You are on page 1of 14

Omega 41 (2013) 665–678

Contents lists available at SciVerse ScienceDirect

Omega
journal homepage: www.elsevier.com/locate/omega

A free gift card alternative to price discounts in the newsvendor problem


Moutaz Khouja n, Sungjune Park, Jing Zhou
Belk College of Business, University of North Carolina at Charlotte, Charlotte, NC 28223, USA

a r t i c l e i n f o abstract

Article history: In this paper, we develop a newsvendor model in which the retailer gives ‘‘free’’ gift cards to consumers
Received 23 November 2011 who purchase a regularly priced product at the end of the selling season instead of discounting the
Accepted 28 August 2012 product. The model is developed for a market with patient consumers. We derive the sufficient
Processed by B. Lev
optimality condition for the retailer’s stocking level in the first period and the optimal gift card value in
Available online 28 September 2012
the second period. We also investigate the conditions under which giving gift cards results in higher
Keywords: expected profits than discounting the product. We find that five factors determine the effectiveness of
Inventory control gift cards. The first three factors are consumers’ valuation per $1 of gift card, gift card redemption rates,
Game theory and the average gross margin of the retailer. The last two factors are the degree to which consumers use
Marketing
gift cards to pay for products which they would have purchased from the retailer in the future with
cash, and the additional spending above the gift card value consumers make when they redeem the
card. The last two factors have a strong interaction. We also find that gift cards can be profitable when
patient consumers consistently value each $1 by their redemption probability, even with 100%
redemption. Numerical analysis shows that in the presence of patient consumers, increases in the
redemption rate may lead to an increase in the expected profit. Similar counter-intuitive behavior of
the expected profit occurs with changes in other problem parameters. The analysis also shows that gift
cards’ profit advantage over discounting increases with the variability of demand. The analysis also
indicates that gift cards are most effective for low to medium priced products sold by high margin
retailers.
& 2012 Elsevier Ltd. All rights reserved.

1. Introduction Retailers can use promotions such as ‘‘free’’ gift cards with the
purchase of a product instead of discounting the product at the
Recently, Cachon and Swinney [8] contributed to our under- end of the period. Fig. 1 shows three examples of gift cards given
standing of the effect of strategic consumers on inventory and free with the purchase of products by large retailers. Gift cards
pricing policy of retailers and on their profits. In their model, have several advantages for the retailer. When used as a promo-
strategic or rational consumers decide between purchasing the tional tool, they are not directly linked to a product and, therefore,
product at the regular price or waiting to purchase later, if the do not create a perception of reduced value of the product like
product is available, at a discounted but uncertain price depend- direct discounts and coupons [11]. The second advantage of gift
ing on which action maximizes their expected gain. The effect of cards is that they can be given by the recipient to another
strategic consumers on retailer pricing and profitability has been consumer, and thus have the potential to expand a retailer’s
the focus of much recent research. Levin et al. [22] distinguished consumer base. A third advantage of gift cards is that some
strategic consumers from others by their awareness that pricing is consumers may redeem only part of the cards’ full value and
dynamic and by timing their purchases accordingly. Strategic others may not use the card at all. Reasons for partial or no
consumers are patient and can weigh the benefit of delaying redemption include loss of cards, expiration of cards, and con-
purchases [5]. Cachon and Swinney [8] concluded from their sumers’ relocation. We refer to this partial redemption as slip-
model that in the presence of strategic consumers it is optimal for page. Horne [17] estimates the slippage rate of the cards that go
the retailer to order a smaller quantity and she will have a lower unredeemed nine month after issuance is 19%. This 19% slippage
expected profit than if there are no strategic consumers. rate is for gift cards which consumers buy and use as gifts. A
fourth advantage of gift cards is that consumers redeeming gift
cards may spend more than the value of the gift card, creating
additional sales. In a survey of more than 500 U.S. consumers
n
Corresponding author. Tel.: þ1 704 687 3242; fax: þ1 704 687 6330.
conducted by Accenture [1], 45% of respondents said they spend
E-mail addresses: mjkhouja@uncc.edu (M. Khouja), supark@uncc.edu (S. Park), more than the value of the gift card when they redeem it. One
jzhou7@uncc.edu (J. Zhou). estimate of the average spending of consumers beyond the card

0305-0483/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.omega.2012.08.005
666 M. Khouja et al. / Omega 41 (2013) 665–678

Fig. 1. Examples of product-specific gift cards.

value when redeeming cards is about 20% of card value [30]. optimal order quantity and gift card value? How do the results
Finally, when gift cards are redeemed, the retailer incurs only the change under increasing redemption rate in gift card value and
cost of the goods sold with the card rather than the full face value decreasing consumer valuation for gift card dollars as gift card
of the card. The last two benefits are realized by the retailer when value increases?
consumers use gift cards to make purchases they would not have We find that the most important factor in determining gift
made at the retailer in the future with cash. On the other hand, if cards profitability is the proportion of redeemed card amounts
purchases made with gift cards result in a reduction in future used to buy products which consumers would not normally buy
purchases that would have been made at the retailer with cash, from the retailer with cash. When this proportion is large, gift
then the last two benefits may altogether vanish. cards’ profit advantage over discounting for retailers with high
We focus on the use of ‘‘free’’ gift cards to sell excess news- average margin can be substantial. This profit advantage of gift
vendor type products in the presence of patient consumers. In our cards over discounting also increases with the variability of
model, patient consumers are similar to strategic consumers in demand. When demand variability is high, the retailer can use
the sense that they act strategically in evaluating their ‘‘buy now’’ the gift cards to mitigate the risk of overstocking and, hence, she
vs. ‘‘buy later’’ decision. These patient consumers may not behave orders a larger quantity under a gift card strategy relative to the
strategically in the redemption period because of the temporal quantity she orders under a discounting strategy.
separation between the purchase decision and the redemption The remainder of this paper is organized as follows. In Section 2,
time. We use the model framework developed by Cachon and we review the relevant literature. In Section 3, we describe our
Swinney [8] to analyze the effectiveness of gift cards. Using their model, consumer choice, and demand. In Section 4, we solve the
framework, we derive the sufficient optimality condition for the problem with a gift card strategy to sell excess inventory. We
retailer’s stocking level and the optimal gift card value in the extend the gift card model in Section 5 to allow for a step-wise
presence of patient consumers. We identify the conditions under increasing gift card redemption rate and a piece-wise linear
which offering gift cards outperforms discounting the product. In concave consumer utility in gift card value. Discussion and numer-
particular, we analyze how consumers’ valuation of gift card ical analysis are presented in Section 6. The paper is concluded in
dollars, gift card redemption rate, additional spending above the Section 7.
cards’ value at redemption, the type of purchases made using
redeemed gift cards, and the retailer’s profit margin influence the
effectiveness of gift cards. 2. Literature review
Using our model we address several questions about the use of
‘‘free’’ gift cards. What are the most important factors in deter- Since patient consumers in our model bear some resemblance
mining the profitability of using ‘‘free’’ gift cards to sell excess to strategic consumers, we begin with a review of the literature
inventory? Under what conditions do gift cards provide a more on the latter. There is a growing literature in Operations Manage-
profitable strategy for selling excess inventory than discounting? ment and Marketing on developing models that incorporate the
What is the effect of the presence of patient consumers on gift behavior of strategic consumers [18,22,28,32]. A review of recent
card profitability? How do consumers’ valuation for gift card work in Operations Management that incorporates strategic
dollars, their gift card redemption rate, spending above the gift consumer behavior can be found in Netessine and Tang [24].
cards’ value at redemption, and other factors affect the retailer’s The key feature of strategic consumers is that they can decide to
M. Khouja et al. / Omega 41 (2013) 665–678 667

