Buying glasses in general is a high involvement process, therefore rewards that relate to the firm can be expected to result in higher redemption rates than non-firm related rewards.
As previously discussed, non-monetary rewards and firm related rewards can be seen as most effective. Therefore it is easily derived that a reward possessing both these characteristics will be most effective;
H1: Non-monetary firm related rewards, compared to monetary firm related/ non-monetary non-firm related/ hedonic rewards, will result in the highest redemption.
Firm related Non- firm related
Economic:
Monetary e.g. discount on glasses e.g. discount at other retailer Non-monetary e.g. free glass cleaning spray e.g. a gift (bottle of wine) Hedonic:
e.g. Free eye check x
17 The remaining economic rewards include both an ideal as well as an imperfect aspect of rewards as mentioned by the literature. However, no research exists on how these characteristics influence redemption. This makes it hard to make inferences about the effect on redemption of the remaining categories. One study on the effect of rewards towards brand associations within loyalty programs by Roehm et al., (2002) might provide guidance for hypotheses. The authors argue that an incentive might be a motive for a customer to make repurchases if they strengthen brand associations. This depends on two dimensions; cue compatibility and tangibility (Roehm et al., 2002). Cue compatibility is the degree to which the reward activates brand associations. The higher the compatibility the more likely a customer is to make another purchase. Rewards might differ in the degree of tangibility. Roehm et al., (2002) state that relatively intangible rewards might lead to a sense of belonging and positive attitudes to the brand.
Following this reasoning, firm related rewards can be seen as high degree of cue compatibility and monetary rewards as highly intangible. This might indicate that monetary firm related rewards might be more successful in triggering repurchases than non-monetary non-firm related rewards.
H2: Monetary firm related rewards, compared to non-monetary firm related/ non-monetary non-firm related/ hedonic rewards, will result in the second highest redemption.
H3: Non-monetary non-firm related rewards, compared to non-monetary firm related/ monetary firm related/ hedonic rewards, will result in the fourth highest redemption.
2.2.3 Hedonic firm related rewards
Hedonic rewards can be seen as special treatment rewards, they provide personal service to loyal customers (Furinto et al., 2009). Such rewards provide emotional and psychological benefits to customers (Mimouni-Chaabane and Volle, 2010). In the current study hedonic benefits include personalized or free services which are all firm related. Even though Melancon et al., (2011) demonstrates that hedonic rewards are more effective in creating strong emotional relationships with customers, many studies provide evidence that economic benefits are most effective in explaining satisfaction and loyalty (Furinto et al., 2009; Keh and Lee, 2006; Mimouni-Chaabane and Volle, 2010). As the other rewards in this study are all economic benefits. It can be argued that in line with these studies, it might be expected that hedonic rewards will be less effective than economic rewards however due to the firm related characteristic they might be more effective than economic non firm related rewards. Therefore, comparing hedonic to economic benefits it can be stated that;
H4: Hedonic firm related rewards, compared to economic (non-monetary firm related/ monetary firm related/ non-monetary non-firm related) rewards, will result in the third highest redemption.
2.2.4 Direct marketing
18 relationships (Gázquez-Abad et al., 2011). It has been proven that addressing customers personally build trust and long term relationships (Gázquez-Abad et al., 2011). Next relational marketing influences purchase behavior (Gázquez-Abad et al., 2011; Verhoef, 2003) as communication is tailored towards individual needs, it causes a desire to fulfill these needs (Verhoef, 2003). Therefore it might be assumed that direct marketing itself might be a trigger to visit the store. Verhoef (2003) concludes that reward based loyalty programs positively affect both customer retention and purchase behavior while direct marketing does not influence customer retention. Arora and Stoner (1992) found that response rates are higher for direct mails with an incentive compared to those without. Therefore the following hypothesis is derived to describe the effect of direct marketing in relation to the different rewards. H5: Direct marketing communications without a reward, leads to lower redemption rates than mailings with a (monetary firm related/ monetary firm related/ monetary non-firm related/ hedonic) reward.
