The Effect of Loyalty Program Divisibility on Consumer Purchase Behavior
XAVIER DRÈZE JOSEPH C. NUNES*
* Xavier Drèze is Assistant Professor of Marketing, the Wharton School of the University of Pennsylvania, Philadelphia 19104-6340. Joseph C. Nunes is Associate Professor of Marketing, Marshall School of Business, University of Southern California, Los Angeles, CA 90089-0443.
Both authors contributed equally and are listed in alphabetical order. Questions should be directed to either Xavier Drèze at [email protected] or Joseph C. Nunes at
This research demonstrates how the effectiveness of a loyalty program depends on its divisibility, or the number of rewards offered and the level at which they are awarded. Three studies investigate the impact of divisibility on program participation. Study 1 demonstrates how increased divisibility can inspire loyalty by making rewards seem more attainable to consumers.
Study 2 reveals how decreased divisibility helps firms create lock-in by increasing the likelihood that future purchases go where progress already has been made. Study 3 investigates the
combined effects of the opposing forces of reward attainability and consumer lock-in, illustrating how firms utilizing alterative currencies must carefully choose the right amount of divisibility.
The increasing popularity of frequency programs has given rise to hundreds, if not thousands of different alternative currencies used to record a consumer’s ongoing purchase activity. Consumers can accumulate, budget, and spend these alternative currencies much like they would traditional paper money. Money, however, is much different in that it is almost always perfectly divisible. Economists consider money divisible when “in particular, a money holder can exchange any fraction of his money holdings” (Shi 1997). Alternatives currencies are neither universally accepted, nor can they be spent in any increment. Hence, alternative
currencies are unlike legal tender in that their use is constrained; they are not perfectly divisible.
For the purposes of this research, we define divisibility as the number of unique exchange opportunities available utilizing a particular currency. For example, all else equal, a program in which awards are available at 50- and 100-point increments is more divisible than a program where awards are available at only 100-point increments. However, a program in which awards are available at 50- and 100-point increments is not more divisible than a program where awards are available at only 50-point increments. Adding another reward at 75 points would increase divisibility, while adding a reward for 150 points would not. In short, divisibility addresses the variety of spending opportunities within a loyalty program. While research examining consumer behavior within loyalty programs has begun to emerge, we know of no work that looks at the effectiveness of loyalty programs based on their divisibility.
In this research, we propose that reward levels serve as goals that influence purchase decisions. Each purchase affects a customer’s proximity to the next closest reward level or possible exchange opportunity, thereby affecting the consumer’s motivation to make a subsequent purchase in order to acquire more of the alternative currency. Therefore, the effectiveness of a loyalty program depends in large part on the number of rewards offered and
the amount of the alternative currency required to redeem a particular reward—its divisibility.
For a program to be successful, it must have a reward structure that motivates customers to view purchases as a sequence of interrelated transactions leading to a goal.
The remainder of this paper is organized as follows. First, we review the relevant literature on goals and the psychology of motivation and persistence. We then propose our conceptualization of divisibility as it applies to alternative currencies from which we derive four testable hypotheses addressing goal attainability and consumer lock-in. In study 1, we focus on the benefits of greater goal attainability brought about by increasing divisibility. More accessible rewards inspire consumers who might otherwise ignore their progress within a loyalty program.
In study 2, we focus on the benefits of greater consumer lock-in brought about by decreasing divisibility. When confronted with competing yet equivalent loyalty programs, customers are more prone to consolidate purchases with a single seller when the programs have fewer exchange opportunities. In study 3, we pit the benefits to the customer from greater goal attainability against the benefits to the firm from greater consumer lock-in to show how programs must strike a balance between providing too much and too little divisibility. We conclude by pointing out some of the limitations and suggesting avenues for future research.
GOALS, MOTIVATION AND PERSISTENCE
At its most basic level, a goal is what an individual is trying to accomplish. In most real- life situations, people maintain several often opposing goals (e.g. enjoy dinner and lose weight), and often cope by either alternating between potentially contradictory goals or maintaining the pursuit of a single goal (Fishbach and Dhar 2005). Research in consumer behavior has focused
more and more on how goal-related mind-sets influence decision-making processes (Chandran and Morwitz 2005). A fundamental premise underlying our research is that loyalty program rewards serve as goals, or incentives, that motivate consumers to continue purchasing from a particular seller or supplier. Soman and Shi (2003) have shown that consumers derive value from making progress toward goals, and this value enhances motivation and performance.
Theories of motivation have been used to explain the persistence people exert towards achieving their goals. The concept of persistence as a component of goal-directed behavior is based on the premise that effort is frequently a function of: (1) external inducements such as money, and (2) the likelihood of success in reaching the goal. Pritchard and Curtis (1973) have shown that people’s effort often depends on the size of the incentive offered and larger
incentives inspire greater effort. Mento, Cartledge, and Locke (1980) have shown that individuals are more likely to work towards a goal when they have high rather than low expectation of reaching it. In fact, the probability of choosing a given goal is increased if the individual thinks it can be attained (Bouffard-Bouchard 1990; Latham and Locke 1991). Models in this tradition have been labeled expectancy-value models (see Klein 1991 for a review).
