Investigating the value of corporate loyalty programs at the store level.

(1)

Investigating the value of corporate loyalty

programs at the store level.

The Vector Autoregressive Analysis of corporate loyalty program growth and

the moderating role of store contingency factors on store revenue

Pornthip Piyarittipong

S3628000

p.piyarittipong@student.rug.nl

09-06-2020

Master Thesis MSc. Marketing Intelligence

Department of Marketing

Faculty of Economics and Business

Rijksuniversiteit Groningen

Supervised by: dr. A. Bhattacharya

(2)

ACKNOWLEDGEMENT

Given the pandemic situation, this thesis was a very challenging but rewarding final step towards completing my Double Degree journey in Master Marketing Intelligence at the University of Groningen and Master Strategic Marketing Management at BI Norwegian Business School.

Firstly, I would like to thank my supervisor, dr. A. Bhattacharya, for your guidance and valuable feedback through the process of doing this thesis. Secondly, I thank prof. dr. T.H.A. Bijmolt for your time as a second reader. Furthermore, I have also been helped and inspired by many academics, especially prof. dr. J.P. Elhorst and dr. J.P.A.M Jacobs from Economics, Econometrics & Finance department at RUG, prof. dr. R.D. van Oest from BI, Norway, dr. P. Suttiwan and T. Rojkangsadan from CU, Thailand. I sincerely thank you all for knowledge, enthusiasm, and valuable contribution.

(3)

ABSTRACT

Purpose: To assess the dynamic effects of the growth in disaggregated loyalty program attributes designed by the firm, namely the number of monthly in-store enrollment and the ratio of reward member’s expenditure to the total transaction on average store revenue. Additionally, this study further examines how this effectiveness would change when accounting for store heterogeneity across outlets of the firm.

Originality: Existing literature predominantly focuses on the design of loyalty programs in the launch phase or adoption, mostly at the customer- and firm-level. To the best of the author’s knowledge, this paper is the first attempt to demonstrate varying values of established corporate loyalty programs’ growth on sales across stores using longitudinal analysis. Specifically, it examines store-specific factors that foster or lessen store sales performance of those corporate loyalty programs for the entire database to provide insights on competitive advantages from a store perspective. Lastly, this study responds to a call for more nuances of the firm’s performance, especially in a firm-specific industry in activities of the continuous loyalty program.

Methodology: Based on the Resource-based view framework, this study investigates the short- and long-term lagged effects of the corporate loyalty program growth and average store revenue relationship for 1,162 retail outlets of the giant U.S. stationery and office supply company, Staples. Data over 39 monthly periods have been used to build the Vector autoregressive (VAR) model. Further, the Ordinary Least Square (OLS) regression is utilized to examine the contingency effects of heterogeneous characteristics for individual stores.

Results: The findings empirically demonstrate a variance within the same corporate loyalty program across the firm’s outlets and, the effect of each program's attributes on average store revenue varies over time. The sale effectiveness of in-store enrollment growth is not positively significant again until six months later, while those of reward mix growth have alternating patterns throughout the time horizon. Moreover, this dynamic effectiveness is disproportionately contingent on the store’s competitive intensity, the age of the store, and manager tenure, leading to the differences in sales performance across outlets within the firm.

(6)

growing a reward usage of members increase store revenue performance. Incorporating with store characteristics, managers could enhance or mitigate the lagged effects, which allow them to have more control in their own store’s destiny. In the bigger picture, the result helps the firm employ an appropriate policy to adopt for newly-opened store locations and revise for older stores. Theoretically, it provides initial evidence of the varying value of disaggregated corporate loyalty program attributes on the financial performance at the store-level, which does not reflect the classical positive effects in previous literature. Further, store characteristics in combination with corporate loyalty program could be considered as a unique source of sustainable competitive advantages, and thereby a tailoring store-level strategy is needed to achieve different loyalty program’s value proposition.

(7)

1 INTRODUCTION

Loyalty programs have become increasingly widespread among retailers because retail strategies are now geared towards a customer-orientation, rather than focusing on selling products (Sheth et al., 2000). A company wants to go beyond a single transaction to long-term customer relationship to achieve customer loyalty, sales revenue (Gupta et al., 2004), the share of wallet (Verhoef, 2003), market share (Ferguson & Hlavinka, 2006; Leenheer et al., 2007) and ultimately long-term competitive advantage (Wierenga & van der Lans, 2017). McKinsey's research indicates that about half of the ten largest U.S. retailers have introduced loyalty programs, and the growth rate is increasing, on average, 9 percent per year (Huang et al., 2018). The adoption rate is also similar among retailers in the United Kingdom and The Netherlands (Cigliano et al., 2000). In 2018, The Loyalty Management Industry estimated that the loyalty program worth $2,133.2 million and is expected to hit $7,126.8 million by 2026 (Fortune Business Insights, 2019). Although such extensive program investment in many firms is initially often driven by competitive parity and consumer needs for better deals (Meyer-Waarden & Benavent, 2006), the loyalty program serves more than just a defensive strategy of the firm.

General evidence has been provided that loyalty programs have a positive effect on attitudinal and behavioral loyalty at the customer level (Leenheer et al., 2007; Lewis, 2004; Meyer-Waarden & Benavent, 2006; Sharp & Sharp, 1997; Taylor & Neslin, 2005). An attitudinal loyalty makes customers often consider a particular brand in a purchase decision (Dick & Basu, 1994). Even more, loyalty program often leads to behavioral loyalty of customers reflecting in more frequent purchase cycles, larger amount spending (Taylor & Neslin, 2005), and positive words of mouth (Miguéis et al., 2012) despite competitive efforts to cause a switching behavior (Yi & Jeon, 2003). Further, loyalty program enhances the relationship of current customers by increasing the consumer lifetime value (Ou et al., 2011). Ultimately, these customers’ loyal behaviors assumingly lead to better firm performance (Chaudhuri et al., 2019).

(8)

2006) and even lower rate of 18 percent found in another study (CodeBroker, 2018). In line with Gordon and Hlavinka (2011) that one-third of loyalty program rewards issued in 2010 remained unredeemed with a value of $48 billion. Overall, a sales contribution from the loyalty program to the firm can range from 0 to 100 percent across customer segments (Liu, 2007).

Given an inconsistency in the existing literature and 90% of current concern of retail executives over a slow growth with substantial investment in the loyalty program (KPMG, 2016), store managers as a key player on the front line may face with operational challenges and be reluctant in strategies whether there is a true value of growing a loyalty program to their stores. Moreover, most of current literature has been conducted at either customer- or firm-level, mainly focusing on the early stage of the program (Table 1). The previous findings then provide a limited answer of how per-store established loyalty program enrollments and engagement interact with store factors influence store sales and contribute a success or failure to a corporate loyalty program. This study attempts to reconcile prior findings on loyalty program contributions and provide more nuance theoretical and managerial insights on the value that such programs can bring to a store and ultimately to the firm. To answer this question, it will be presented and tested at the macro-level (i.e., multi-outlet) evaluation of the benefits of having a loyalty program by demonstrating the dynamic effect of program enrollment growth and engagement growth on average store revenue. Further, store heterogeneity will be incorporated to identify a moderating effect of store-specific characteristics on the sales effectiveness of such programs.