2.3 Customer characteristics
In order to provide a relevant LP to members, a deeper understanding of individual customer needs is required (Ferguson and Hlavinka, 2008). Over the past years a shift took place from mass marketing towards a customer centric marketing approach. This change lead to an emphasis on meeting individual needs rather than those of mass markets or segments (Seth, Sisodia and Sharma, 2000). The development in customer relationship management taught marketers not to consider all customers as equal (Leenheer, 2004). Database technology makes it possible to analyze individual customer behavior (Bell et al., 2002), this would also be favorable for LPs.
First of all, customers’ different needs can be explained through demographic variables as gender, income and age variety (Seth, Sisodia and Sharma, 2000). Typically, LP members are distinguished by profitability and preferences. This is often done through the establishment of customer tiers, which provide customers with a certain status (Voorhees et al., 2011). Here customers differ on frequency of visiting, relationship duration or spending levels (Tanford and Malik, 2015). However, a broader set of customer characteristics should be taken into account in order to meet individual needs (Voorhees et al., 2011). As customers differ on many aspects, it is purposed to include individual behavior (Rust and Verhoef, 2005) and psychographic variables as lifestyle and values (Levin and Zahavi, 2001; Tanford and Malik, 2005; Voorhees et al., 2011). LPs that take into account background and purchase information for targeting are able to generate incremental business from their members (Wansink, 2003).
19 2.3.1 Age
The age of an LP member expressed in years might influence redemption behavior. It was found that older members decrease their spending patterns over time as well as their redemption behavior (Leenheer et al., 2007). However when looking at coupon redemption, results show that females over 55 years old are most likely to redeem while females below 40 years are least likely to redeem coupons (Lee and Brown, 1985). There seems to be a mixed effect of age on redemption behavior. When looking at loyalty program adoption, studies indicate that there is no significant effect of age found (Demoulin and Zidda, 2009; Dorotic et al., 2012). As the LP studied at hand is in the optics industry, where mainly elderly people need glasses, a hypothesis can be derived in line with the results of Lee and Brown (1985). H6: Age positively influences redemption behavior.
2.3.2 Gender
Little is known about gender differences in loyalty programs especially in redemption of rewards. Still this is an important aspect to investigate. If there is a difference in redemption behavior between men and women, it means they respond differently towards loyalty programs and therefore require different marketing approaches (Melnyk, van Osselaer and Bijmolt, 2009). Several studies investigated the difference in satisfaction and loyalty between men and women (Ladhari and Leclerc, 2013; Melnyk et al., 2009; Melnyk and van Osselaer, 2012). However mixed results were found; Ladhari and Leclerc (2013) show higher levels of satisfaction and loyalty for women compared to men. Melnyk and van Osselaer (2012) state that the difference in loyalty between men and women can be explained through the offered reward. Men respond more positively towards a reward that is associated with status, whereas women prefer personalized services (Melnyk and van Osselaer, 2012). As none of these studies directly relate gender to reward redemption it is difficult to hypothesize the effects. However, in this LP mainly economic rewards are offered which might illustrate a certain social status, it might be expected that man in this case redeem more than women.
H7: Gender has a significant effect on reward redemption; men redeem more often than women.
2.3.3 Behavioral factors
20 H8: Average time between subsequent purchases negatively influences reward redemption. H9: The number of different products a customer bought positively influences reward redemption.
2.3.4 Psychographic factors
Psychographic variables differentiate customers on a deeper level, in terms of personality, lifestyles, values and interests (Levin et al., 2001; Voorhees et al., 2011). Customers might be the same in terms of store spending and relationship length, but might differ in terms of their values and lifestyle (Tanford and Malik, 2015). In this research a segmentation variable is used to identify a customer based on a set of characteristics. Market segmentation is a commonly used tool in marketing (Dolnicar et al., 2014). The main idea is to create groups of customers with similar characteristics, needs and purchase behavior (Levin and Zahavi, 2001). The segmentation codes used in this research are determined by the company and used by retailers to classify their customer base among. The segments are constructed based on various databases of the Dutch population including demographic and psychographic aspects. Next the customer base of the retailer is compared with the overall population and customers can be classified among one of the fourteen possible segments with their own unique characteristics. As this is a very company and industry specific variable no inferences can be made about the effect on redemption behavior. However due to the differences among the segments it might be assumed that there is a significant difference between the different segments and redemption. Previous studies on perceived value show that price sensitive segments prefer economic benefits as gifts or discounts, while segments that value quality give preference to hedonic rewards (Kopalle et al., 2012; Meyer-Waarden, 2013). As the segments differ in income, family size, lifestyle and values there might be a difference in redemption behavior (Tanford and Malik, 2015). Especially for the retailer it might be valuable to investigate which segment is most likely to redeem. Based on previous findings, it might be assumed that the segments with lower incomes who use public transportation, rent their houses and do not live a luxurious life redeem more because the LP mainly offers economic rewards. Those segments are seen as less wealthy and therefore it can be stated that; H10: The different segments significantly influence reward redemption; the less wealthy segments redeem more often than the high wealthy segments.