Within this framework, both desirability and the feasibility of completion are believed to increase as the distance from the goal decreases (Lewin 1935; Atkinson and Birch 1974). Hence, proximity to a goal is at the heart of what has been labeled the goal gradient effect: organisms exert more effort towards reaching a goal the closer they come to the goal. The earliest
demonstrations of the goal gradient effect involved timing rats as they ran towards food. Hull (1932) and his colleagues found that rats closer to food ran faster. Brown (1948) found that rats nearer to food pulled harder on special harnesses that measured the strength with which they were pulling. While the goal gradient effect was documented in several early studies involving
human participants (e.g., Losco and Epstein 1977; Fenz and Epstein 1967), more recent work by Nunes and Drèze (2006) has demonstrated the effect within the context of a loyalty program. In a field study conducted at a professional car wash, patrons were offered a free wash after
purchasing a requisite number. The time between visits for patrons decreased by 0.5 days on average with each additional wash purchased. The closer consumers got towards earning the reward, the more effort they were willing to exert in terms of hastening their next purchase.
Within the context of loyalty programs, proximity to a reward essentially depends on the program’s divisibility. Increasing the number of unique exchange opportunities inevitably means people are closer to earning a reward and inherently more likely to reach their goal. It has been hypothesized that sub-goals facilitate performance by helping people maintain effort over long periods of time. In line with this, Bandura (1997) found that weekly goals led to weight loss only when daily goals were set as well. On the other hand, too many sub-goals may impede progress.
Locke and Bryan (1967) found that on a 2-hour arithmetic task, setting 15-minute sub-goals led to poorer performance than simply setting end goals. It is important to point out that in this work sub- goals deliver rewards that when redeemed may impede progress towards a more difficult goal.
Consider McDonald’s promotional tie-in with Sony’s online music store whereby customers received a free downloadable song with every Big Mac bought. Surely consumers prefer this promotion to one that awards 10 free songs only after 10 Big Macs are purchased, as consumers are expected to prefer programs with fewer required purchases to earn a specific reward. Consequently, five free songs after five Big Macs is preferable to 10 free songs after 10 Big Macs. Therefore, by increasing the divisibility of the rewards available, we expect those consumers who are brought closer to a reward, to be more inclined to continue their participation
within a program, subsequently exerting more effort consistent with the goal gradient effect. This leads to our first hypothesis.
H1: When increasing divisibility, persistence will increase for consumers who are brought closer to a reward.
In the aforementioned McDonald’s example, consumers are assumed to prefer rewards handed out more readily. Yet, rewards dispensed too readily lose their ability to serve as goals that inspire consumers to direct future purchases towards the awarding firm. Compared to providing a song with every Big Mac, awarding two songs only after two Big Macs are purchased might actually prompt consumers to buy more burgers at McDonald’s than Burger King as they make the additional effort to return for a second meal in order to receive the reward.
Hence, decreasing divisibility can induce a consumer to consolidate purchases with a single supplier. To understand the competitive benefit of decreasing divisibility, consider the following illustration of consumer lock-in.
Imagine a scenario whereby a customer can patronize one of two stores. For simplicity, assume the two stores provide undifferentiated goods and each offers the exact same “Buy N get 1 free” loyalty program. Note that, per our definition, programs with larger Ns are less divisible.
Ceteris paribus, the consumer would be indifferent between the stores and would choose randomly between the two. Yet, according to theories of persistence and expectancy-value, the consumer would also consider relative progress within each store’s loyalty program, and choose according to the expected value of the reward she anticipates getting from each store (i.e., the higher the expected reward from a store, the more likely she is to visit that store on her next trip).
In addition, if an initial action is used to infer one’s general level of commitment to a goal (i.e., buying from a particular store), it is likely to be followed by a similar choice of actions (Soman
and Cheema 2004). Therefore, action towards a certain goal can be seen as increasing the commitment to future actions that favor the same goal (i.e., purchases from the same store).
Clearly, consumers who are equally far along in two competing loyalty programs, would be indifferent between the two (no loyalty). Yet, consumers will be more likely to direct their next purchases to the firm possessing the program in which they have made the most progress.
Consistent with the goal gradient effect and the escalation of commitment, we would then expect the likelihood of directing one’s next purchase towards a particular firm at which the buyer has progressed further to increase with the size of the gap in progress between the programs. In other words, consumers who have completed five of 10 purchases at Store 1 and only one of 10
purchases at Store 2 are much more likely to steer their next purchase towards Store 1. As that gap grows larger, so would the propensity to shop at Store 1.
By definition, the gap is greatest (and thus motivation to visit Store 1 is at its peak) when the consumer is only one purchase away from earning a reward at Store 1 and has made no purchases at Store 2. Compare two shoppers in two hypothetical loyalty programs with different reward requirements. The first shopper has accumulated 9 purchases at Store 1, and 0 purchases at Store 2, and both stores require 10 purchases in order to earn a reward. The second shopper has accumulated 10 purchases at Store 1, and 0 purchases at Store 2, in programs where both stores require 11 purchases in order to earn a reward. The gap is larger for shopper 2 than for shopper 1; consequently shopper 2 is more likely than shopper 1 to visit Store 1. In other words, the likelihood of making a purchase at Store 1 when it is at its greatest (i.e., peak motivation) increases as divisibility decreases. This leads to our second hypothesis:
H2: Peak motivation towards completion is greater in less divisible programs.