delay their purchasing decision if they expect that delaying and period, the product sells for p per unit. In the second period, any
purchasing later will give them higher surplus [21]. These con- remaining inventory is discounted and sold. The cost of the
sumers are forward looking and are willing to wait to make their product is c per unit. There are three types of consumers in the
purchase if they anticipate a discount or a promotion. Unlike market: bargain-hunting, myopic, and strategic (patient in our
myopic consumers, a strategic consumer may not buy the product model) consumers [8]. Bargain-hunting consumers only buy
in the first or second period. The presence of strategic consumers discounted products. These consumers do not visit the retailer
generally leads the retailer to decrease her stocking level and in the first period because they know that the price is too high
make less profit. Quick response capability, i.e., the retailer’s relative to their valuation, and they do not buy products even in
ability to order additional inventory after obtaining updated the second period unless they believe they are getting a ‘‘deal’’.
demand information, was shown to be of significant value to The size of the bargain-hunting consumer segment is very large
the retailer and to mitigate the effects of strategic consumers [8]. and its consumers have a product valuation of vB o c, which
Strategic consumers’ valuation for the product changes over makes them the usual segment for salvaging excess inventory in
time. This assumption has been adopted for newsvendor-type the newsvendor model [8]. Methods for estimating the correct
products in previous dynamic pricing models [33,15]. Strategic salvage value can be found in Cachon and Kök [7]. There are D 4 0
consumers’ valuation for the product may increase or decrease patient and myopic consumers with product valuation of vM 4p
overtime. Bitran and Mondschein [6] suggest that the valuation of in the first period. D is a random variable with distribution FðÞ,
travelers for airline tickets is an example of valuations which may complementary cdf F ðÞ ¼ 1FðÞ, and density f ðÞ. Furthermore,
shift to the right (increase) over time as the departure date similar to Cachon and Sweeny [8], we assume D satisfies the
approaches. They also suggest that consumers’ valuations for monotone scaled likelihood ratio (MSLR) property, i.e., for all g r 1
fashion goods shift to the left (decrease) overtime since early and d in the support of D, f ðgdÞ=f ðdÞ is monotonic in d. Many
purchase allows consumers to enjoy these items earlier and commonly used non-negative distributions, including the gamma,
longer. The focus of most models with strategic consumers, Weibull, uniform, exponential, power, beta, chi, and chi-squared,
including ours, is on valuations which decrease over time. satisfy the MSLR property.
Retailers can also behave strategically. Gaur and Park [13] Of the D consumers, ð1aÞD are myopic consumers and all of
describe a retailer as strategic if she recognizes that its demand them purchase in the first period because they are not willing to
distribution depends on its inventory level through consumer wait. These consumers do not visit the retailer in the second
learning. In that sense, the retailer in Cachon and Swinney’s [8] period. The remaining aD are patient consumers. A summary of
model behaves strategically. In our analysis, a retailer is strategic the size and characteristics of the three consumer segments is
not only if she recognizes that its demand distribution depends on shown in Table 1.
its inventory level because of the presence of patient consumers Each patient consumer determines if she should buy the
but also if she uses promotional tools to mitigate the effects of product in the first period or wait until the second period.
patient consumer behavior. A patient consumer’s decision to buy the product in the first period
Retailers have used various promotional tools to increase their or to wait is based on her second period’s valuation, her forecast of
sales and profits for a long time. These include discounts and the retailer’s second period’s product price, and her chance of
coupons [3]. Another strategy which has gained popularity in having a unit available. The actual purchase decision in the second
recent years has been the use of gift cards. Gift card sales grew to period depends on the realization of the retailer’s price. Let v be the
$91 billion in 2010 (Mui 2010), a $3 billion increase over their patient consumers’ reservation prices in period 2, a random
previous peak of $88 billion reached in 2007 [16]. variable with a uniform distribution function GðÞ and support on
While the areas of price discounts and coupons have been ½v,v. A patient consumer buys a unit of the product in the first
extensively investigated [2,3,10], little work has been done in the period if her surplus from buying in the first period is larger than
area of gift cards. Khouja et al. [19] analyzed the effectiveness of her anticipated surplus from buying a unit in the second period
gift cards which are given free to consumers whose total purchase weighted by the probability of a unit being available. Otherwise,
amount in a single store visit reaches or exceeds a specified she waits until the second period in anticipation of the discount.
threshold. The criteria for getting the gift card in their model is Following the framework of Cachon and Swinney [8], we
the total price of the purchased basket of goods and is not assume a simultaneous game between the retailer and patient
contingent on purchasing a particular product. consumers. The values of vB, vM, v, v, and a are known to the
An important factor in determining the profitability of gift retailer and consumers and each patient consumer has private
cards is the type of goods consumers purchase with the cards. knowledge of his own second-period valuation at the start of the
When consumers use gift cards to pay for goods they would have game. The retailer and each patient consumer possess beliefs
purchased with cash in the future from the retailer, the retailer’s about the actions of the other players. We use the (^) over the
future cash sales decrease which causes future profit to decline. variable to denote beliefs and subscripts of d and g to present
We are unable to identify any empirical research estimating the discount and gift card cases, respectively. For the case of discounts,
amount of gift cards redeemed which consumers use to offset Cachon and Swinney [8] identified a subgame perfect Nash
future purchases they would have made without the cards at the equilibrium with rational expectations in which each consumer
retailer. However, in the Accenture survey, 56% of respondents and the retailer choose optimal actions given their belief about
stated that when redeeming gift cards, they buy themselves
something they would not normally buy.
Table 1
Description of consumers.
3. Consumer choice and demand
Segment Valuation Number