1.1 Problem statement and research questions

A large body of empirical studies in marketing and consumer behavior has shown that membership enrollment (Leenheer et al., 2007) and promotion inside loyalty programs can lead to substantial increases in customer engagement in the form of customer spending (Drèze & Nunes, 2011; Wagner & Hennig-Thurau, 2009) and thus leading to the firm profitability. However, it is unclear to what extent these can be translated into a store-level practice. Specifically, how and

when the growth of these corporate programs creates financial gains at the store-level, considering

extensive heterogeneity among firm’s outlets.

(9)

across firm variance among loyalty programs to evaluate their success, there might be substantial within firm variance as well. Next, most of the current literature focuses on the program at the launch or adoption such that research that examines how the loyalty program can improve sales performance at the maturity phase is scant (Bolton et al., 2000; Chaudhuri et al., 2019). With significant costs of loyalty program maintenance (Bijmolt & Verhoef, 2017), managers may be reluctant with their strategy, whether growing this program brings competitive advantages over disadvantages. Further, most of existing studies on loyalty programs are cross-sectional or at least have a short time horizon. This study aims to fill these research gaps and bring a store perspective on how and when the growth of mature loyalty programs can be beneficial or harmful to sales effectiveness at different stores within the firm over the time horizon.

To address these gaps efficiently, this study identifies two critical attributes of loyalty programs. By decomposing them into in-store enrollment (enrollment), and the ratio of member’s expenditure to a total transaction (reward mix), different effects on the store’s revenue contribution can be separately examined. The former will capture how attractive the loyalty program is in getting new members to the store while the latter, which separates a store engagement into those of members versus nonmembers, essentially captures how good the loyalty program is in the eyes of members. Thus, this study intends to answer two main research questions (RQs).

RQ1: How do loyalty program attributes affect an average store revenue and when such attributes bring a financial gain to the store?

RQ2: How do store’s characteristics influence the relationship between these loyalty program attributes and store revenue?

The first research question investigates the dynamic effectiveness of corporate loyalty program attributes at a store in both short- and long-term, while the second research question accounts for store heterogeneity across outlets within the firm.

(10)

marketing efforts in loyalty programs, this model also accounts for the possible marketing feedback loop of sales and the decisions of loyalty program growth through a system of equations, enabling a basis for comparison (Bezawada & Pauwels, 2013). Hence, the relationship over time between one variable to another is simultaneously revealed. Lastly, the long-term consumer response is the main focus of any marketing actions that aim to create a sustainable competitive advantage. With that, the lag structure of the VAR model makes it easier to examine long-term marketing effectiveness than multivariate models in specifying a series of own- and others- past (Hanssens et al., 2001). Thus, this approach has been increasingly used in the recent marketing studies due to its structural advantage using system’s approach to the market response (Dekimpe & Hanssens, 2004; Horváth et al., 2001; Nijs et al., 2001; Pauwels et al., 2002; Srinivasan et al., 2004).

1.2 Contributions

Loyalty programs present an important facet of marketing actions for the firm. Within this context, the findings are of relevance to both academics and practitioners. The results reveal multiple previously unknown aspects of corporate loyalty program effectiveness at the store-level. First, the results show that new enrollments to the program may only provide a partial picture of loyalty program success. A new metric that jointly looks at loyalty members to total customers (through reward mix) offers additional insights into the true value of loyalty programs. Second, the author shows that the effectiveness of a loyalty program varies across stores over time and store-specific characteristics such as regional competition, store age, and manager’s experience.

(11)

1.3 Structure

(12)

2 THEORETICAL FRAMEWORK

In the following, a theoretical background, current studies, and proposition development on the main relation, as well as their short- and long-term effects using lagged variables will be discussed. Next, the author discusses the potential moderators and the control variables in the model. Figure 1 graphically represents the conceptual framework. Based on the dynamic response of customers, the author formulates the model among variables with two loyalty attributes’ growth that represents how attractive the loyalty program (enrollment) and how good loyalty program (reward mix) is to the store’s effectiveness (total sales). Further, three store-characteristics are evaluated as a contingency effect, namely the competitive intensity, age of the store, and store manager tenure. A detailed explanation will be given below.

[Insert Figure 1 here]

2.1 Resource-based view framework

There have been a marketing strategy and marketing management literature connecting Resource-based View (RBV) and resulting capabilities on firm performance as a source of its sustainable competitive advantage (SCA) (Kumar & Reinartz, 2018). The latest version of RBV refers to the notion that a firm comprises of a set of resources spread heterogeneously through industries to create a unique, sustainable competitive advantage. With regards to the RBV framework, successful retail should possess a collection of resources within the firm that is VRIO in nature; Valuable, Rare, Inimitable, and exploitable by the firm’s Organization (Barney & Hesterly, 2010).

(13)

attractive offers remains challenging for many firms (Kumar & Reinartz, 2018). Thus, a limited number of firms are capable of accessing it to gain a positional advantage over their competitors.

Moreover, an accumulation of member information obtainedover time by the firm makes the loyalty program to be imperfectly imitable in which it is costly for competing firms to acquire or develop. Additionally, resources can be imperfectly imitated due to certain factors such as casual ambiguity and complexity. Lastly, successful loyalty program pertains to organizational

processes, policies, and practices that, in turn, impact resource efficiency (Barney & Clark, 2007).

With that, corporate loyalty programs are perceived as long-term resources and capabilities that allow stores within the firm to develop and utilize differently their customer knowledge in terms of customer needs and behavior (Reimann et al., 2010), which contributes to revenue streams and eventually store performance. Specifically, loyalty programs provide the store with customer data that can be integrated into customized reward deals to satisfy individual consumer’s needs. Hence, the RBV framework is adopted in this study to examine the store’s sales effectiveness of a loyalty program growth and a variety of contingency factors among multi-outlets within the firm whether they can alter the effectiveness of these growth efforts.