Now consider the average likelihood of choosing Store 1 when 10 purchases are required versus the average when 11 purchases are required (a less divisible program). Consistent with the goal gradient effect, the likelihood of making the 11th purchase at Store 1 is greater than the likelihood of making the 10th or any preceding purchase, and thus decreasing divisibility increases the average likelihood of shopping at Store 1. Therefore, holding the reward constant across programs, the likelihood associated with each additional purchase requirement added to N is larger than the last, which in turn elevates the average likelihood. This leads us to hypothesis 3:
H3: Decreasing a program’s divisibility increases the motivation of participants to patronize the store at which they have made the most progress.
Taken together, hypotheses 2 and 3 imply that “N+1” type reward programs, although linear in nature, behave similarly to convex reward programs (whereby the rewards earned after 2 N purchases are more than two times the rewards earned after N purchases). We detail the convex properties of N+1 reward programs mathematically in the appendix.
We proposed in hypothesis 1 that increasing divisibility helps the firm induce increased loyalty by making rewards seem more attainable. In hypotheses 2 and 3, we proposed that, rather than increasing divisibility, firms can benefit from decreasing divisibility as it encourages
consumer lock-in. Accordingly firms must manage the divisibility of the currencies in their loyalty programs such that a balance is struck between offering attainable goals and locking consumers in. The benefits that can be derived by recognizing and responding to these opposing forces form the foundation for hypothesis 4.
H4: Decreasing divisibility in order to encourage lock-in decreases perceptions of goal attainability, and thus firms that balance these opposing forces can be made better off.
Before exploring the interplay between goal attainability and consumer lock-in, we study each effect individually. In study 1, we set out to test how consumers would respond if we increased the absolute number of redemption opportunities, or increased a loyalty program’s divisibility.
STUDY 1 - THE BENEFICIAL EFFECTS OF INCREASING DIVISIBILITY
Method
Participants. Participants were 160 undergraduate business students at a major West Coast
university. Twenty-three of the surveys were incomplete or unusable, leaving 137 valid responses.
Stimuli and design. The design was a 2 (accumulated miles: 5,000 miles, 20,000 miles) x 2
(divisibility: 25,000 miles, 10,000 and 25,000 miles) between subjects design. Participants completing the study were asked to imagine that they were working abroad and had begun flying regularly on an airline on which they had never flown before. They were told that they had accumulated either 5,000 or 20,000 frequent flier miles on this airline and that a round-trip ticket could be redeemed for 25,000 miles and was either the only reward available, or that they could also obtain a first class upgrade for 10,000 miles. We manipulated divisibility by increasing the number of exchange opportunities available to the consumer. In an attempt to reduce the variability in the value different respondents would place on a free ticket as well as manage expectations regarding future participation, we told participants they had paid an average of $500 per round trip in the past and that they expected to continue flying on this airline.
After reading the basic storyline, respondents were told the airline was running a special promotion allowing them to buy a voucher for 2,500 miles (i.e., adding the miles to their
account). They were then asked whether they would purchase the voucher (willingness to buy, WTB). Those who responded that they would buy the voucher were subsequently asked to report the maximum that they would pay (willingness to pay, WTP). Next, participants were instructed to imagine that instead of the previous offer, the airline was selling a voucher for 5,000 miles.
Again, they were asked whether they would purchase the voucher, and if so, the maximum that they would pay. The order of the voucher amounts (2,500 and 5,000 miles) was counterbalanced.
The vouchers served as measures of a consumer’s propensity to steer future business towards the offering firm in order to earn the reward (i.e., persistence).
_______________________
Insert figure 1 about here _______________________
Our predictions were as follows. First, we expected the insertion of a 10,000-mile reward to increase both the willingness to buy (WTB) and the willingness to pay (WTP) for both a 2,500-mile and 5,000-mile voucher (see figure 1). However, as predicted by hypothesis 1, the effect would only be applicable to those in the low accumulation condition (5,000 miles) who were far from the 25,000-mile goal. For them, the additional reward level provides an attainable goal. An attainable goal already existed for those in the high accumulation condition (20,000 miles). Therefore an interaction between divisibility and accumulated miles would provide support for hypothesis 1. Second, we expected consumers to be more likely to buy the 5,000- mile voucher than the 2,500-mile voucher, and to be willing to pay more (i.e., a main effect of voucher size such that WTP2500 < ½ WTP5000). Indeed, the larger voucher allowed the high accumulation people not only to approach their goal, but also to reach it. Yet for low wealth
respondents (5,000 miles), the larger voucher allowed them to reach their goal only when there was a 10,000-mile reward. Again, in line with hypothesis 1, we expected an interaction between voucher size and divisibility. This interaction reflects how proximity to the next closest reward level influences the value associated with successive increments of the currency, or progress towards a reward.
In the second part of this study, a subset of the respondents was asked an additional question. They were told that they were on the phone with a representative of the airline planning a trip across country. They had two options: either a direct flight, which would earn them 4,500 miles, or a connecting flight with a stopover that would add two hours to the trip, but would earn them 5,500 miles. Both tickets were said to cost the same amount, and respondents were asked to specify which flight they would take. The purpose of this question (willingness to connect, WTC) was to replicate our results using time as a medium of exchange with miles rather than money. The demonstration of our effects across multiple currencies further generalizes the notion of consumer effort, operationalized as both time and money, towards a goal.