We assume a risk-neutral profit-maximizing retailer. Models Period 1 Period 2


assuming different risk tolerance and objectives within the news-
Myopic vM n/a ð1aÞD
vendor framework have also been used in the literature [31,4].
Patient vM U½v,v aD
The retailer sells a product whose demand is uncertain during a Bargain hunting n/a vB 1
finite season. The season is divided into two periods. In the first
668 M. Khouja et al. / Omega 41 (2013) 665–678

pv^ g pv^ g
Fig. 2. Consumer choice in periods 1 and 2 with gift cards. (a) Period 1 consumer choice, (b) Period 2 consumer choice, t r and (c) Period 2 consumer choice, t 4 .
d d

how the others will behave and that the beliefs are correct. period is
A patient consumer may not purchase the product in the second 8
> Rv pv pv^ g
period if the retailer’s sale price (s) turns out to be higher than her >
> aD p dt dGðvÞ ¼ aDGðpdtÞ if rtr ,
>
> d d
second period valuation (v). In the second period, both bargain- >
< Rv pv^ g pvB
hunting and patient consumers may buy the product depending on Dg2 ¼ aD v^ dGðvÞ ¼ aDGðv^ g Þ if ot o ,
>
> g d d
the discount. >
>
>
> pvB
: 1, if t Z :
Suppose instead of using discounts, the retailer gives a ‘‘free’’ d
gift card with value t to consumers who buy the product in the
second period while keeping the price at p. Patient consumers With a low-value gift card (i.e., t r ðpv^ g Þ=dÞ, the retailer creates
anticipate this promotion1 and evaluate their expected surplus in demand only to serve a portion of patient consumers. With
the second period. If the retailer provides a gift card of value t medium-value gift card, she is able to serve all patient consumers.
with the purchase of a product, since consumers value gift card Finally, she can serve even bargain hunters with a high-value gift
dollars by a lower amount than its cash equivalent [25], the card. This happens when the effective benefit of the gift card dt is
consumers’ valuation for the gift card is only dt, where d o 1 is equal to or greater than the gap between the price and the
consumers’ valuation of $1 gift card.2 This discounting of gift card bargain hunter’s valuation of the product.
dollars is due in part to consumers realizing that their probability The retailer decides her optimal stocking level in the first
of redeeming the card is less than 1. Therefore, the effective price period and sale price if she uses a discount strategy, or gift card
in the second period is pdt. Similar to the discount case in value if she uses a gift card strategy to sell excess inventory in the
Cachon and Swinney [8], we have the following lemma when the second period. This decision is analyzed in the next section.
retailer uses a gift card to salvage the product.
4. Model solution—gift cards or discounts

Lemma 1. If the retailer gives a gift card when consumers buy the
In this section, we analyze the retailer’s problem when gift
product in the second period at a price of p, then in equilibrium, there cards are used. The solution to the retailer’s problem when a
exists some vng A ½v,v such that all patient consumers with second
discount is used can be found in Cachon and Swinney [8]. Although
period value less than vng purchase in the first period, and all patient patient consumers’ belief may include whether the retailer would
consumers with value greater than vng wait for the second period. A
use the discount strategy or the gift card strategy, we assume that
consumer with value vng is indifferent between purchasing in the first the retailer commits to one strategy at the beginning of the first
or second periods.
period. Hence, as a consequence of Lemma 1, the retailer’s belief
Fig. 2a illustrates Lemma 1. The demand in the first period is about consumer behavior is v^ d if the discount strategy is used and
v^ g if the gift card strategy is used, respectively. If the gift card
Z v^ g strategy is used, then the retailer’s expected profit is
Dg1 ¼ ð1aÞDþ aD dGðvÞ ¼ ð1a þ aGðv^ g ÞÞD ¼ ð1aGðv^ g ÞÞD  xg D:  
v
pg ðq, v^ g Þ ¼ E p minðq, xg DÞcqþmaxNRðt,Ig Þ , ð2Þ
ð1Þ t

where xg ¼ 1Gðv^ g Þa, the on-hand inventory at the beginning of


In the second period, the retailer announces the value of the gift
the second period is Ig ¼ ðqxg DÞ þ , and NRðt,Ig Þ is the expected net
card. Patient consumers buy the product if they did not buy in the
revenue offset by future gift card redemptions. Unlike the case in
first period and if the effective price pdt is less than their
Cachon and Swinney [8] where the revenue stream stops at the
valuation v. Thus, as Fig. 2b, c shows, demand in the second
end of second period, the redemption of gift cards creates negative
revenue after the second period.
The profitability of gift cards depends on five factors: consumers’
1 valuation per $1 of gift card, the redemption rate, the cost of goods
Patient consumers may only anticipate that there will be a promotion of
some type equivalent to a specific monetary value in the second period. This sold with redeemed gift cards, the type of purchases made with gift
promotion may be in the form of a price discount, a gift card, a rebate, a discount cards, and the additional spending above the gift card value con-
on the second unit, etc. sumers make when they redeem their cards. If gift cards are used by
2
This is based on the assumption that consumers have an increasing linear consumers to buy goods they would not have bought without the gift
utility in gift card dollars which implies that gift cards are used on a limited
number of products. If that is the case, the total amount of gift dollars a consumer
cards, then future sales are not affected by gift cards. If redeemed gift
may receive is small and utility per $1 of gift card is constant, d. Later, we examine cards are fully or partially used by consumers to pay for goods they
the effect of strictly concave consumer utility in gift card dollars. would have purchased with cash in the future, then future sales and
M. Khouja et al. / Omega 41 (2013) 665–678 669

profits are reduced. We refer to the reduction of future cash sales in DL DM DH


the latter case as the future cash sales offset. Consumers’ utility from
each $1 of gift card, gift cards redemption rate, the future cash sales tH
offset, and additional spending above the cards’ value may be
different for different individuals. For simplicity, we assume that all