2.3.5 Total store spending
21 H11: Total store spending positively influences reward redemption.
H12: The number of purchases positively influences reward redemption.
2.3.6 Relationship length
In this study relationship length will be measured by the number of years a certain person is a customer at the firm. This is indicated by the number of years since the very first purchase. Redemption effects differ across LP members, especially when looking at the length of the membership. Dorotic et al., (2014) state that the length of the membership negatively influences redemption decision. The longer a member of an LP, the less he/she redeems. This phenomenon can be explained by the concept of behavioral learning. Rothschild et al., (1981) state that when a customer buys for a longer period of time at the same firm, the likelihood of redemption becomes lower due to behavioral learning where the behavior is reinforced by the product or service and not by the reward. In such a case the company or product becomes the main motivator and not the reward. Rewards play an important role in the early stages of the relationship, when customers gain more experience with the firm they rely less on promotional marketing (Gázquez-Abad, De Canniére and Martinez-Lopez, 2011).
H13: Relationship length negatively influences reward redemption.
2.3.7 Customer lifetime cycle
The customer lifecycle stage indicates in which phase of the relationship the customer is situated. In this study retailers use four different lifecycle stages; “after sales”, “strengthen relationship”, “shorten repurchase”, “re-activation”. The further the customer is located in the cycle, the longer it has been since the last purchase. Customers enter the cycle once they made their first purchase. Within each stage different mailings will be send after various time frames (months). A customer who made a purchase three to nine months ago will be in the “after sales” stage. One who made their purchase twelve to twenty-four months ago will be situated in the “strengthen relationship” stage. Thirty to forty-five months ago indicates “shorten repurchase”, and those who made their purchase forty-eight till sixty months ago will be in the “re-activation” stage. Once a second purchase has been made, the customer will move to the first stage in the cycle again. The lifecycle does not indicate the length of the relationship and therefore might be valuable to add as a variable to this study. It can be assumed that when a customer just made a purchase, and therefore is situated in the first stage of the lifecycle will not rapidly make another purchase again. As glasses are often replaced or need maintenance after three to four years it might be assumed that customers who are situated further in the lifecycle will be more likely to respond to the mailing and therefore are more likely to redeem.
22
Segmentation variables
Dimension Variable Definition
Demographics Age The age of a customer in the LP expressed in years Gender Whether the customer is a Male or Female Behavioral Average time between
purchases
The average time before a customer makes a subsequent purchase measured in months.
Variety of products The different products a customer bought at the retailer this can be glasses, lenses or both.
Psychographics Segment code Indicates on a set of demographic and psychographic factors to which segment the customer belongs. Codes are used by retailers and provide information on lifestyle, values, and socio-demographic aspects.
Financial value Total store spending The total amount of money, expressed in euro’s (€), the customer has spend during his/her relation with the firm Number of purchases The total number of purchases a customer made at the firm
during his/her entire relationship.
Relationship length The amount of time a customer has a relation with the firm measured in years since the first purchase.
Customer lifetime stage The stage of customer lifecycle in which the customer is situated depending on the time since the last purchase.