2.2 Effects of loyalty programs on store revenue

(14)

Related to the notion of RBV, loyalty programs could be considered as a valuable resource to grow a store's SCA and create a long-term mobility barrier for new entrants (Aaker, 1996). This is because attitudinal and behavioral loyalty can reduce customers’ responsiveness to competitive offerings and also lock in present customers (Leenheer & Bijmolt, 2008). For loyalty programs to be successful, consumers first need to adopt and later engage them. Thus, two attributes, namely enrollment and reward mix, are considered as separate important elements to access the success of the programs. Firstly, stores can expand their loyal customer base by increasing the number of loyalty program enrollment using attractive marketing campaigns. The key advantage a membership enrollment could bring to the store is customer knowledge enhancement, which strongly related to customer acquisition, retention, and thus profitability (Stahl et al., 2012). Once a customer becomes enrolled in a program membership, stores are able to access and track a certain data regarding personal information such as demographics and specific transactions in order to understand customers’ actual needs and wishes at individuals (Bolton et al., 2000; Leenheer & Bijmolt, 2008). Hence, with times, stores would be better off in providing personalized offerings (e.g., up-sell, cross-sell) and further enhance their sales figures.

By growing the number of new customer enrollments, stores then build on their loyalty programs resource stocks through insights gained from those customer data such as (email) address of the customer and detailed purchasing information. As such, the more customers participate in loyalty programs, the better stores can possess individual customer knowledge and gain subtle insights from it. Getting to know individual preferences further enables effective differentiation in the store’s marketing activities at the individual level (Mauri, 2003). Subsequently, stores are likely to be better in targeting customers, which in turn increasing revenue within stores. However, when little to no effort is put to adopt a free membership in most loyalty programs, it is unclear whether customers would keep up with the stores (McCall & Voorhees, 2010). The members are, on average, actively using only half of those loyalty programs they enrolled (Ferguson & Hlavinka, 2006). Hence, increasing membership rates may or may not imply the sales effectiveness.

(15)

Taylor & Neslin, 2005) because it enhances a value perception in economic aspect and beyond (Morgan & Hunt, 1994). The loyal behaviors of the members can be expressed through an increase in purchase frequency and volume, cross-buying, less price-sensitive, advocating firm with positive word of mouth (WOM), and strong identification with the firm (Berman, 2006; Bijmolt et al., 2010; Oliver, 1999).

At a firm level, retail loyalty programs lead to a fragmentation of members and nonmembers in the customer base. Kumar and Leone (2010) underline that loyalty members have a stronger business relationship with firms than nonmembers. Specifically, members pay price premiums and spend more than nonmembers on average (Drèze & Hoch, 1998; Leenheer, 2004; van Heerde & Bijmolt, 2005). A study by Bolton, Kannan, and Bramlett (2000) in the financial sector, pointed out that the loyalty members extensively use their credit cards than nonmembers. In the U.S. grocery setting, Meyer-Waarden and Benavent (2009) found that loyalty members’ expenditure increases substantially because members maintain a greater share of spending in a specific focal outlet. Taylor and Neslin (2005) further elaborate that a loyalty program is successful in increasing sales through two underlying mechanisms of membership point pressure in the short-term and rewarded behavior in the long-term. In sum, there might be a difference in marketing response between members and nonmembers. Given that loyalty program customers have a greater motivation to purchase than those of nonmembers (Lal & Bell, 2003), they are thereby likely to contribute more to the firm revenue.

In line with the previous findings, it seems logical that the larger proportion of transactions made by members per overall store transaction (i.e., reward mix) would lead to the larger total sales of the focal store. Moreover, the percentage of transaction size from members is considerably important to the store’s marketing decision. The transaction details of members will also enrich the database and eventually used for precisely targeting members such as via direct mailings (Blattberg et al., 2008). This again, in turn, increases average store revenue within the firm.

2.2.1 The short-term and long-term effect

(16)

reward system before adjusting their behavior (Drèze & Nunes, 2011). Meyer-Waarden and Benavent (2008) show that after six to nine months of the membership, the shopping behavior of customers starts changing obviously. Thus, success or failure in the short-time-interval provides a limited foresight for a whole loyalty program and thus may not give much insightful managerial implication (Liu, 2007) such as for resource allocation (Dekimpe & Hanssens, 1999).

Similarly, the effect of marketing actions in the loyalty scheme is not limited to that period but still remains for a number of future periods before the effect dies out. For example, there is evidence that a purchase volume of loyalty members appeared to increase just right after the loyalty card adoption and decrease during subsequent periods (Demoulin & Zidda, 2008). This may occur due to the initial impulse buying and stockpiling, which similar to the sales dip phenomenon after price promotion. Moreover, literature also found the customer loyalty inertia where members still patronage a focal store without thinking (White & Yanamandram, 2004) and a rewarded behavior where customers keep spending in order to benefits from the next rewards (Taylor & Neslin, 2005). These thus impact the program effectiveness when comparing dynamic effect overtime.

Moreover, current literature indicates the mixed effects of the loyalty program. Some literature found little (Verhoef, 2003) or no obvious effect on purchase behavior as a proxy of the general effectiveness of loyalty program (De Wulf et al., 2001; Mägi, 2003) while others found positive effects (Lewis, 2004; Liu, 2007; Meyer-Waarden, 2007, 2008). Furthermore, the research also indicates differences between short- and long-term sales effects of loyalty programs such as positive short-term changes but not in the long-term (Meyer-Waarden & Benavent, 2009) or both short- and long-term effects on profit gains (Chaudhuri et al., 2019).

Based on the aforementioned mixed effects, the author argues that both loyalty program attributes can favor both positive and negative effects over the time horizon. To account for different evolutions in sales effectiveness of loyalty program attributes, this existence of different immediate and delayed effects, as well as their own past, will be evaluated by using a lag operator to capture dynamic response function in the model (Dekimpe & Hanssens, 1999).

(17)

effect and put more weights on the actual empirical evidence. Therefore, the author purposes the following:

P1: An increase in the number of in-store enrollment leads to an increase in average store

revenue.

P2: An increase in the reward mix leads to an increase in average store revenue.

2.3 Moderating effects on the relationship between loyalty programs and store revenue Given the fact that different store possesses different characteristics as a unique resource, these factors interact with loyalty programs could enhance or hinder sales effectiveness. Thus, in this section, the moderating role of a variety of store heterogeneity will be discussed.

2.3.1 Competitive Intensity

The loyalty program is a potential marketing tool for retails to create a strategic differentiation (Nastasoiu & Vandenbosch, 2019), distinguishing themselves in a way that is both uniquely valuable for consumers and costly to imitate by surrounding competitors (Porter, 1997). Luo and Homburg (2007), however, indicate that the degree of competition intensity might negatively affect the relationship between customers and a firm. Thus, it is likely that the firms will gain positional advantages when fewer to no competitors operate such programs. On the contrary, many competing loyalty programs available in the market make perceived benefits of loyalty programs less distinctive as customers can easily benefit from comparable rewards (Dowling & Uncles, 1997; Liu & Yang, 2009). Hence, the stores are less likely to fully benefit from loyalty program revenue due to the likelihood of customers adopting more than one membership.