Optimal gift card value


consumers (bargain hunters and patient) are the same with regard to
these four aspects. This assumption allows us to identify the condi-
tions under which it is better to use gift cards instead of discounting tM
the product.
Let m be the average gross margin per $1 of sales (0o mo 1),
tL(D)
then 1m is the cost of goods sold per $1 of redeemed gift card.
Also, let Z be the fraction of each $1 of redeemed gift card that goes
to future cash sales offset (0r Z r1). In other words, Z is the
decline in future sales per $1 of redeemed gift card which results in
an mZ loss in future profit. The total cost per $1 of redeemed gift
card to the retailer is r ¼ 1m þmZ ¼ 1mð1ZÞ. Since not all gift
cards are redeemed, the total cost per $1 of gift card is $rr. Some
Demand realization (D)
consumers may spend more than the gift card value when they
redeem the gift cards. Let y be the average additional spending Fig. 3. Optimal second-period decision as a function of D.
paid for in cash per $1 of redeemed gift card. Then, the additional
revenue per gift card is ryt which is reduced by Zryt to account for
purchases which consumers would have made at the retailer is the low gift card value, which is contingent on the demand realization
without the gift card. This results in a net increase in sales of and remaining inventory.
rytZryt ¼ ð1ZÞryt per gift card. The contribution of the addi-
Proposition 1 is illustrated in Fig. 3. It is reasonable to assume
tional spending to the retailer’s expected profit is mð1ZÞryt per
that the retailer’s profit from each unit sold to the bargain hunters
gift card. Since mð1ZÞ ¼ 1r, mð1ZÞryt can be rewritten as
satisfies 0 r pbrt H o c, i.e., ðpcÞ=ðpvB Þ o br=d rp=ðpvB Þ.
ð1rÞryt. Therefore, the expected net revenue function in (2) is
8 If pbrt H 4c, then the retailer’s order quantity in the first period
> pv pv^ g will be infinity, i.e., q ¼ 1.
>
> ðpðrð1rÞyÞrtÞminðaDGðpdtÞ,Ig Þ if rtr ,
>
> d d
>
< By substituting the optimal gift card value from Proposition 1
pv^ g pvB into the retailer’s expected profit function in (2), we can show
NRðt,Ig Þ ¼ ðpðrð1rÞyÞrtÞminðaDGðv^ g Þ,Ig Þ if ot o ,
>
> d d
>
> that the retailer’s expected profit function with a gift card
>
> pvB
: ðpðrð1rÞyÞrtÞIg if t Z : strategy is unimodal as follows.
d
ð3Þ Proposition 2. The retailer’s expected profit pg ðq, v^ g Þ is quasicon-
We assume y o r=ð1rÞ. Otherwise, it is optimal for the retailer to cave in q, and the optimal order quantity is determined by the unique
order infinity in the first period. Let b  rð1rÞy, which is the net solution to the first-order condition,
cost per $1 of redeemed gift card. It can be shown that b decreases
dpg ðq, v^ g Þ
with the gross margin m and the additional spending rate above ¼ pcpFðDH Þ þðpbrt H ÞFðDL Þ
dq
the gift card value y, and increases with the future cash sales offset Z DH  
per $1 of redeemed gift cards Z. Eq. (3) becomes brðpvÞ þ dp
þ 2br t L ðxÞ dFðxÞ ¼ 0: ð5Þ
8 DM 2brd
> pv pv^ g
>
> ðpbrtÞminðaDGðpdtÞ,Ig Þ if rtr ,
>
> d d
>
<
pv^ g pvB
NRðt,Ig Þ ¼ ðpbrtÞminðaDGðv^ g Þ,Ig Þ if oto , ð4Þ If we assume that v^ d ¼ v^ g  v, ^ i.e., no matter which strategy
>
> d d
>
> the retailer commits in the first period, the discount strategy or
>
> pv B
: ðpbrtÞIg if t Z : the gift card strategy, a patient consumer whose second-period
d
valuation is less than v^ purchases the products in the first period,
We can find the retailer’s optimal second-period gift card whereas a patient consumer whose valuation is greater than v^
policy when the gift card strategy is used as follows: wait for the second period. A consumer with valuation of v^ is
indifferent between purchasing in the first or second periods. We
Proposition 1. When the gift card strategy is used, define the critical
can show that when ðpcÞ=ðpvB Þ o br=d o 1, the retailer orders
demand levels DL ¼ q=ðxg þ ðpbrt M ÞGðpdt M Þa=ðpbrt H ÞÞ, DM ¼
more in the first period and obtains larger expected profit under
q=ðxg þ Gðpdt M ÞaÞ, and DH ¼ q=xg , where L, M, and H stand for
the gift card strategy than the discount strategy, i.e., qng Zqnd and
low, medium, and high, respectively. Then, given a demand level D,
png ðqng , v^ g Þ Z pnd ðqnd , v^ d Þ. When 1 o br=d o p=ðpvB Þ, the retailer
there is a unique optimal gift card value determined by
8 orders more in the first period and obtains larger expected profit
< t L ðDÞ if DM o D r DH ,
> under the discount strategy than the gift card strategy, i.e., qnd Z qng
t ðDÞ ¼ t M
n if DL o D rDM , and pnd ðqnd , v^ d Þ Z png ðqng , v^ g Þ. Therefore, five factors, consumers’
>
:t
H if D rDL , valuation per $1 of gift card (d), the redemption rate (r), the cost
of goods sold per $1 of redeemed gift cards (1m), the future cash
where t H ¼ ðpvB Þ=d is the high gift card value, sales offset per $1 of redeemed gift cards (Z), and additional
tM ¼ argmax ðpbrtÞðvp þ dtÞ spending rate above the gift card value (y) interact to determine
ðpvÞ=d r t r ðpv^ g Þ=d gift card profitability. It is noteworthy to point out that slippage
may not be the determining factor of gift card profitability. Even
is the medium gift card value, and
with 100% redemption rate of gift cards, i.e., r ¼ 1, offering gift
pv^ g ðDqÞðvvÞ cards can still be more profitable than discounting the product.
t L ðDÞ ¼ 
d aDd For r ¼ 1, the condition for gift cards to be more profitable than
670 M. Khouja et al. / Omega 41 (2013) 665–678

discounting the product is ðpcÞ=ðpvB Þ o b=d o1. Substituting


for b and simplifying gives dðpcÞ=ðpvB Þ o 1mð1ZÞð1 þ yÞ o d
which is satisfied for many retailers with large m, small Z, large y,
and selling products with reasonable p, c, and vB. Therefore, for
retailers with large margins who can control future cash sales
offset, gift cards are more profitable than discounting.
If patient consumers behave consistently by valuing each $1 of
gift card by their actual card redemption probability, then
Corollary 1 follows.