Results
The order in which the voucher was presented had no discernible effect and thus the data were collapsed. The results are summarized in table 1. The data related to willingness to pay were analyzed using ANOVA while the data related to willingness to buy were analyzed using logistic regression.
_______________________
Insert table 1 about here _______________________
In terms of willingness to pay, our interest was in how people valued miles conditional on their intent to buy the voucher. In order to compare marginal value across voucher size we
compared willingness to pay for the 2,500-mile voucher with one-half the amount they were willing to pay for the 5,000-mile voucher. For WTP there was a main effect for accumulated miles (WTP20K miles = 20.9 versus WTP5K miles = 13.7, F(1, 231) = 28.7, p < .01), but not for voucher size. We find support for hypothesis 1 in that the main effect of divisibility (WTP1 level = 15.4 versus WTP2 levels = 19.2, F(1, 231) = 7.84, p < .01) was qualified by a significant interaction with accumulated miles (F(1, 231) = 17.9, p < .01). As expected, the effect of divisibility on willingness to pay only applied in the low accumulation condition. Adding a reward at 10,000 miles mattered for those for whom a 25,000-mile reward might otherwise seem unattainable.
In terms of willingness to buy, respondents were more likely to buy the voucher the closer they were to their goal, resulting in a main effect of accumulated miles (WTB20K miles = 95.3% versus WTB5K miles = 77.6%, χ2 = 15.83, p < .01). In addition, there was a main effect of voucher size such that people were more likely to buy the voucher the closer it brought them to their goal (WTB5K voucher = 90.2% versus WTB2.5K voucher = 81.8%, χ2 = 4.12, p < .05). More importantly, there was a main effect for divisibility such that the insertion of a 10,000-mile reward increased the likelihood of purchase (WTB1 level = 81.7% versus WTB2 levels = 90.6%, χ2 = 5.96, p < .05). As was the case with willingness to pay, we tested for an interaction between divisibility and accumulated miles. While the pattern of results was consistent with the results for WTP, the interaction only approached statistical significance (χ2 = 2.47, p < .11).
With regard to the second part of this study, we find an identical pattern of results (see table 2). There were significant main effects for accumulated miles (WTC20K miles = 87.5% versus WTC5K miles = 46.3%, χ2 = 51.31, p < .01) and divisibility (WTC1 level = 61.3% versus WTC2 levels =
72.5%, χ2 = 16.89, p < .01). These main effects were qualified by a significant interaction (χ2
=13.39, p < .01), which provides additional support for hypothesis 1. People were more inclined to take a connecting flight with a 2-hour stopover in order to earn an additional 1,000 miles when they already possessed 20,000 miles, or when they possessed 5,000 miles and there was a reward offered at 10,000 miles.
_______________________
Insert table 2 about here _______________________
Discussion
In study 1, we expected the insertion of a second reward (increased divisibility) to increase the likelihood of purchase, the willingness to pay, and the likelihood of making an inconvenient stopover, but to do so only in the low accumulation condition; when the additional reward introduced a goal that appeared attainable. Changes in these dependent measures reflect the increase in consumer persistence caused by increasing divisibility as predicted by hypothesis 1. We find a significant interaction between divisibility and wealth for both willingness to pay and likelihood of making a stopover, and a weak interaction for willingness to buy. By
increasing divisibility we have provided an attainable reward level that can serve as a realistic goal to those consumers who otherwise might consider their 5,000-mile holdings as valueless.
Further, those in the high accumulation condition were not affected by the addition of a less valuable reward.
In study 1 we find that doubling the exchange opportunities from one to two encourages consumers’ participation within the program. Yet, we do so in the context of a single provider
without testing the limits of the beneficial effects of increasing divisibility. This issue is
addressed explicitly in study 3. First, however, we explore the potential benefits associated with reducing divisibility independently in study 2.
STUDY 2 - THE BENEFICIAL EFFECTS OF DECREASING DIVISIBILITY
Study 2 is comprised of two parts. In part A, which examines the peak probability across loyalty programs with different purchase requirements, we test hypothesis 2. In part B, which examines the average probability across programs with different degrees of divisibility, we test hypothesis 3.
Method
Participants. Participants in part A of this study were 142 undergraduate business students at a
major West Coast university. In part B, participants were 120 MBA students at the same university.
Stimuli and design. In part A, we were interested in assessing the peak probability associated
with loyalty programs that differ in their divisibility. Recall from the discussion leading to hypothesis 2 that the greatest likelihood of making a purchase at Store 1 is reached when 0 purchases are made at Store 2 and N-1 purchases, or one purchase short of the reward, have been made at Store 1. Therefore, we manipulated whether the consumer had accumulated 5 of 6
(83%), 7 of 8 (88%), 9 of 10 (90%) or 19 of 20 (95%) of the purchases required to earn a reward.
The paper and pencil based study utilized the following scenario:
Imagine you are at a local mall and want to stop for lunch. You are in the mood for a sandwich and there are two shops in the food court that could satisfy you craving. At the first, Store 1, you are a member of their frequent buyers’ club. You need to purchase a total of 6 [8, 10, 20] subs total to get a free sub. You already have 5 [7, 9, 19] stamps on your frequent diner card. There is a line at Store 1, which will take at least 25 minutes.