(18)

where rivals actively provide comparable marketing offers. Subsequently, the customer purchasing behaviors may also spread over the market.

A market competition might also reduce a switching cost in a way that allows enrolled customers to move freely from one store to another to maximize their benefits (Feng et al., 2016) so that members are not exclusive for the focal store. This situation leads to widespread spending across stores and brands, which makes the expected cash flow of the focal store become uncertain (Srivastava et al., 1999). Therefore, the effect of loyalty program enrollment on the focal store’s sales is expected to be diminished and not fully advantaged if they located in such high competition areas. In the same vein, the intensity of market rivals potentially moderates the relationship between the reward mix and the sales of stores. Because the competition may reduce the perceived value proposition of the loyalty program through ‘cherry-picking’ opportunities (Leenheer et al., 2007) such that the same benefit for a member may not be considered as good as before. Therefore, customers have a high likelihood of visiting other stores and not remains loyal, resulting in more variance in the expected return of the focal store.

P3a: The competitive intensity negatively moderates the effect of in-store enrollment on

average store revenue to the extent that stores located in a more competitive environment will experience less benefit from enrollments in the loyalty program.

P3b: The competitive intensity negatively moderates the effect of reward mix on average

store revenue to the extent that stores located in a more competitive environment will experience less benefit from member engagement in the loyalty program.

2.3.2 Age of Store

(19)

Understanding customer beyond a transaction to create customer engagement value over a long period of time helps stores to gauge effective marketing campaigns for attracting new customers and retaining the existing ones (Kumar, Aksoy, et al., 2010). Therefore, the well-established stores are expected to have a unique competency for a loyalty program to be leveraged better in terms of marketing strategies compared to the younger ones. For example, with data collected over time from several buying cycles, the established store can draw a critical insight from numerous customer information and leverage in store’s pricing strategy, marketing mix, store’s promotion activities, and customized offerings and services to activate their enrolled customers. Such that it helps the effect of more customer enrollments result in more store revenue.

In the same vein, the established store should have better product selection, assortments, store layout, and benefits package through several years of trials and errors, which can improve the store’s value proposition (Dowling & Uncles, 1997). With more accurate product selection and the opportunities to conduct competitive analysis, it is not hard to resemble the members’ value. Further, a store with more experiences would be better able to create a higher customer involvement and enhance members' satisfaction, which could moderate the reward mix and sale effectiveness (Yi & Jeon, 2003).

(20)

Thus, the author proposes that store age may have a positive effect on the sales effectiveness of the loyalty program. The opportunity to experience more purchasing cycles and have data collected for a longer time provides more insight for the store to facilitate the effective loyalty program to the customer via marketing mix, product selection and assortments, which leads to the higher store income. Even though the younger store has an advantage of unsaturated satisfaction from customers, which will make the loyalty program more effective, the author expects that the effect of the unsaturated satisfaction cannot overcome database and marketing insights.

P4a: The age of the store positively moderates the effect of in-store enrollment on average

store revenue to the extent that older stores experience greater benefit from enrollments in the loyalty program.

P4b: The age of the store positively moderates the effect of reward mix on average store

revenue to the extent that older stores experience greater benefit from member’s engagement in the loyalty program.

2.3.3 Manager Tenure

The existing literature provides mixed effects of top manager experience on store performance. Some findings reveal that managers are willing to take a risk at the beginning of their tenure and actively learn from strategic experimentation (Finkelstein & Hambrick, 1990; Luo et al., 2014). The managers are then less flexible and more risk-averse as time passes. They might become less motivated in engaging in their role, which has a negative effect on overall performance (Levinthal & March, 1993).

(21)

Hence, store managers with more years in the market should be able to uncover and accumulated information about past decision and informative market-specific idiosyncrasy as growing their tenures (Yang, 2019). The multiple cycles of service and marketing processes by a manager can boost sales revenues by attracting new customers and adding value for existing customers through increasing repurchase and customer loyalty (Hallak et al., 2018). Therefore, experienced managers tend to learn better from previous marketing success and failure, employ appropriate strategies, and make better marketing decisions that appear to influence the whole process of loyalty diffusion.

A study by Allaway, Berkowitz, & D’Souza (2003) found that fewer retails effectively convey the benefits that consumers will earn when they adopt and use the loyalty card. Thus, a longer-tenured manager should be better off in implementing and leveraging the effective strategic marketing campaign (Arnold et al., 2009) to get enrolled customers come actively spending in their store. Likewise, the manager with more experiences should be able to communicate to and stimulate reward members to increase a spending level via cross-selling and up-selling. Effective communication is crucial to underline a privileged status and value of the relationship to the loyalty members (Shugan, 2005). As such, a manager with more tenure years should strengthen the effect of in-store enrollments and the effect of the ratio of reward member’s expenditure to total transactions on average store sales. Thus, this led to the following proposition:

P5a: The manager tenure positively moderates the effect of in-store enrollments on average

store revenue in a way that stores with a longer-tenured manager lead to greater benefit from enrollment in the loyalty program.

P5b: The manager tenure positively moderates the effect of reward mix on average store

(22)

3 METHODOLOGY

In this section, the research methodology will be presented including the data description, measurement, model specification, estimation, and validation of the results.

3.1 Data description

The data were provided by the giant U.S.-based stationery and office supply retailer, Staples. It is among the first company in their industry (Neighbor, 2013) to provides a basic loyalty program for customers with no annual fee. The members can benefit from the loyalty program by 1) Having up to 5% cashback, 2) Free delivery with a minimum purchase, and 3) Reward from the recycle ink cartridge. Members will receive rewards earned for every purchase, and then the rewards will be converted to reward coupons (i.e., rebate check) issued in the following month. For example, after customers place an order today, a rewards coupon will be issued at the end of next month and available for the whole month after next with an assigned expiration date. Rewards are paid out in $5 increments, and monthly balances of less than $5 will be expired within a company policy time (Staple, 2020).

A preliminary assessment of data quality is undertaken prior to arrival to ensure an appropriate level of data quality. Thus, the data cleaning part is not directly relevant in this part. The dataset contains store-level data from 1,162 stores operating in the United States, offering both business-to-consumer (B2C) and business-to-business (B2B) customers. For each store, the information consists of the number of in-store loyalty program enrollments, the reward member transactions, and total store transactions in monthly periods ranging from June 2007 – August 2010, totaling 39 observation periods for each store (Figure 2, Figure 4, and Figure 3). Moreover, the loyalty program was in the maturity phase of their program development as it can be noticed from the stability of in-store enrollment and reward mix growth. Additionally, the loyalty program was introduced several years before the data collection and had no changes during the study window with the average store age of 13.58 years.