Corollary 1. If consumers valuation per $1 of gift card is equal to


their redemption probability (d ¼ r) and do not fully offset future
cash purchases with gift cards redemption (Z o 1), then the order
quantity is equal to or larger than quantity under the discounting Fig. 4. Step-wise redemption rate.
strategy and the optimal expected profit is equal to or larger than
under the discounting strategy.
U (t)
Corollary 1 shows that gift cards can outperform discounts
when consumers act patiently in evaluating their ‘‘buy now’’ vs.
‘‘buy later‘‘ decision as well as act consistently by valuing each $1
p − vB
of gift card by their actual redemption probability. For this out- δT
performance to occur, consumers must buy some goods with gift δT
cards they would not have bought in the future with cash from p−v
the retailer. We have assumed that v^ d ¼ v^ g in our discussion thus p − vˆg
far. Since v^ d and v^ g are the results from the subgame perfect Nash δ
equilibrium with rational expectations, they are not equal in
general. We numerically investigate whether the gift cards
strategy or discount strategy is better for the retailer to salvage
excess inventory in Section 6. t
tM T tH
Cachon and Swinney [8] derive patient consumers’ best
response and show the existence of the subgame perfect Nash Fig. 5. Piece-wise linear utility in gift card dollars.
equilibrium for the discount strategy. Similar results can be
shown for the gift card strategy. We omit the results here to
minimize repetition. condition for t M rT ot H to hold:
pv pvB
rT o : ð6Þ
d d
While more consumers gain utility by redeeming their gift
5. Model extensions cards, we also expect the marginal utility from gift card dollars to
become decreasing at large gift card amounts. Suppose there is a
In this section, we consider changes in consumer behavior single threshold value4 of t at which consumers’ marginal valua-
when a higher-valued gift card is given. Normally, when gift card tion for gift card dollars decreases and this threshold value is the
value becomes salient, consumers are more likely to redeem the same as the threshold value where the redemption rate increases,
cards. Suppose the redemption rate increases from r to rT if the i.e., T. The consumers’ utility function is
gift card value is greater than or equal to a threshold value of t, (
denoted by T.3 From the results in Section 4, while the value of tH dt if t r T,
UðtÞ ¼ ð7Þ
is independent of r, the values of tM and tL(D) depend on r and dT þ dT ðtTÞ if t 4 T:
the resulting tM and tL(D) must be consistent with the values of r
Let t M and t H denote the optimal medium and high gift card
used to compute them. The gap between tH and tM is usually large.
values, respectively, for the problem with one-threshold step-
Numerical analysis in Section 6 revealed that for reasonable
wise increasing redemption rate and a one-threshold piece-wise
problem parameters, the value of tH is much larger than the value
linear concave consumer utility for gift card dollars. Lemma 2
of tM. Therefore, it is safe to assume that there is a single threshold
shows that condition (6) is also a sufficient condition to make
value T, at which gift cards’ redemption rate increases. If t H o T
t H 4T and t M o T.
then the problem is solved using Eq. (5) with a single value of r. If
t M r T o t H , then as Fig. 4 shows, one redemption rate, r, is
Lemma 2. Condition (6) is a sufficient condition for t M r T o t H to
applicable at tM and tL(D), and a larger redemption rate, rT , is
hold.
applicable at large gift card values including tH. In this case, the
medium-value gift card for the one-threshold step-wise increas- Fig. 5 shows an example when t M rT o t H . In this case, one gift
ing redemption rate problem is the same as the tM in Section 4, card valuation, d, is applicable at tM and tL(D), and a smaller gift
i.e., t M ¼ minfðbrðpvÞ þ drÞ=2brd,ðpv^ g Þ=dg. For t M rT to hold, a card valuation, dT , is applicable at large gift card values including
condition of ðpv^ g Þ=d rT is sufficient. Since v^ g results from the t H . When the redemption rate is step-wise increasing and utility
game equilibrium and v o v^ g , ðpvÞ=d rT is sufficient for t M r T for the gift card is piece-wise linear concave at the single thresh-
to hold. Substituting t H ¼ ðpvB Þ=d gives the following sufficient old of T, we have the following results.

3 4
There can be many gift card value thresholds at which the slope of the There can be many gift card value thresholds at which the slope of the
redemption rate changes. We consider only one value. consumer utility function changes. We consider only one value.
M. Khouja et al. / Omega 41 (2013) 665–678 671

Proposition 3. If condition (6) holds, then the retailer’s expected from offering a high value gift card and the second part implies
profit p g ðq, v^ g Þ with a one-threshold step-wise increasing redemption that this revenue is less than unit cost.
rate and a one-threshold piece-wise linear concave consumer utility
for gift card dollars is quasiconcave in q, and the optimal order
quantity is determined by the unique solution to the first-order 6. Discussion and numerical analysis
condition,
In this section, we perform sensitivity analysis on the model
dp g ðq, v^ g Þ  LÞ
¼ pcpFðDH Þ þðpbrT t H ÞFðD with gift cards and then compare the performance of using free
dq
gift cards for selling any remaining inventory to discounts. We
Z DH  
brðpvÞ þ dp use the example with the parameters shown in Fig. 6 to illustrate
þ 2br t L ðxÞ dFðxÞ ¼ 0, ð8Þ some of the results. Figs. 6 and 7 show some counter-intuitive
DM 2brd
results. As gift card redemption rate increases, the retailer’s
where expected profit from giving gift cards is expected to decrease.
q However, Fig. 6c shows that this may not be the case and that the
L¼
D expected profit can be increasing for a range of the redemption
xg þ ðpbrtM ÞGðpdtM Þa=ðpbrT t H Þ
rate. The reason for this increase in expected profit is patient
and consumers’ belief about q.^ As the redemption rate increases, the
retailer orders a smaller q as shown in Fig. 6a. As a result, patient
ðpvB ÞðddT ÞT consumers have a smaller probability of having a unit available in
t H ¼ ,
dT the second period and, thus, more patient consumers buy the
product in the first period, i.e., a larger v^ g as shown in Fig. 6b. This
and DM, DH, tM and tL(D) are the same as those in Section 4.
causes the expected profit to increase for some interval of r.
Note that we still assume that 0 r brT t H o c. The first part of Similar counter-intuitive results are shown in Fig. 6d–f with
the inequality implies that the retailer gets some revenue per unit respect to future cash sales offset (Z). As Z increases, expected

Fig. 6. Optimal order quantity, indifferent patient consumers’ valuation, and expected profit vs. redemption rate and future cash sales offset, a ¼ 0:5, p ¼ 75, c ¼ 45, vB ¼ 30,
vM ¼ 93:75, v ¼ 45, v ¼ 67:5, r ¼ 0:7, d ¼ 0:7, Z ¼ 0:4, y ¼ 0:2, m ¼0.4, Gamma distribution with m ¼ 100 and s ¼ 25.
672 M. Khouja et al. / Omega 41 (2013) 665–678

η
η

η
η

η
η

η
π η

Fig. 7. Optimal order quantity, indifferent patient consumers’ valuation, and expected profit vs. additional spending and future cash sales offset, a ¼ 0:5, p ¼75, c ¼ 45,
vB ¼ 30, vM ¼ 93:75, v ¼ 45, v ¼ 67:5, r ¼ 0:7, d ¼ 0:7, m¼ 0.4, Gamma distribution with m ¼ 100 and s ¼ 25.