At Store 2, which makes equally good sandwiches, there is also a loyalty program.
You get a free sub after purchasing 6 [8, 10, 20], just like at Store 1. Although you possess a frequent diner card for Store 2, you have 0 purchases recorded on it. The line at Store 2 is moving much faster and will take about 5 minutes.
At which store would you buy your lunch?
Store 1 _____ Store 2 _____
This scenario forces respondents to make a tradeoff between waiting 20 additional minutes to accumulate the final stamp necessary to receive the reward at store 1 and spending 20 minutes less in line, but not earning the reward. We can assess the impact of program divisibility on store choice by measuring the percentage of people willing to wait in each condition.
In part B, we were interested in measuring the average likelihood across different points of progress. Recall from our discussion leading to hypothesis 3 that the likelihood associated with each additional purchase requirement added to N is larger than the last, and therefore elevates the average likelihood. If this is true, ceteris paribus, the average likelihood of making another purchase at a store requiring four purchases in order to earn a reward should be greater than at a store requiring only three purchases. With this in mind, we created the following scenario:
Imagine you need to fly to Northern California for the weekend. Two local budget carriers – Airline 1 and Airline 2 – have nearly identical itineraries. The tickets are
priced the same, the service and all other aspects of the trip are nearly identical, and you are only deliberating on which carrier to fly.
Each airline has a loyalty program that offers a free flight after 3 [4] paid tickets. On Airline 1, you have flown 1 [2; 1, 2, 3] flight[s] recorded towards earning a free ticket. On Airline 2, you currently have 0 flights recorded towards a free ticket. The airport at which Airline 1 arrives is 60 minutes from your destination. The airport at which Airline 2 arrives is 20 minutes from your destination.
Which airline would you fly?
(Please put an “X” next to your choice.) Airline 1 ____ Airline 2 _____
In this scenario, we force respondents to make a tradeoff between an additional flight on the carrier with which they have progress and an additional 40 minutes of drive time. There were five conditions based on the program requirement [3 or 4 flights] and the number of qualifying flights completed [1 or 2 in the 3 flights condition and 1, 2, or 3 in the 4 flights condition]. This is not unlike Southwest Airline’s frequent flier program, which issues free flights based on the number of trips taken rather than the number of miles flown. A priori, we expected that when the program requirement was less divisible, or required 4 flights [averaging across people with 1, 2 and 3 flights completed], respondents would be more likely to endure the longer drive and choose Airline 1, than when the program requirement was only three flights [averaging across people with 1 and 2 flights completed]. All of the studies were run between subjects.
Results
The data were analyzed utilizing logistic regression. In order to test hypothesis 2, we regressed the likelihood of choosing Store 1 on the number of purchases needed to earn a reward.
As expected, the probability of choosing Store 1 in part A depended on the number of purchases
required (slope = 0.08, χ2 = 6.0, p = 0.01). This result provides support for hypothesis 2 in that the peak motivation—defined as the likelihood of making a purchase at Store 1 when only one purchase is required to get the reward—increased as divisibility decreased (see figure 2).
_______________________
Insert figure 2 about here _______________________
In part B, we find that the probability of steering business towards Airline 1, where driving 40 extra minutes is required, is higher when the divisibility is decreased. In support of hypothesis 3, the average likelihood of choosing Airline 1 was higher when consumers earned a reward after accumulating four flights (μN=4 = 59.5% versus μN=3 = 42.6%, χ2 = 7.2, p < .01).
Also, reinforcing the results supporting hypothesis 2, the likelihood of flying Airline 1 was highest for those who completed three of four flights, as it was for those who completed two of three, when three flights were required (see figure 3). In addition, those who were 3/4ths complete were more likely to choose Airline 1 than those who were 2/3rds complete (i.e., greater peak probability) respectively (PA3/4 = 79.6% versus PA2/3 = 58.3%, χ2 = 5.0, p < 0.05).
_______________________
Insert figure 3 about here _______________________
Discussion
This study suggests that reduced divisibility can be beneficial to the firm in two
straightforward ways. First, in accordance with hypothesis 2, peak motivation, or the likelihood of making a purchase at a particular store when it is greatest, increases as divisibility decreases.
We see this pattern in the results of part A of study 2. Second, in accordance with hypothesis 3, the average motivation or likelihood of visiting a store where progress has been made is greater for customers participating in a less divisible program. We see this pattern of results in part B of study 2. Taken together, this shows how less divisible programs can benefit the firm. We should point out that this is not necessarily the case when competitive programs offer differing rewards structures. Instead, these results suggest that for firms with equivalent reward structures (e.g., for all intents and purposes United and American Airlines), less divisible programs will inspire greater loyalty. Of course, if this result were pushed to its limit, with one million miles being required for one free flight (i.e., making attainability seem unlikely), the result would probably no longer hold. Consumers must believe that goal attainment is possible.
In study 1, we demonstrated how increased divisibility often helps the firm induce increased loyalty by making rewards seem more attainable. The results of study 2 demonstrate how reducing divisibility helps firms increase consumer loyalty by raising the likelihood that future purchases go where progress already has been made. In study 3, we investigate the
combined effects of these opposing forces. Just as in study 1, we increase the number of rewards available. Just as in study 2, we vary consumer progress within competing programs. We explore choice among retailers as a function of the purchases made and the number of purchases required simultaneously; we look at the choice probabilities associated with choosing between retailers based on divisibility.