(23)

descriptions of each variable are presented in Table 2. In total, this resulted in a total set of 45,318 usable observations (1,162 stores across 39 time periods). In addition, an overview of descriptive statistics and correlations for store variables are provided in Table 3 and Table 4. Noting that, the total sales, which represent the store revenue, are the focal dependent variable in this study. The two main drivers are reward mix and in-store enrollment, and other variables are included as moderators and control variables in this study.

[Insert Table 2 here] [Insert Table 3 here] [Insert Table 4 here]

3.2 Measure

All measures were captured using secondary data from Staples. In the VAR system, three variables were constructed; total sales, in-store enrollment, and reward mix as follows.

3.2.1 Total sales (TS)

The total sales (TS), which is the total revenue of a particular store, is one of the most widely-used measures of loyalty program effectiveness (Berman, 2006). Table 3 shows the average number of total sales is slightly right-skewed as its mean ($486,870.4) is a bit larger than the median ($451,104.91). There is a moderate dispersion (S.D. = $185,697) that suggests that there is sufficient variation in the total sales to provide a reliable test. The correlation between total sales and in-store enrollment is weakly positive (r = 0.2823) while total sales and reward mix is weakly negative at -0.2655. Figure 2 depicts our data.

3.2.2 In-store enrollment (EN)

(24)

same direction. The pattern of enrollment overtime seems they have a little seasonal trend, and again the author will account for it in the analysis section. Figure 3 presents the in-store enrollment data.

3.2.3 Reward Mix (RMix)

The ratio of loyalty (reward) member’s expenditure to total transactions (RMix) represents the proportion of the transactions completed by the members of the loyalty program to the total transactions in a given store at a specific month. The higher reward mix indicates the higher percentage of transactions in such months completed by the store members. Deriving such a ratio is the way to leverage the appended raw data to meaningful implied data (Verhoef et al., 2016). Thus, this ratio essentially captures a full picture of the loyalty program quality and market penetration of the program for focal stores. An overview of the reward mix in Table 3 shows the normal distribution with mean (0.6851), while the standard deviation is relatively small (0.0948). Lastly, since it seems a growing proportion of member’s transaction overtime, trend specification will be determined when testing the stationarity. Figure 4 presents the reward mix data.

3.2.4 Control variables

Since the sales figure may also depend on extraneous variables, the author introduces four control variables as covariates in this study, namely years of marketing experience, years of retailing experience, years of sales experience, and the number of marketing degrees.

(25)

3.2.5 Multicollinearity

Before starting the analysis, the author checks for the possible multicollinearity due to strong correlations between retailing experience and sales experience (r= 0.9865, p<.05) using Variance Inflation Factor (VIF) for all independent variables in Table 5. The VIF score obviously turned out to be high for one pair of the variables; retailing experience (VIF = 42.71) and sales experience (VIF = 42.43). Thus, the sales experience variable is excluded to solve the multicollinearity issue. After removing, the remaining variables have VIF smaller than 5 (Leeflang et al., 2015).

[Insert Table 5 here]

3.3 Econometric modeling (VAR Model)

Vector-autoregressive (VAR) is proposed in this study using STATA version 16 for modeling and investigating the sales effectiveness of loyalty program attributes due to two important reasons.

First, the VAR model allows the immediate and delayed time-lagged effect from the previous marketing actions (Dekimpe & Hanssens, 1999), which are expected in the loyalty program setting. Information such as benefits from the loyalty program and membership promotion may not reflect in sales immediately, rather slowly influence store income via customer satisfaction and behavioral loyalty (Bolton et al., 2000). For instance, a loyalty program requires new customers to learn about the reward benefits, which may take a certain period of time (Drèze & Nunes, 2011). Also, due to the nature of the office supplies business, it may take several months before a new buying cycle begin anew. Hence, the author believes that capturing these lagged effects is important and ignoring these dynamics can lead to biased conclusions.

(26)

Base on focal variables and proposed direction, the VAR model can be specified as below. ! "#_! $%_! &'()_!* = ! ,_"# ,_"$ ,_"%* + . / ,_##& _, $#& ,%#& ,_#$& _, $$& ,%$& ,_#%& _, $%& ,%%& 0 ! "#_!'& $%_!'& &'()_!'&* ( &)# + ! 1_#! 1_$! 1_%!*

Where TS stands for total sales in dollars, EN for a number of in-store enrollment each month, and RMix for the monthly reward mix. The subscript t stands for time in months, and p is the lag order of the model, e is the error term, which is normally distributed with means of 0 and variance Σ. In addition, a*+ represent the effects of variables ( in equation 4.

The procedures of the analysis are organized as follows. First, after specifying variables that are included in the model, a unit root is tested to determine if variables are stationary, or evolving (Enders, 2004). Second, determining the optimal lag length (p), based on the information criteria and permitting trades off between proper inference, autocorrelated error (Colicev et al., 2018). Third, estimating the final model. Forth, Granger Causality is conducted to ensure the primary causal relationship between proposed explanatory variables and total sales. Fifth, model assumptions (serial correlation, normality, and stability) are to be checked to ensure the quality of the final model.

3.3.1 Stationary, Unit root test

The estimation begins with the unit root test for the stationary process. This is important to rule out spurious regressions (Engle & Granger, 1987) for the relationship between variables in time series analysis where the assumption of covariance stability overtime must be held. The result of the well-known Augmented Dickey-Fuller (ADF) has the null hypothesis of the unit root, and the alternative hypothesis is that the series is stationary. The result rejects the null hypothesis of non-stationary at a 95% confidence interval for the total sales and 90% for the variables enrollment and reward mix. In other words, the former is stationary at a 5% critical value while the latter two show marginal stationary at 10% critical value. As such, all of them are stationary at their levels before any difference is taken (i.e., covariance-stationary, I(0) processes). Thus, there is no need for further cointegration tests, and it can proceed with the VAR model. The full result refers to Table 6.

(27)

3.3.2 Optimal Lag

There is no universal consensus of choosing the right order for the lagged variables in any case (Wooldridge, 2016). Therefore, the optimal lag length for the VAR(p) model may be determined by the most applied information criteria, such as the Akaike Information Criterion (AIC), Schwarz Bayesian information criterion (SBIC), and Hannan–Quinn information criterion (HQIC). The result (Table 7) shows that all information criteria have chosen a model with seven lags, whereas SBIC has selected a model with two lags.

Due to results inconsistencies, it is helpful to check the likelihood-ratio test (LR) test which gave seven lags orders in our case (Hatemi-J & S. Hacker, 2009) as well as taking into account the residual autocorrelation Lagrange multiplier test to balance between appropriate lags and autocorrelation bias (Breusch, 1978). In other words, with this number of lag orders, the model ruled out the problem of autocorrelation while maintaining lag optimization. Thus, the author chose seven lag orders as a choice of lag length for the rest of the analysis. With seven-time periods are excluded to account for lag order, the following analyses will proceed with 32 remaining time periods.