2.4 Moderation effect
LP rewards should add value to the customer in order to be effective (Tanford and Malik, 2015). It is unrealistic to assume that value perception and therefore customer behavior (e.g. loyalty, word of mouth, willingness to pay or redemption) are similar across all customers (Floh et al., 2014). A customers’ decision to enroll in an LP is determined by the benefits of the program (Dorotic et al., 2012). Perceived benefits refer to the perceived value customers get from the program, so what the program provides its members (Mimouni-Chaabane and Volle, 2010). Xie and Chen (2013) state that in order to go from acquisition to retention it is necessary to create program value and a customer-program fit. This is an LP which fits with its’ customer needs and where customers are able to identify with the benefits (McCall and Voorhees, 2010). There should be a balance and match between an LP’s reward structure and the customer’s goal or motive (Ha and Stoel, 2014). Marketers should develop loyalty programs that appeal to the target customers by providing rewards that are relevant to individual customers, in this way the program really matters for customers (Ha and Stoel, 2014.
Various types of rewards induce varying levels of perceived value (Meyer-Waarden, 2013). If customers perceive a reward as value adding, it will positively influence behavioral loyalty (Matilla, 2006). Floh et al., (2014) support this finding and add to it that the effect differs across customers. Specific heterogeneous customer goals and needs influence the preference for LP rewards (Daryanto et al., 2010). Therefore, the perceived value of rewards differs across customers (Meyer-Waarden, 2013). Most reward programs offer a “mass” approach, while ideally LPs should tailor rewards to meet the needs of individual customers (Kumar and Shah, 2004; Voorhees, et al., 2011). To improve resource allocation, customers
23 should be targeted according to reward preferences (Berman, 2006; Meyer-Waarden, 2013; Voorhees et al., 2011).
One major advantage of LPs is the availability of customer data. The data can provide insights into customer buying behavior and spending levels (Berman, 2006; Nunes and Dréze, 2006), but it also allows the seller to tailor offerings, or assign rewards to individual customers (Berman, 2006; Nunes and Dréze, 2006; Xie and Chen, 2013). The usage of such data allows personalization of the LP. Tesco, a UK grocery store, is a well known example for using data from its Clubcard members (Berman, 2006; Kumar and Shah, 2004; Nunes and Dréze, 2006). The firm used LP data to create customer segments in order to offer a personalized coupon program. “While the industry average for redemption is between 1 and 2 percent, between 15 and 20 percent of Tesco coupons are redeemed” (Berman, 2006).
As not all customers relate to the same benefits, it might be relevant to research the moderation effect of customer characteristics in order to identify how customers respond to different benefits (Mimouni-Chaabane and Volle, 2010). Customer characteristics, in combination with previously described rewards, will be used to investigate how certain combinations strengthen or weaken reward redemption. As far as known, no research investigated these effects. Therefore, little inferences can be made on the direction or the influence of the different variables. Most studies recognize individual differences however which ones are not clear (Meyer-Waarden, 2013; Taylor and Neslin, 2005). Especially the effect of demographics and psychographics remains unclear (Dorotic et al., 2012; Dorotic et al., 2014). This research will investigate whether moderation effects of customer characteristics, with respect to redemption of various rewards, exist and how redemption is influenced by it.
3. METHODOLOGY
3.1 Study setting
The current research focuses on redemption in LPs in the Dutch optics sector. The data for this study was derived in cooperation with the company Loyalty Lab. The optics sector is a unique and interesting sector to investigate as consumers do not frequently visit or spend money at an optics store. Therefore understanding what customer’s value in an LP in this sector can provide insightful information as a base for future investments.
24 The overall structure of the LP is the same across the different retailers. The LP works with mailings on fixed periods in time based on the customer lifetime cycle. This means that depending on the stage (i.e. After sales, Strengthen of relation, Shorten repurchases, Re-activation) the customer will receive a suitable and personalized message with the intention to increase store spending and reduce re-purchase time. The mailings can be unique per retailer and include different types of rewards for the customers. As each retailer can offer its own chosen reward, the reward scheme is very heterogeneous. When customer’s purchase at one of the participating retailers, they will be automatically part of the mailing cycle and therefore the LP. From that point on longitudinal data is gathered from the touch points the customer has with the company. This indicates that customers do not participate voluntarily. Especially for this reason effectiveness among different customers should be measured in order to make the LP as effective as possible for the retailer.