profit increases for a range of Z before it starts decreasing. The valuation for the product in the first period and more of them buy
reason is the patient consumers’ belief about q,^ which is similar to in the first period. This causes the expected profit to be decreasing
the case of increases in the redemption rate. for a range of consumer gift card valuation starting at about
The effect of consumer spending above the gift card value by dT ¼ 0:45.
100y% for different future cash sales offset Z is shown in Fig. 7. As To examine the performance of using gift cards to sell excess
can be seen, the effect of y strongly depends on Z. The results are inventory vs. discounting it, we conducted a numerical experi-
again counter-intuitive in the sense that increased spending ment using the parameters shown in Table 2, which resulted in
above the gift card value may lead to a decrease in the expected 14,400 problems. Similar to Cachon and Swinney [8], we dis-
profit for a range of y as shown in Fig. 7c. For Z ¼ 0:4, the retailer carded uninteresting problems where all patient consumers buy
has lower cost for gift card redemption and rapidly increases the in the first period (i.e., act as myopic) or wait till the product is
order quantity with increases in y as shown in Fig. 7a. As a result, salvaged using a large discount or high-value gift card (i.e., act as
patient consumers have a larger probability of finding a unit bargain hunters). We begin by discussing the results for con-
available in the second period and, thus, fewer patient consumers sumers who act patiently in deciding to buy in period 1 or wait
buy the product in the first period, i.e., a decreasing v^ g as shown until period 2 as well as act consistently by carrying through with
in Fig. 7b, which results in a decrease in expected profit. For their redemption intentions, i.e. r ¼ d. The results in Table 3
Z ¼ 0:8, much of the additional spending above gift card value shows the average png =pnd over coefficient of variations of
would have been made with cash in the future and the retailer s=m ¼ 0:25,0:5, and 1. The results of Table 3 are illustrated in
has a larger cost for gift card redemption. Therefore, the retailer Fig. 9. As the table and figure show, much of the advantage of gift
increases the order quantity only slightly with increases in y as cards is eliminated when b ¼ 1mð1ZÞðy þ 1Þ gets close to 1, and
shown in Fig. 7a. As Fig. 7b shows, indifferent patient consumer’ at b ¼ 1, the optimal expected profits from the gift card and
valuation in the second period v^ g decreases only slightly and discounting strategies become equal. A b ¼ 1 implies that all
expected profit increases. consumers use gift card redemptions to offset future cash pur-
The effect of increasing redemption rate and decreasing con- chases they would have made at the retailer, i.e. Z ¼ 1. If
sumer valuation of gift cards shown in Fig. 8 is again counter- administering a gift card program has a fixed cost larger than
intuitive. An increase in the redemption rate and/or a decrease in the cost of administering an instant discount, then discounting
consumer valuation of gift cards at high gift card values may may have larger optimal expected profit even at b o 1. Fig. 10
increase the expected profit. For the example in Fig. 8a–c, the shows the average png =pnd over all values of a when b ¼ 0:9 vs.
larger redemption rate causes the retailer to order a smaller coefficient of variation (s=m). As the coefficient of variation
quantity. As a result, patient consumers have a lower probability increases, gift cards’ profit advantage improves. When demand
of finding a unit available in the second period. Therefore, they has high variability, overstocking cost can be substantial. Using
have a higher valuation for the product in the first period and gift cards, the retailer can create some second period demand
more of them buy in the first period. This causes the expected from patient consumers without having to attract the bargain
profit to increase for a range of the larger redemption rates up to hunters which reduces the overstocking cost and increases the
about rT ¼ 0:76. Similarly, in Fig. 8d–f, smaller consumer gift card expected profit. Fig. 11 shows the average png =pnd over three values
valuation causes the retailer to order a smaller quantity. As a of the coefficient of variation (0.25, 0.50, and 1.0) and b ¼ 0:9 vs.
result, patient consumers have a lower probability of finding a the proportion of patient consumers (a). As the figure shows, the
unit available in the second period. Therefore, they have a higher profit advantage of gift cards decreases as a increases. This can
M. Khouja et al. / Omega 41 (2013) 665–678 673

Fig. 8. Optimal order quantity, indifferent patient consumers’ valuation, and expected profit vs. higher stepwise redemption rate and lower valuation, a ¼ 0:5, p ¼75, c¼ 45,
vB ¼ 30, vM ¼ 93:75, v ¼ 45, v ¼ 67:5, r ¼ 0:7, d ¼ 0:7, m¼ 0.4, Z ¼ 0:4, y ¼ 0:2, Gamma distribution with m ¼ 100 and s ¼ 25.

Table 2 Table 3
Parameter values used in numerical experiments. Mean Optimal expected profit using gift cards/Optimal expected profit using
discounts vs. b, r ¼ d ¼ 0:7.
Parameter Values
Parameter, Mean value of png =pnd Average value
Demand Gamma b png =pnd
distribution a ¼ 0:01 a ¼ 0:25 a ¼ 0:50 a ¼ 0:75 a ¼ 1:00
m 100
s f25,50,100,150g 0.3 2.114 1.850 1.636 1.488 1.381 1.694
p 20 0.5 1.465 1.356 1.288 1.239 1.210 1.312
c f5,10,15g 0.7 1.268 1.189 1.131 1.099 1.082 1.514
vM f24,28g 0.9 1.086 1.046 1.028 1.019 1.016 1.039
vB f2,4g
½v,v f½4,20,½6,8,½12,14,½18,20g
a f0:01,0:25,0:50,0:75,1g
b f0:10,0:30,0:50,0:70,0:90g
have to offer a large value gift card to motivate them to buy the
½d, r f½0:7,0:4,½0:7,0:7,½0:7,1:0g leftover product inventory at the regular price, which decreases
the profitability of the gift card.
We also examined the effect of one additional aspect of
also be observed for other values of b, albeit png =pnd are larger for consumer behavior, time-inconsistent preferences and procrasti-
smaller values of b. This is because as the proportion of patient nation, on the profitability of gift cards. The economic literature
consumers increases, there are fewer myopic consumers who buy provides evidence that people could have a bias for the present
the product if their valuation is met. With a larger the number of [29,23,12], which will negatively impact gift cards since they are
patient consumers, it is likely that more of them will wait to for future redemption. If consumers have a preference for
purchase the product. Thus, it is more likely that the retailer will immediate utility over delayed utility, then they value each $1
674 M. Khouja et al. / Omega 41 (2013) 665–678

of the gift card, which is to be redeemed in the future, by a lower consumers may procrastinate in redeeming gift cards, in parti-
value than the redemption probability, i.e. d o r. On the other cular gift cards with long deadlines, and end up not redeeming
hand, other researchers [27] provide evidence that in some cases, them, thus in some cases d 4 r. Both present-bias and procrasti-
nation have been found to have a strong influence on the
performance of cash mail-in rebates [14,20]. Table 4 shows
png =pnd over coefficient of variations on s=m ¼ 0:25, 0.5, and 1 for
present-biased consumers (d ¼ 0:4 and r ¼ 0:7), consistent con-
sumers (d ¼ 0:7 and r ¼ 0:7), and procrastinators (d ¼ 0:7 and
r ¼ 0:4). The table shows that if consumers are present-biased,
πg*/πd*