3.3.3 Granger causality tests

Endogeneity issues in a loyalty program are not trivial because it leads the managers to overestimate their loyalty program performances. At the customer level, the loyalty programs created to boost customer loyalty; however, the best customers of the firm are most likely to subscribe to the loyalty program. Likewise, at the firm level, while loyalty programs influence the sales revenue of the firm, the higher sales revenue might also impact on how managers reinvest in loyalty programs in the next period. Thus, to account for the self-selection bias and potential endogeneity of such programs, the causality can be employed (Leenheer et al., 2007).

(28)

Table 8 shows support for a forward direction of the dynamic relationships in our conceptual framework. That is both in-store enrollment (5$_{= 150.52, p < .000) and reward mix (5}$_{= 141.67,}

p < .000) significantly and strongly Granger-cause total sales. Specifically, the dependent variable

(i.e., total sales) can be better predicted by having past data of IVs (reward mix and in-store enrollment). In addition, the result also reveals the feedback system coming from total sales to in-store enrollments and reward mix as expected. However, the proposed direction fits the data better than the reversed pattern where total sales predict in-store enrollments and reward mix.

3.3.4 Assumptions diagnostics

To statistically validate the model after estimation, the diagnosis of residual assumption violation has been tested to rule out the bias in variance estimations. Firstly, the serial autocorrelation test is applied to make sure there is no systematic bias in the residuals. Then, the normality test has been conducted to ensure the distribution of the residuals. Lastly, the model stability test is utilized to validate the stability of the model.

3.3.4.1 Autocorrelation test

Remaining autocorrelation is mostly a sign of an incorrect number of lags in the model (Wooldridge, 2016). Further, the model that includes finite lag can also have serially correlated errors even if there is no underlying misspecification issue (Leeflang et al., 2015). Therefore, it is important to perform the diagnostic for spatial dependence and spatial heterogeneity. To detect a violation of the assumption, the Lagrange-multiplier test is used for investigating any serial correlation. The result is reported in Table 9. Because the null hypothesis of the Lagrange-multiplier test is no autocorrelation at all lag order, the test fails to reject the null hypothesis and can be concluded that there is no autocorrelation with 95% confidence. Hence, the model assumption regarding serial correlation is satisfied.

3.3.4.2 Normality

(29)

The result provided in Table 10 indicates that the null hypothesis cannot be rejected. Therefore, the model does satisfy the normality assumption.

3.3.4.3 Model Stability

Assuming the stability condition does not hold, it implies that the variables entered in the VAR system are probably non-stationary (Wooldridge, 2016), and thus the model may need to be re-specified. In eigenvalue stability analysis, the stability condition of the model requires that all the eigenvalues are smaller than one in absolute value on Modulus. The result of the model shows a stable process (Table 11), which assuming that the error terms are independent of the dependent variable, total sales (i.e., E(y0) = 0 given at t starts at 0).

3.4 Moderator analyses

Thereafter, the author investigates the moderating effects using the procedure applied in previous literature (Bezawada & Pauwels, 2013; Dekimpe & Hanssens, 1999; Pesaran & Shin, 1998). After obtaining the impulse response function, the parameters are further examined. Three moderators, competitive intensity, store age, and manager tenure, are used as explanatory variables (Xs), and each store’s VAR parameters are used as dependent variable (Y). The general form of the moderator analysis is presented below.

,_*+& _{= 6}

"+ 6#78!+ 6$#9!+ 6%'"!+ :!

Where ,_*+& _{is lag ; parameter of variables ( in equation 4 in VAR Model. 6}

" is an intercept and 6_, is the slope for explanatory variable <. 78_! stands for competitive intensity at the focal month period =, #9! is store age in months at the focal month period =, and '"! is manager tenure in years at the focal month period =. :_! represents error term, which is normally distributed with mean 0 and variance >$_{. The Ordinary Least Squares (OLS) technique is utilized for obtaining the} moderating parameters which are presented in the results section. An overview of the analysis is presented in Figure 5.

(30)

4 RESULTS AND DISCUSSION

To assess the propositions, two steps of testing are utilized. Although the VAR system yields the equation system, the author will only present and interpret the forward-direction equation of interest.

At the first step, the author estimated a VAR model in Level 1 that only included the main dynamic effects of independent variables (in-store enrollment and reward mix) to support Proposition 1 and 2. Then, parameters at each store are obtained, totaling 1,162 stores with 22 parameters per store. Following this, the author regresses each store the lagged parameters on store factors, namely competitive intensity, age of the store, and manager tenure to assessing the impact of store-level characteristics (Proposition 3a – 5b).

The overall model is significant (p = 0.000) with R2_{= 0.9582. In other words, 95.82% of the}

total variance in average store sales is explained by the lagged series of enrollment and reward mix variables in a model.

4.1 In-store enrollment (P1)

The full result in Table 12 and Figure 6 present the coefficient of lagged enrollments on sales for seven past periods. Of seven lags, three lags, which are L2, L6, and L7, significantly explain the previous enrollment effectiveness of average current store sales. The lagged of enrollment does not drive an increment in sales until the sixth period. Afterward, the marketing cycle begins again. The results in short- and long-terms are analyzed as follows.

4.1.1 Short-term effect of in-store enrollment

(31)

One possible explanation would be relevant to a post-enrollment dip phenomenon as mentioned earlier and a purchasing pattern of these product categories. Firstly, the effect of the short-term lagged enrollment demonstrates that enrolling new customers in the loyalty program might pull them ahead in the buying cycle, resulting in revenue losses in the next two months following the enrollment. This is because when the customers enroll to the loyalty program, they are usually attracted by financial, psychological, sociological incentives (Rosenbaum et al., 2005) that may lead to an immediate purchase upon visits (Meyer-Waarden, 2007; Yi & Jeon, 2003) such as reward points or free gift promotion. Behavior prior to redeeming a reward may temporarily shock a purchase resulting in a larger purchased quantity than they would do without reward otherwise. Thus, after initial enrollment, a detrimental effect on their next purchase periods would be expected. Put differently, the lagged number of enrollments potentially creates a post-enrollment dip in sales in the following months, lowering the average revenue of the stores in the short run (Liu, 2007).

Another explanation is regarding the nature of buying behavior in product categories of stationery and office supplies. Consumers usually do not purchase these products heavily on a regular basis but rather on a periodicity such as during the back-to-school-period in July to September (Zmuda, 2009). Stationery and office supplies such as copy paper and ink cartridges are clearly a long-lived product and not easily perishable. These products, by nature, can last for several months so that it is expected to have a pause from customers before repurchasing the products again. As such, customers may not have many opportunities to express their behavioral loyalty (i.e., through more purchase occasions), which affects the loyalty program's effectiveness (Kumar & Reinartz, 2018).