3.2 Data description
From the entire database of 250 participating retailers, data from five large retailers was made available by Loyalty Lab in order to perform this research. This resulted in a dataset of around 100,000 customers who made a purchase at one of the retailers. Data includes information on consumer characteristics and marketing initiatives over a time period of three years (2013 - 2015), and purchase history of the last seven years (2008 – 2015). The database contains customer-level information on which mailing has been received, when the mailing has been send and the content/ reward of the mailing. Furthermore it is known when a certain customer made a purchase and the amount spend in the store. In order for redemption to happen, it is necessary that a customer received a mailing at least once during his/her relationship with the firm. Therefore only those customers who received mailings will be included in the sample size. An important aspect of the data includes the fact that some customers occur more often in the data set, this happens due to the fact that one customer might have received several different mailings. As the interest lies in response towards different mailings, it is allowed for a customer to occur more often.
3.2.1 Data purification
Before analyzing the data, some data cleaning has been performed by checking irregularities and missing values. First of all, every customer in the database had to be provided with a unique case number as the different retailers in the dataset used the same customer id in some cases. This would result in duplicate cases while those customers are not identical. In order to do so, the highest occurring customer id was looked up and added by a larger value. In this case 100,000 was added to customers for retailer 1; 200,000 for retailer 2 and so on. Next it occurred that some customers had untraceable customer id’s (e.g letters instead of numbers), in total 52 cases all from retailer 2. These customers could not be identified or matched to mailing data and had to be deleted from the dataset.
25 purchases over the period of time. Furthermore, the total store spending showed some outliers. It has been decided to leave those outliers in the dataset in order to prevent loosing valuable information. Because the retailers are all operating in the high end segment, and customers with those outliers are long term customers, it might seem reasonable that purchase amounts around €10,000 occurred. Especially when several purchase have been made and when a customer bought both glasses and lenses.
When looking at the missing values, many variables showed unknown or unobserved values which have not been inserted into the database. Each variable has to be considered carefully in order to not distort the data or delete valuable information.
For redemption it is necessary that the customer actually received a mailing. This means that new customers who made their first purchase and therefore did not receive a mail jet, as they just entered the mailing cycle, will be excluded from the sample. Also the cases of which mailing information is missing will be deleted. In total 31,422 cases have been deleted because no redemption conclusions can be made about these cases.
For the variable age 116 cases across different retailers showed incomplete information. As for many of these cases a segment code is available which is based on demographic and psychographic factors, it has been decided to look up the segment code and replace the missing value with the average age of the segment. This is preferred over replacement with the overall mean as this might distort the standard deviation and reduce the variance even more (Silva and Zárate, 2014). By using this procedure, the missing values of age were reduced to 33 cases and the rest is replaced by the overall mean (i.e. 58) as the mode and the mean do not differ much.
The variable gender contained 62 missing values, therefore in this case an extra category is added labeled “unknown”. The same has been done for segment code. For this case 13600 cases have been labeled as “unknown”. Around 2700 cases within the variable total purchase amount were missing. When this was the case, it would indicate that the customer made a purchase before the year 2008, because customers only receive mailings when they made a purchase. In this case there is also no purchase date available. Therefore it has been decided to set the value of these cases to €0.00 as no purchase occurred within the observed period. These customers can be considered as inactive customers.
3.2.2 Descriptive statistics
26
Total nr. of different rewards sent per retailer with redemption rates Non-Monetary Firm related Monetary Firm-related Non- Monetary Non- Firm related
Hedonic reward Direct Mail (no reward)
Retailer N %Y=1 N %Y=1 N %Y=1 N %Y=1 N %Y=1
1 1191 2.4% 1374 2.0% 920 2.8% 2571 2.6% 4037 2.6% 2 508 4.1% 794 1.6% 0 0.0% 1659 2.2% 751 1.1% 3 2158 3.0% 1898 2.0% 1531 2.8% 4882 2.6% 5274 3.7% 4 0 0.0% 4055 1.8% 0 0.0% 8748 2.6% 7050 2.3% 5 2122 4.4% 2200 3.4% 1883 2.7% 6691 3.8% 3962 2.4% Total 5979 3.3% 10321 2.1% 4334 2.7% 24551 2.8% 21074 2.5%
When looking at redemption rates, out off all mailings sent 2.8% of the customers went to the store and claimed the reward. Comparing this with averages in the retail industry which run from 1.3% to 14% (Hozier and Fernando, 1985) and direct mail coupon response (around 3%) (Bawa, 1996; Beasley, 2013) it can be said that response is low but around average. It has been found that redeemers differ on purchase behavior compared to non-redeemers. From table 4 it can be seen that on average redeemers spend more and made more purchases than non-redeemers. In general, the customers who made a purchase with the firm in the observed period on average spent €1,066.21. Purchase amounts vary between €10 and €15,000 over the eight year period. And the average number of purchases across all customers is three. It can be seen that those customers who did redeem spent more and made more purchases than the overall average across all customers as well.