STUDY 3 – BALANCING THE OPPOSING FORCES OF DIVISIBILITY
Method
Participants. Participants were 300 undergraduate business students enrolled at a major West
Coast university. They completed the survey voluntarily. Seven respondents failed to answer the questions resulting in 293 usable surveys.
Stimuli and design. We utilized a 3 (divisibility: reward every $100, $500, or $1,000 spent) x 4
(accumulated spending: $25, $425, $525, or $925) between subjects, full factorial design.
Participants were asked to imagine that they had moved to a new city and there were two large grocery stores near their house carrying essentially the same assortment of goods, and that the stores were no different in terms of price, layout, cleanliness, or service. However, they were told the two grocers did differ significantly in their frequent shopper programs. Store B, “offers a 10% cash refund on every dollar” spent at the store, which is issued immediately. For example, if a shopper spends $50, they would receive a $5 refund immediately (i.e., thereby only paying
$45). Store A offers cash back at various increments. These increments or reward levels varied according to the degree of divisibility ($100, $500, or $1,000). It is important to note that the cumulative rewards offered by the two stores are identical; only the frequencies at which the rewards are dispersed differ.
The first question in the study asked them which store they believed they would shop at more frequently. The study went on to describe their progress towards the goal at Store A. It explained that: “one store is near the freeway, which you often take to work, while the other is closer to a main surface street, which you take to and from work instead of the freeway when you leave during rush hour. While they are equal distances from your house, they are in opposite directions. As a result, you have shopped at both stores repeatedly during the past six months, as
convenience has often been a factor.” This scenario was consistent with the metropolitan community in which most students resided. We then manipulated accumulated spending by stating: “at store A you have currently accumulated $25 [$425, $525, or $925] in purchases.
(Remember, when you reach an accumulated spending of $100 [$500, $1,000] they will refund
$10 [$50, $100].)” The scenario then told them that it was the weekend (rush hour would not be a factor) and asked them to imagine that they were heading out to buy $50 in groceries for a party they were planning that evening. They were asked to indicate at which store they would shop.
There are three fundamental predictions we make based on our theorizing. First, in accordance with the goal gradient effect, when the only reward available is at $1,000, consumers will become increasingly inclined to go to Store A as their accumulated spending increases from
$25 up to $925. In other words, the value of each dollar in accumulated spending increases as they near the $1,000 goal. Second, in accordance with hypothesis 1, by inserting a sub-goal at
$500, we expect the propensity to visit Store A (i.e., continue progress towards the goal) to increase for those who were far from the $1,000 goal. In particular, we expect those with an accumulated spending level of $425 to be more likely to visit store A when rewards are issued for every $500 spent than when they are issued for every $1,000 spent. In addition, changing the payoff scheme as such should also make shopping at Store A more attractive at the $425
accumulated spending level (in anticipation of the goal) than at the $525 accumulated spending level (after achieving one goal and far from the second). Third, in accordance with hypothesis 4, we expect a small increase in divisibility (i.e., going from rewards issued for every $1,000 to every $500) to increase the overall attractiveness of the program. Conversely, we expect the large increase in divisibility (from rewards issued for every $1,000 to every $100) to negatively
impact the overall attractiveness of the program, as the rewards are no longer viewed as goals worth striving for and there is no lock-in.
Results
Figure 4 depicts the reported probability of going to Store A for each of the three reward levels ($1,000, $500 and $100) and each of the four levels of accumulated spending ($25, $425,
$525, $925). As is evident from the graph, the likelihood of going to Store A is affected by both the level of reward and accumulated spending. As predicted, the likelihood of going to Store A increases monotonically in the $1,000 reward condition. Tests of proportion revealed that those who spent $525 were more likely to choose Store A than those who spent only $25 (PA525 = 28%
versus PA25=4%, χ2 = 3.9, p < .05), and those who spent $925 were more likely to go to Store A than those who spent $525 (PA925 = 100% versus PA525 = 28%, χ2 = 10.2, p < .01). Although the likelihood of choosing Store A in the $425 condition (16%) falls halfway between those who spent $25 (4%) and those who spent $525 (28%), neither difference is statistically significant.
Nevertheless, the overall pattern of results supports our first prediction.
_______________________
Insert figure 4 about here _______________________
With regards to our second prediction, by increasing the divisibility, we increased the likelihood of going to Store A for those with $425 in accumulated spending from 16% when the reward level was at $1,000 to 76% when it was reduced to $500 (χ2 = 15.3, p < .01). The
likelihood of going to Store A was also larger for those who spent $425 and had not yet achieved the $500 goal than for those who spent $525 and had earned the reward at $500, but were now
$475 away from earning another $50 (PA425 = 76% versus PA525 = 44%, χ2 = 5.1, p < .05).
These results replicate those of study 1, adding further support for hypothesis 1.
Most importantly, our results support hypothesis 4. The average likelihood of going to Store A was largest (57%) when divisibility was moderate, and shoppers were rewarded for every $500 they spent. The average likelihood was smallest (11%) when the divisibility was increased to reward shoppers for every $100 they spent. The loyalty program that rewarded shoppers for each $1,000 they spent fell in between at 36%. In other words, the $500 reward level induced greater loyalty than both the $1,000 reward level (χ2 = 8.6, p < .01) and the $100 reward level (χ2 = 15.2, p < .01).