4.1.2 Long-term effect of in-store enrollment

(32)

Overall, the in-store enrollment drives sales in the focal time period because the more members register for the program 1) the better the store can gain from their purchase, 2) the deeper insights from customer data, and thus 3) the better targeting. However, a positive effect seems not to reappear as positively significant again until six months later, consistent with the study of Chaudhuri, Voorhees, and Beck (2019) that gross profits do not become significant until the second quarter following the participation of the program. Not surprisingly, the nature of the Staples program rewards is that it comprises cashback and is mailed to its members as a rebate check (Berman, 2006). For example, members spent $400 today and they would receive a rebate check $25 in the next month for using in another month. Hence, it takes approximately three months for customers to have an opportunity to learn how the redemption system works. The experiences at the first cycle allow the customer to start altering their buying behavior and returning a value to the store six months later, which would be in the second cycle for their purchases. This underlying explanation is in line with Drèze and Nunes (2011) who demonstrate that learning is a key process for consumers to increase spending in reward programs. Further, the finding from previous literature mentioned that loyalty program participation mainly provides benefits in the long-term perspective through delayed rewards, which members can receive for continuous purchases (Dorotic et al., 2014).

4.2 Reward mix (P2)

The coefficients of the reward mix are presented in Table 12 and Figure 7. Individually, six out of seven lag variables have significant effects on average store sales at a 95% confidence interval, except L3. Hence, the results indicate that the model mostly contains many significant parameters. Taking a closer look at Figure 7, it can be noticed the alternating patterns of up and down trend of coefficient towards average total sales during the observation periods. Specifically, starting at L1, as the reward mix increases, the stores experience a dramatic drop in total revenue and a rise in the subsequent month. Thereafter, it dips into the negative in the fourth month before increasing again to its peak in the fifth month, dropping in the subsequent month six before increasing again in the seventh month.

(33)

4.2.1 Short-term effect of reward mix

Table 12 shows that an overall net effect of the reward mix is positive in the short run. There are more significant and positive effects of the lagged numbers of reward mix on current total store revenue as presented in L1 and L2, which supports P2. Specifically, a one percent increase in reward mix in the past one and two months led to a decrease of $812,471.1 and an increase of $1,626,163 in the average store revenue in a focal month, respectively. However, L3 is not significant and thus the interpretation is not meaningful.

This result finds general support within loyalty program literature in terms of increased spending levels in the short run as members usually visit stores more often than nonmembers and, on average, spend more than nonmembers who buy (Leenheer, 2004; Liu, 2007). Further, the result implies that design of loyalty program creates an immediate incentive to buy and temporary commitment among members resulting in a spike in sales when customers raise their level of purchase to qualify for a reward (Kivetz et al., 2006).

4.2.2 Long-term effect of reward mix

Overall, in the long-run, the past percentage of transaction size by loyalty members significantly and negatively explains the store revenue in the current period at 95% confidence interval (see Table 12), which does not support P2. Individually, an additional percentage of reward mix in the last four, five, six, and seven months significantly impact store revenue of the focal month by -$2,757,019, $2,476,556, -$2,486,979, and $1,517,614 respectively. The most likely reason to see a surge and drop pattern in the effect of the reward mix can be explained by the fractions of purchasing patterns between members and nonmembers.

(34)

nonmembers are, however, lower and fluctuate over time. Thus, an increase in reward mix that contributes to a lower sale mainly resulting from the dramatic change in the revenue stream of nonmembers.

This indicates that during surge periods with scant income from nonmembers, the store revenue decreases. The possible explanation of stable members’ transactions is that the reward program in terms of rebate check with an expiration date successfully keeps members doing business with the store and redeeming for a benefit (Dorotic et al., 2014). On the other hand, the nonmembers who do not have such point pressure and rewarded behaviors do not regularly visit the store and hence influence the change in the store revenue. This aforementioned phenomenon is supported by the empirical research stating that store revenue can be influenced by nonmember customers (van Heerde & Bijmolt, 2005).

4.3 Moderator analyses

After analyzing the main effect, the author further investigates a moderating effect of store contingency. In this study, the loyalty program attributes are at the corporate level with an average store’s total sales. However, the store characteristics are exogenous at the store level. Thus, we could not directly include the moderating effect on the main model. Instead, we regress the obtained parameters of each store on the moderator variables (competitive intensity, age of the store, and manager tenure). The full results of moderating effects are available in Table 13 and Table 14.

4.3.1 Moderating effect on in-store enrollment effectiveness

4.3.1.1 Competitive intensity

(35)

The interpretation of the interaction effect indicates that one more percent increase in competitive intensity have strengthened the effectiveness of enrollment on L1 and L2 by $59.4317 and $293.3983 at a 95% confidence interval. Likewise, in the long run for L4 – L7, one percent increases in competitive intensity has changed the enrollment effectiveness on store revenue by $80.9484, -$63.9511, -$203.1026, and $313.1534 at 95% confidence level. Therefore, the stores located in high competition areas will financially pay off in terms of revenue from the effect of both immediate and delayed enrollment effectiveness.

This implies that the enrollment of a well-known store could create high switching costs for store members. As Staples is the top three largest stationery and office supplies brand, rewards from being membership with such superstore could be substantially attractive to lock-in members. For example, a well-known store implies good fit loyalty programs and a strong sense of belonging among members (McAlexander et al., 2002). Customers, therefore, assume that they benefit more by staying loyal to their store than by switching to another store (Demoulin & Zidda, 2008). Plus, in the high competition areas, there is also a greater probability of customers. A loyalty program allows the store a greater share of those customers and keeps them loyal. Furthermore, a well-known store with a diverse customer base can benefit highly from the loyalty program’s opportunities for customer differentiation. Instead of treating all loyalty cardholders similarly, a retailer could better target promotions and rewards to specific customer segments. As a result, the higher competitive intensity, the better for the focal store because it will create a larger brand differentiation between the focal store and others. Thereby, the impact of competitors on the lagged effectiveness of enrollment is minimized or even positive to the focal store.

4.3.1.2 The age of the store

(36)

those from L4, L5, and L7 are decreased significantly by $0.4866, $0.41063, and $0.2337 respectively.

The possible explanation is that in the mature industries such as stationery and office supplies, customer adoption and purchasing pattern are relatively unchanged and thus easily understand and predict (Audretsch & Feldman, 1996). Therefore, with more data available on hand, the older stores may not have an additional advantage than the younger store in attracting and stimulating new members’ behaviors. Specifically, as mentioned that the older store may experience a diminishing marginal utility in relying solely on enrolled data in their marketing mix (e.g., segment and target members with tailored incentives). If everyone knows the older store, then the old store may face diminishing returns as gaining additional loyalty is difficult. In addition, attaching to the relatively static demographic data, the older store is likely to fail in adapting to the market change. Hence, the older store may not accurately resemble the enrolled customers’ value and thus weaken the sales effectiveness of enrollment as a result.