Variable Redemption No redemption
Total store spending
Mean € 1,811.03 € 1,044.89 St. Deviation € 1,203.14 € 972.56 Number of purchases Mean 4.70 3.15 St.Deviation 2.50 2.41 3.3 Research method 3.3.1 Model choice
This research investigates individual customer behavior, where the dependent variable indicates redemption behavior. As the dependent variable can only take two forms; 1 if redemption happened and 0 if no redemption occurred, the variable is a so called binomial variable (Leeflang et al., 2015). In such a case nonlinear regression models should be used, as the error terms are not normally distributed (Franses and Paap, 2001). A binary choice model will be specified to examine the problem. The binary choice model uses probabilities of the redemption event to happen given several observed variables.
Table 3, Reward type included in mailing with redemption rates per retailer
27 Most commonly used choice models are the Logit model or the Probit model (Franses and Paap, 2001; Leeflang et al., 2015). In order to pick the model best suited to explain redemption, both models are estimated and compared on their results. Outcomes of both models provide similar results. When looking at the fit statistics it can be seen that the Logit model has a slightly better fit (Logit: AIC = 11891.94; BIC = 12192.38, Probit: AIC = 11923.04; BIC = 12223.48). As both functions provide quite similar results, it has been chosen to proceed with the Logit model due to its mathematical convenience and slightly better fit. Furthermore, the Logit model is often preferred because the parameters are easier to interpret and the calculation of the probabilities is simpler compared to the Probit model (Leeflang et al., 2015).
3.3.2 Model specification
The Logit model represents the probability of redemption to happen (Y=1) by looking at the unobserved utility explained by several explanatory variables. The probabilities of the model are provided by the cumulative distribution function represented by F, where probabilities always lie between zero and one (Leeflang et al., 2015). This results in the following model specification;
The model for Ui will be constructed twice in order to test the influence of customer
characteristics, different types of reward and the moderation effect on redemption. The first model will only contain the main effects of the customer characteristics and the different rewards. The second will include the interaction effects. In both models the dependent variable will be a binomial variable which indicates redemption (1=yes/ 0=no).
3.3.3 Checking assumptions
28 As data is derived from different retailers it has to be checked whether there are differences in results between the different retailers, in order to see if it is allowed to pool the data into one model. This can be done through estimating the model separately per retailer and comparing the overall results1. When looking at the signs and significance of the main effects no outstanding differences were found. Therefore it has been decided to proceed with an overall pooled model.
3.4 Model construction 3.4.1 Dependent variable
The dependent variable used in the model is a binary variable indicating whether a customer redeemed a reward in the observed period of time or not. This variable has been constructed based on a comparison between the mailing date and the purchase date, as no direct redemption data was available. The company tracks whom receives which type of mailing but does not register whether the reward has been claimed. Loyalty lab uses a 60- 70 day response frame for the effects of direct mailing. As mailing date and purchase date are known per customer, a so called “max response” date was constructed which adds 60 days to the mailing date. Next, all purchase occasions of one customer are compared to this date. When the purchase date lies within the mail date and the max response date it can be assumed that a purchase has been done due to the mailing, and therefore redemption took place. A limitation of this method is that in some cases a purchase happened by chance within the given period, but no reward has been claimed. However it is still a reasonable assumption as many rewards can only be collected when an additional purchase has been made (e.g. get 20% on your next purchase). Also when a customer visits the store for a free service this will be indicated as a purchase date, but then with a purchase value of zero. The redemption variable takes one when a customer made a purchase within 60 days from the mailing date. If the customer did not make a purchase or the purchase date lies further from the mailing date, redemption did not happen, and the variable takes a value of zero.