Discussion
Study 3 clearly shows the importance of divisibility, or the number of rewards and the level at which they are awarded, on the ability of a loyalty program to incite consumers to shop at a particular store. While the overall amount a shopper would receive remains constant for anyone spending $1,000, two effects are at play here. If one makes the program too divisible (but not perfectly, as in Store B) then the rewards structure is unlikely to be attractive enough to induce loyalty. In contrast, if the program is not divisible enough, it may offer great incentives, but only for those who are already committed to the program, and thus will be de-motivating for relative newcomers. An adequate balance must be struck between the two forces.
Another point made salient by study 3 is that loyalty programs benefit from a strong memory effect – people take into consideration whether they have successfully reached the goal in the past. Indeed, one would expect that in the $500 reward condition the percentage of
respondents who would travel to Store A would be equal in the $25 and $525 accumulated spending conditions (each $475 away from earning $50). Likewise, we would expect the percentages to be equal in the $425 and $925 accumulated spending conditions (each $75 away from earning $50). This is not what occurred. Instead, those who accumulated $525 in spending were more likely to choose Store A than those who accumulated $25 in spending (44% versus 12.5% respectively, χ2 =5.35, p < .05). Similarly, those who accumulated $925 in spending were marginally more likely to choose Store A than those who accumulated $425 in spending (96%
versus 76% respectively, χ2 = 3.11, p < .10).
This suggests that consumers benefit from reaching worthwhile goals in more than a monetary way. It may be that affect regulation, or a process in which the positive affect associated with the reward represents a goal in itself, plays a role and individuals attempt to protect it once the state has been attained (Andrade 2005). But whether the benefits come from positive affect, a sense of accomplishment, an updated likelihood of success, or another reason is not clear from our results, but appears worthy of further investigation. It is important to
remember that this effect did not occur in the $100 reward condition indicating that consumers do not seem to derive any secondary benefits from achieving easy goals. One should also note that at the $525 level, the respondents in the $500 condition were actually more likely to go to Store 1 then the respondents in the $1,000 condition, even though their accumulated spending is lower ($25 vs. $525).
CONCLUSION, LIMITATIONS AND FUTURE RESEARCH
The objective of this research was to demonstrate how the effectiveness of a loyalty program in spurring consumers to consolidate purchases depends directly on the number of rewards offered and the amount of an alternative currency required to redeem each reward, what we have labeled a program’s divisibility. We adopt a goal-based framework to describe how consumers respond to increasing and decreasing divisibility. We argue that salient rewards serve as goals, and each goal’s attainability (proximity) as measured by the reward level and the amount of the currency thus far accumulated helps determine whether a consumer is more or less inclined to continue working towards the goal (i.e., steering her purchases such that she
continues to accrue the alternative currency that takes her closer to her goal). Valuable and attainable rewards serve to increase the marginal value of the currency used to achieve that reward. While we demonstrate that increasing divisibility can allow for increased loyalty among those with low asset levels in the alternative currency, this research also reveals how too much divisibility can be de-motivating as it diminishes the effectiveness of rewards as goals. This last result is counter-intuitive, as one would expect a currency to be more highly valued as the exchange options available for that currency increase. This is the first research we know of to test the effects of increasing and decreasing the number of rewards available and the level of effort required on effort exerted.
This research is not without its limitations. While in our studies, divisibility was increased when the amount of an alternative currency required to redeem each reward was lowered, increasing divisibility alone does not necessarily imply easier attainability (i.e., a lower N implies easier attainability). Consider study 1, in which we doubled the exchange opportunities by offering a 10,000-mile upgrade as well as a 25,000-mile free ticket. We could just as easily have offered a second reward at 30,000 miles. While this would increase divisibility, it may not
make the program any more appealing. While we show that more rewards can be better than fewer rewards, and too many rewards can be a bad thing, the precise numbers and levels associated with two few and too many is likely to vary by product or service category, rewards offered, medium of exchange (miles, points, purchases) and a slew of other factors too numerous to integrate into our studies.
From a practical perspective, it is worthwhile to note that the divisibility of any
alternative currency may not be entirely within the control of the firm issuing that currency. For example, frequent flier miles, the most ubiquitous alternative currency, are becoming
interchangeable with several other alternative currencies and can now be redeemed at numerous second-party vendors. Some firms, such as Southwest Airlines, limit the divisibility of their alternative currency by issuing credits in non-divisible increments (flight segments).
Finally, we believe the notion of divisibility applies to marketers in many domains and is worthy of study outside of the realm of loyalty programs. For example, when people receive gift certificates denominated in specific amounts (e.g., $5) that are good only at the issuing vendor, they may encounter divisibility issues. Indeed, if they can not find an item they desire costing exactly $5, they may be forced to decide whether to make a larger purchase and spend their own money, or a smaller purchase and risk wasting a portion of the certificate. One only needs to recall the time spent in a foreign airport trying to spend the remainder of the local currency one held before departing a country they were unlikely to visit again to get a sense of the power of limited divisibility.
REFERENCES
Andrade, Eduardo B. (2005), “Behavioral Consequences of Affect: Combining Evaluative and Regulatory Mechanisms,” Journal of Consumer Research, December, forthcoming.