In contrast, the youngest store in this dataset is almost four years which implies that preliminary data may be sufficiently variant in extracting an insight to encourage new members’ purchase. On top of that, the younger stores may find themselves to be more attractive in terms of newness (e.g., pleasurable shopping environment), which can be considered as a value-creating asset. Hence, the younger store marketing action seems to enhance hedonic perceived values in the eye of new members. This attitude can impact anticipated future consumption with the store (O'Curry & Strahilevitz, 2001), increasing the enrollment effectiveness in-store sales.

4.3.1.3 Manager tenure

(37)

effectiveness on current store revenue is reduced by $7.3523. Hence, the more job experience manager has, the smaller sales effectiveness of past-month enrollment in the focal period will be.

On the other hand, the results indicate that the total carry-over effect of enrollment between four to seven months (L4-L7) have positively and significantly moderated by the manager tenure. One more additional job experience of manager changes the effect from long-term lagged enrollment by -$2.9373, -$2.2208, $2.3875, and $6.7202, respectively. Therefore, the long-term effects of enrollment become stronger as manager tenure increases.

The unexpected result in the short term can be explained by the fact that store managers who have more years of working experience may fail to proactively adapt in how to communicate loyalty program’s value to new customers. This loyalty program communication is considered to be vital to make customers feel special and lead to loyal behaviors (Morgan et al., 2000). Rather, experienced managers tend to be passive and commit to their own prior practice in general (Salancik, 1977). With a status quo-oriented, the advertising campaign and communication styles produced by long-serving managers may not be catchy, and thus new customers may not find it to be attractive. Hence, a store may not fully financially benefit from having new customers enrolled as managers with more tenure significantly diminish the enrollment effectiveness of the store.

However, long-tenured managers tend to have persistent, unchanging strategies (Finkelstein & Hambrick, 1990). With that, a communication and marketing activities are rather consistent overtime. Prior research shows that consistent marketing communication has a greater effect on cultivating loyalty in the long run (Navarro-Bailón, 2012). Maintain consistency in marketing message and communication about the loyalty program facilitate customers in recognizing the brand easily and consolidating information they received (Smith et al., 1999). Thus, customers would become actively loyal to a particular store when they start looking for items to purchase. Ultimately, after a certain period of time, the manager tenure becomes positive to the sales effectiveness of lagged enrollment.

4.3.2 Moderating effect on reward mix effectiveness

(38)

addition, the age of store and manager tenure positively moderate the effectiveness of reward mix to store revenue in the short run, and the opposite effect is true for the long run. The detailed explanations are as follows.

4.3.2.1 Competitive intensity

Overall, the net moderating effect of competitive intensity does not support P3b. The author finds that the short-term net effect of competitive intensity is significantly negative to the relationship between lagged numbers of reward mix and store revenue based on the coefficients of L1 and L2. The competitive intensity negatively moderates the reward mix effectiveness at a 95% confidence interval. Specifically, one percent increase in the competitive intensity, the benefit of reward mix during the past one, and two months on store revenue at the focal month changed by $113,795 and -$460,214.10, respectively. The result indicates that the stores located in a more competitive environment will experience less short-term sales effect from the reward mix.

On the other hand, the cumulative long-term net effects turned out to be positive and significant at a 95% confidence interval. One percent increases in competitive intensity has impacted the benefit of reward mix in the past four to seven months on the current sales by -$100,794, $67,984.62, $208,143.7, and -$107,246.2 respectively. This means that, in the long run, the store sales will be, on average, even more benefited from the past reward mix effectiveness given high competition circumstances.

(39)

as well as psychological and social benefits that could dilute the effectiveness of the competitive intensity. Moreover, the market leaders often possess competitive advantage in terms of economies of scale and brand equity making smaller brands suffer from a Double Jeopardy effect, that is the small brands sell less at a lower price with a fewer customers (Ehrenberg, Goodhardt, & Barwise, 1990). Hence, in the long term, the higher competitive intensity will favor the focal store.

4.3.2.2 The age of the store

As proposed in the earlier section, it seems that the moderating effect of the age of store on the reward mix effectiveness is mixed; therefore, the net results do not directly support P4b. In relation to the short-time interval, the age of the store has a net positive moderating effect on the effectiveness of the reward mix. Specifically, L1 and L2 are statically significant at 95% confidence level. One additional month of store age changes the effectiveness of the reward mix from the last one and two months by -$432.5123 and $1,221.488,respectively. Hence, in the short term, the older store gains more current financial benefits from the past reward mix than the younger ones.

The long-term net effect of the store age has statistically and negatively moderated the relationship between reward mix and total sales at a 95% confidence level. An additional month of store age has changed the benefit of reward mix to the store revenue by -$2,422.436, $ 736.1957, -$1,215.632, and $482.5458, meaning that the effect of past reward mix to the store revenue become weaker, in the long term, when stores are getting older.

(40)

Furthermore, due to the strong historical transactional data, the older store may continue to stick to their current status quo – not changing or seeking new products and retaining the current product assortments and layouts. In contrast, a preliminary database that the younger store obtains allows the store to be more flexible to the market change. Specifically, every new data collected, a relatively small database of the younger store has changed substantially. Also, the younger one tends to accept new items in order to survive and make growth (Dawar & Frost, 1999). Therefore, in the long term, the younger store may actively adapt for better product assortments and selections that satisfy members’ needs, enable them to provide better program quality to the members. Therefore, the younger store age strengthens the sales effectiveness of the reward mix compared to the established ones.

4.3.2.3 Manager tenure

Generally, the moderating effect of manager tenure on the effectiveness of the reward mix has supported the proposed effect on P5b, only in the short run. Given the short-term lagged path, the effectiveness of the reward mix positively and significantly moderated by the years of manager tenure. The L1, L2, and L3 effects have changed with an additional length of store manager tenure by $34,118.28, -$3,364.887, and -$13,865.92, respectively. In total, the stores with a greater number of years manager have gained more benefits from the effect of short-term past reward mix. For the significance in the long run (L5 – L7), the total moderating effect becomes negative on average. An additional year of manager tenure alters the effectiveness of the reward mix by -$16,125.44, $13,984.64, and -$23,233.42, respectively. These results point an interesting interaction result that the stores tend to experience more current revenue gains from the delayed reward mix effectiveness when they hire a store manager with fewer job tenures than those who have a lengthy job tenure.

(41)

Investigating the value of corporate loyalty programs at the store level.