3.4.2 Explanatory variables
The model includes possible main effects that several explanatory variables might have on redemption. Next the different independent variables will be explained.
Customer characteristics:
Agei = age of customer i measured in years. This variable has been constructed by
subtracting the date of birth from the current date (2015).
Genderi = indicator variable for customer i gender, this variable can take three forms;
Gender0i: male, Gender1i: female, and Gender2i: unknown (benchmark).
Segmenti = indicator variable for the general customer segment to which individual i
belongs. In the original dataset, the segment code consisted out of 14 categories. In order to keep the model simple jet complete, the number of categories has been reduced to six through the combination of categories based
29 on the similarity across the categories. The following categories will be added to the model;
SEG0i: Unknown (benchmark)
SEG1i: Segment A and B combined
SEG2i: Segment C and D combined
SEG3i: Segment E and F combined
SEG4i: Segment G, H and N combined
SEG5i: Segment I and L combined
SEG6i: Segment J, K and M combined
AVPurchi = the average time between the different purchase occasions of customer i
measured in months. This variable has been constructed by adding the number of months between the different purchase occasions of an individual and dividing it by the total number of purchases made by that individual. This variable is only applicable if several purchases have been made therefore a dummy variable has to be added to the model.
DUM1i = a dummy variable indicating one if more than one purchase has been made,
and zero if the customer made only one or no purchase.
DIFProdi = a categorical variable indicating whether customer i bought glasses, lenses or
both at the retailer store. Where; DIFProd1i: Glasses only
DIFProd2i: Lenses only
DIFProd3i: Both glasses and lenses
DIFProd4i: Unknown (benchmark)
TotalSpendi = the total amount of money customer i spent at the retailer in the observed
period of time, measured in euro’s. The variable is measured in 1000 of euro’s (total amount / 1000). The effect of total store spending itself is taken into account by adding an extra dummy to the model. When a customer has no purchase amount within the observed period of time, the dummy should only count when a purchase has been made.
DUM2i = dummy variable indicating whether customer i made a purchase in the
observed period or not.
RelLenghti = categorical variable indicating the number of years customer i has a
relationship with the retailer. This variable has been constructed based on the first available purchase date. As the customer enters the mailing cycle only after the first purchase has been made this might serve as an indicator of the length of the relationship. Those customers, of whom no purchase date is available, made purchases before the observed period. As the exact time is in that case not known the variable needs to be constructed as a categorical variable where their relationship length is set to over seven years. The customers who made their first purchase in the year 2015 are seen as short term customers of less than one year. The variable can take three different levels; RelLenght1i: Short term customers (<1 year – 2 years)
RelLenght2i: Medium term customers (3-6 years)
30 NRPurchi = the total number of purchases customer i made in the observed period of
time.
StageLCi = categorical variable indicating in which stage of the customer lifecycle
customer i was situated when receiving the mailing. The lifecycle stage can take four forms;
StageLC1i: After sales
StageLC2i: Strengthen relationship
StageLC3i: Shorten repurchase
StageLC4i: Reactivation
Reward characteristics:
Rewardi = categorical variable indicating which type of reward the mailing contained
received by customer i. The reward type can take the following forms; Reward1i: Non-Monetary firm related
Reward2i: Monetary firm related
Reward3i: Non-Monetary non-firm related
Reward4i: Hedonic reward
Reward5i: No reward
The previous described variables result in the following Logit regression model which indicates the probability that customer i claims the reward;
Ui = + + + + (2) 4. RESULTS
Estimation of the results will be done on an estimation sample, including 95% of all cases. Within marketing modeling, it is very common to split the data and use a validation sample to estimate the usefulness of the model (Leeflang et al., 2015). In this case a validation sample is selected of around 5% of the cases in order to test the predictive validity of the model. There is no general standard on the size of the validation sample, however in order to obtain reliable results a minimum of 100 cases is required (Vergouwe et al., 2005).
4.1 Goodness of fit