the optimal expected profit with gift cards is only 14.4% above
discounting’s expected profit when the cost of each $1 of gift card
to the retailer is only $0.50. When the cost of each $1 of gift card
to the retailer is $0.90, the optimal expected profit with discounts
is larger than with gift cards for present-biased consumers.
To identify the type of products for which using free gift cards
to sell excess inventory is more profitable than discounting it, we
examine the solutions to the problems shown in Table 5 under
Fig. 9. Mean Optimal expected profit using gift cards=Optimal expected profit
increasing redemption rate and diminishing piece-wise linear
using discounts vs. b, r ¼ d ¼ 0:7. consumer utility in card value. The low value for card redemption
is r ¼ 0:7 and the high consumer valuation for gift cards is d ¼ 0:7.
The gift card value threshold for the redemption rate increase and
consumer gift card valuation decrease is T¼60. Because of the
threshold of T¼60, low-ticket products I and II are not affected by
the change in the redemption rate and the consumer valuation of
gift cards. In the first case there is no increase in the redemption
rate and consumer gift card valuation (i.e., dT ¼ 0:7 and rT ¼ 0:7),
which is our basic model. For this case, gift cards have a 0.5% profit
advantage over discounting. In the second case where rT ¼ 0:8 and
dT ¼ 0:6, gift cards still outperform discounting. Furthermore, the
increase in dT and rT results in better performance of gift cards for
products III and IV. This occurs because the retailer reduces the
order quantity, causing more patient consumers to buy in the first
period. For the third case where rT ¼ 0:9 and dT ¼ 0:5, gift cards
outperform discounting for only products I and II whereas gift
Fig. 10. Mean Optimal expected profit using gift cards=Optimal expected profit
cards under-perform discounting for product IV. This indicates that
using discounts vs. s=m, r ¼ d ¼ 0:7, b ¼ 0:9. gift cards should not be used for high-ticket products where the
gift card value is large, rT is very high, and dT is very low.

7. Conclusions

We developed a model in which the retailer offers a ‘‘free’’ gift


card along with a regularly priced product at the end of the selling
season. We followed the model framework developed by Cachon
and Swinney [8] and developed the sufficient optimality condi-
tion for the order quantity. We also compared the performance of
the ‘‘free’’ gift card strategy with that of the discounting strategy.
We found that under certain conditions, the gift card strategy can
be a viable alternative to a discounting strategy for retailers facing
patient consumers.
Numerical analysis shows that in the presence of patient con-
Fig. 11. Mean Optimal expected profit using gift cards/Optimal expected profit sumers, the expected profit function may exhibit a counter-intuitive
using discounts vs. a, r ¼ d ¼ 0:7, b ¼ 0:9. behavior. For example, increases in the redemption rate in some

Table 4
Mean value of Optimal expected profit using gift cards/Optimal expected profit using discounts.

Consumer type Parameters Mean value of png =pnd Average value png =pnd

dr b a ¼ 0:01 a ¼ 0:25 a ¼ 0:50 a ¼ 0:75 a ¼ 1:0

Present-biased  0.30 0.50 1.251 1.174 1.122 1.093 1.078 1.144


0.70 1.000 0.999 0.998 0.997 0.997 0.998
Consistent 0.00 0.50 1.465 1.356 1.288 1.239 1.210 1.312
0.70 1.268 1.189 1.131 1.099 1.082 1.154
Procrastinators 0.30 0.50 2.255 1.954 1.700 1.526 1.399 1.767
0.70 1.687 1.534 1.419 1.346 1.286 1.455
M. Khouja et al. / Omega 41 (2013) 665–678 675

Table 5
Performance of gift cards for different products: a ¼ 0:50, d ¼ 0:70, r ¼ 0:70, T¼ 60, y ¼ 0:20, m¼ 0.40, Z ¼ 0:4, Gamma with m ¼ 100 and s ¼ 25.

Product p c vB v v vM rT ¼ 0:70 & dT ¼ 0:70 rT ¼ 0:80 & dT ¼ 0:60 rT ¼ 0:90 & dT ¼ 0:50

png =pnd qng =qnd png =pnd qng =qnd png =pnd qng =qnd

I 25 15 10 15 22.5 31.25 1.005 1.091 1.005 1.091 1.005 1.091


II 50 30 20 30 45.0 62.50 1.005 1.091 1.005 1.091 1.005 1.091
III 75 45 30 45 67.5 93.75 1.005 1.091 1.008 1.044 1.000 1.006
IV 100 60 40 60 90.0 125.00 1.005 1.091 1.007 1.036 0.994 0.991

range may lead to an increase in the expected profit. Also, the strategic consumer with a second period valuation of v is
expected profit may decrease with increases in consumer spending Z t
beyond gift card value at redemption and decreases in future cash cg ðv, v^ g , q^ g Þ ¼ ðvp þ dtÞhg ðt, v^g , q^g Þ dt:
sales offset. This is because when these changes in problem ðpvÞ=d