Atkinson, John W. and David Birch (1974), “The Dynamics of Achievement-Oriented Activity,”
in Motivation and Achievement, Eds. J.W. Atkinson and J.O. Raynor, Washington D.C.:
Winston, 271-325.
Bandura, Albert (1997), Self-Efficacy: The Exercise of Control, New York, NY: Freeman.
Bouffard-Bouchard, Thérèse (1990), “Influence of Self-efficacy on Performance in a Cognitive Task,” The Journal of Social Psychology, 130 (3), 353-63.
Brown, Judson S. (1948), “Gradients of Approach and Avoidance Responses and their Relation to Level of Motivation,” Journal of Comparative and Physiological Psychology, 41 (6), 450-65.
Chandran, Sucharita and Vicki G Morwitz (2005), “Effects of Participative Pricing on Consumers' Cognitions and Actions: A Goal Theoretic Perspective,” Journal of Consumer Research, 32 (2), 249-60.
Fenz, Walter D. and Seymour Epstein (1967), “Gradients of Physiological Arousal of Experienced and Novice Parachutists as a Function of an Approaching Jump,”
Psychosomatic Medicine, 29 (1), 33-51.
Fishbach, Ayelet and Ravi Dhar (2005), “Goals as Excuses or Guides: The Liberating Effect of Perceived Goal Progress on Choice,” Journal of Consumer Research, forthcoming.
Hull, Clark L. (1932), “The Goal Gradient Hypothesis and Maze Learning,” Psychological Review, 39 (1), 25-43.
Klein, Howard J. (1991), “Further Evidence on the Relationship between Goal-Setting and Expectancy Theory,” Organizational Behavior and Human Decision Processes, 49 (2), 230-57.
Latham, Gary P. and Edwin A. Locke (1991), “Self-regulation through Goal Setting,”
Organizational Behavior and Human Decision Processes, 50, 212-47.
Lewin, Kurt (1935), A Dynamic Theory of Personality, New York, NY: McGraw-Hill.
Locke, Edwin A., Bryan, Judith F. and Kendall, Lorne M. (1968), “Goals and Intentions as Mediators of the Effects of Monetary Incentives on Behavior,” Journal of Applied Psychology, 52, 104-121.
Losco, Jean and Seymour Epstein (1977), “Relative Steepness of Approach and Avoidance Gradients as a Function of Magnitude and Valence of Incentive,” Journal of Abnormal Psychology, 86 (4), 360-8.
Mento, Anthony J., Norman D. Cartledge, and Edwin A. Locke (1980), “Maryland vs. Michigan vs. Minnesota: Another Look at the Relationship of Expectancy and Goal Difficulty to Task Performance: 1966-1984,” Organizational Behavior and Human Decision Processes, 25 (3), 419-40.
Nunes, Joseph C. and Xavier Drèze (2006), “The Endowed Progress Effect: How Artificial Advancement Increases Effort,” Journal of Consumer Research, March, forthcoming.
Pritchard, Robert D. and Michael I. Curtis (1973), “The Influence of Goal Setting and Financial Incentives on Task Performance,” Organizational Behavior and Human Decision Processes, 10 (2), 175-83.
Shi, Shouyong (1997) “A Divisible Search Model of Fiat Money,” Econometrica, 65, 75-102.
Soman, Dilip and Amar Cheema (2004), "When Goals are Counter-Productive: The Effects of Violation of a Behavioral Goal on Subsequent Performance," Journal of Consumer Research, 31 (June), 52-62.
Soman, Dilip and Mengze Shi (2003), “Virtual Progress: The Effect of Path Characteristics on Perceptions of Progress and Choice Behavior,” Management Science, 49 (September), 1229-50.
TABLE 1
Study 1: Percentage Buying Vouchers and Willingness to Pay
Wealth Goal(s)
% Willing to Buy
2,500 voucher WTP
% Willing to Buy
5,000 voucher WTP/2
5,000 25k 65.0% $9.27 75.7% $8.89
5,000 25k & 10k 79.4% $17.07 93.8% $19.73
20,000 25k 93.8% $19.87 97.0% $23.75
20,000 25k & 10k 93.5% $20.15 96.8% $19.74
TABLE 2
Study 1: Percentage willing to make a connection
Wealth Goal(s)
% willing to make a connection
5,000 25k 32.5%
5,000 25k & 10k 60.0%
20,000 25k 90.0%
20,000 25k & 10k 85.0%
Figure 1
Study 1: Impact of Divisibility on the Desirability of Accumulated Miles
25K 10K
5K 7.5K 20K 22.5K
Low Divisibility High Divisibility Desirability
Accumulated Miles
Figure 2
Study 2: Likelihood of Visiting Store 1
Number of Purchases Required before Earning a Reward
0%
10%
20%
30%
40%
0 5 10 15 20 25
Figure 3
Study 2: Likelihood of Flying on Airline 1
Percentage of Trips Required Taken Towards Earning the Reward
0%
25%
50%
75%
100%
0% 25% 50% 75% 100%
3 Flights 4 Flights
Figure 4
Study 3: Impact of divisibility on Store Choice
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
$0 $100 $200 $300 $400 $500 $600 $700 $800 $900 $1,000
Accumulated Wealth
P(Visiting Store A)
$100 $500 $1,000 Fungibility