parameters occur the retailer decreases (increases) the order quan- Since hg ðÞ is independent of v due to the rational expectations
tity which decreases (increases) patient consumers’ probability of hypothesis, this expression is increasing in v. Therefore, there exists
finding a unit available in the second period, thus more (less) a unique vng for which vM p ¼ cg ðvng , v^ g , q^ g Þ. All strategic consumers
consumers buy the product in the first period.
with greater second period valuations prefer to wait for the gift card
If patient consumers are consistent in the sense of valuing
promotion, while all strategic consumers with lower second period
each $1 by the probability of redeeming it, offering gift cards can
valuations prefer to purchase in the first period. &
still be more profitable than discounting. The most important
factor to this profitability is that sufficient number of consumers Proof of Proposition 1. Proof is similar to that of Lemma 2 of
use the gift cards to buy something they would not normally buy Cachon and Swinney [8]. When D 4DH ¼ q=xq (i.e., qo xq D), there
from the retailer. When such is the case, this gift card profit is no inventory left at the end of the first period. When D r DH , the
advantage is larger for retailers with high average margins and retailer has two choices: (i) selling to the strategic consumers only
more consumers spending beyond the cards’ value. by setting t oðpvB Þ=d. Any gift card value in the range ðpv^ g Þ=
Gift cards’ profit advantage over discounting increases with the d ot o ðpvB Þ=d is never optimal since t ¼ ðpv^ g Þ=d always yields
variability of demand. Using gift cards can mitigate the risk of greater profit. Therefore, ðpvÞ=d r t rðpv^ g Þ=d (i.e., v^ g r pdt r
overstocking and, hence, the retailer orders a larger quantity relative v) and the optimal gift card value is the solution to
to discounting as the demand variability increases. Gift cards are not
argmax ðpbrtÞminðGðpdtÞaD,Ig Þ:
always a good alternative to discounting. When the product price is ðpvÞ=d r t r ðpv^ g Þ=d
high, the retailer has to offer a high value gift card when the realized
demand is small. As a result, consumers are more likely to redeem If D rq, then the second-period inventory is enough to sell to all
their gift cards while their marginal valuation of gift cards is likely to strategic consumers, i.e.,
be small. Our results indicate that for high price products, the Iq ¼ qxq D Zqxq qZ ð1xq ÞD ¼ Gðv^ g ÞaD ZGðpdtÞaD:
retailer is better off by discounting the product instead of using gift
cards. In addition, gift cards are more likely to result in lower Therefore, the retailer’s optimization problem becomes
expected profit when consumers are present-biased. vðpdtÞ
Some extensions to our models are possible. One interesting tM ¼ argmax ðpbrtÞ aD:
ðpvÞ=d r t r ðpv^ g Þ=d vv
extension would be incorporating a different behavior for bargain
hunters from patient consumers. For example, the redemption Since ðpbrtÞðvp þ dtÞ is concave in t, there may be an interior
rate and valuation for gift cards may be different for bargain optimum determined by the solution to the first order condition,
hunters and patient consumers. Incorporating different redemp- which yields t n ¼ ðpvÞ=2d þp=2br, if t n rðpv^ g Þ=d; otherwise, the
tion rates for bargain hunters and patient consumers would optimal gift card value is on the boundary. Therefore, t M ¼ min
require modeling the arrival process of consumers when a large fðbrðpvÞ þ drÞ=2brd,ðpv^ g Þ=dg. Note that t M is independent of D
value gift card is offered. Demand for leftover inventory with a and a, but does depend on v^ g and v. The optimal profit in this region
large value gift card can be assumed to form a queue of both types is ðpbrtÞððvp þ dtÞ=ðvvÞÞaD.
of consumers with a parameter which represents the proportion Now consider the case in which q oD rDH . In this region, if the
of each type of demand in the queue. retailer offers a high gift card, she is inventory constrained,
whereas if she offers a low gift card, she is demand constrained.
For any demand level D, there exists some critical gift card t L ðDÞ,
Appendix
such that the retailer’s net revenue function is
(
ðpbrtÞIg if t Zt L ðDÞ,
Proof of Lemma 1. Proof is similar to that of Lemma 1 of Cachon NRðt,Ig Þ ¼
and Swinney [8]. We include the proof here for completeness. The ðpbrtÞGðpdtÞaD if t ot L ðDÞ:
surplus to a strategic consumer who purchases in the first period is
In particular, t L ðDÞ is determined by solving GðpdtÞaD ¼ Ig for t,
vM p. In the second period, a strategic consumer only purchases the
which yields
product if (1) the effective price pdt is less than or equal to her
second period valuation, which requires t Z ðpvÞ=d and (2) there is pv^ g vv
Rt t L ðDÞ ¼  ðDqÞ:
inventory available to purchase. Let ðpvÞ=d hg ðt, v^ g , q^ g Þ dt be a d adD
strategic consumer’s belief of the probability that the effective price Recall that tM is the maximizer of ðpbrtÞGðpdtÞaD. Because
is less than or equal to v and the consumer receives a unit, where t ðpbrtÞGðpdtÞ is concave in t, if t L ðDÞ r t M , the optimal gift card
is the upper bound of a gift card value which is assumed to be a value is t L ðDÞ, whereas if t L ðDÞ 4 t M , the optimal gift card value is
common knowledge. Then the second period expected surplus of a tM. Therefore, there exists some critical demand level DM such
676 M. Khouja et al. / Omega 41 (2013) 665–678

that for D oDM , it is optimal to offer a gift card of tM, and for A local optimum is achieved (j0 ðqÞ ¼ 0) if and only if , for any q on the
D 4 DM , it is optimal to offer a gift card of t L ðDÞ. DM is determined interior of the support of f,
by solving t L ðDÞ ¼ t M for D which yields brðpvÞ f ðDH Þ dDH f ðDL Þ dDL
0¼ þðpbrt H Þ
q d f ðDM Þ dq f ðDM Þ dq
DM ¼ :  
xg þ GðpdtM Þa brðpvÞ þ dp dDM
2br t M
2brd dq
Z DH
2br 1 1
(ii) Selling to the bargain hunters by setting t ¼ ðpvB Þ=d. If the  ðvvÞ dFðxÞ: ð11Þ
d f ðDM Þ DM ax
retailer offers gift cards of t ¼ ðpvB Þ=d  t H , the second period net
revenue is ðpbrt H ÞIg . This yields a greater profit than offering Recall that the MSLR assumption implies f ðgxÞ=f ðxÞ is monotonic in x
gift cards of tM if and only if ðpbrt M ÞGðpdt M ÞaD rðpbrt H ÞIg , for all g r 1. Assume that f ðgxÞ=f ðxÞ is weakly increasing in x. (The
i.e., proof is identical if f ðgxÞ=f ðxÞ is weakly decreasing in x. Since v r p,
the first term is negative and increasing in q by MSLR assumption.
q
Dr  DL : Similarly, the second term is positive and increasing in q by MSLR
xg þ ðpbrtM ÞGðpdtM Þa=ðpbrt H Þ assumption. The third term is constant, while the fourth term is
negative. We will now demonstrate that the fourth term is also
Since tM maximizes ðpbrtÞGðpdtÞ in the interval ðpvÞ=d r
increasing in q by performing a change of variable. let yq ¼ x, such
t rðpv^ g Þ=d rðpvB Þ=d ¼ t H ,
that dx ¼ q dy, and let gH ¼ dDH =dq and gM ¼ dDM =dq. Then, the
  
pvB  pv^ g integral in the fourth term is equivalent to
pbr Gðv^ g Þ r pbr Gðv^ g Þ rðpbrt M ÞGðpdt M Þ,
d d Z DH Z gH
f ðxÞ f ðyqÞ
dx ¼ dy:
which implies DL r q. Thus, if demand is less than DL, it is optimal DM xf ðDM Þ gM yf ðgM qÞ
to offer high value gift cards to clear all inventory and sell to the Differentiating with respect to q,
bargain hunters. & ! Z
d
Z gH
f ðyqÞ gH dy  d f ðyqÞ 
dy ¼ r 0,
Proof of Proposition 2. Proof is similar to that of Lemma 3 of Cachon dq gM yf ðgM qÞ gM y dq f ðgM qÞ
and Swinney [8]. The retailer’s expected profit under a gift card
strategy is where the inequality follows from the MSLR assumption combined
Z DH Z 1 with the fact that y Z gM . Thus it follows that the fourth term in (11)
pg ðq, v^ g Þ ¼ p xg x dFðxÞ þp q dFðxÞcq is increasing in q. Each term on the right hand side of (11) is
0 DH increasing in q), and if a solution to the equation exists, it is unique.
Z DL This implies jðqÞ has at most one interior optimum, and conse-
þ ðpbrt H Þðqxg xÞ dFðxÞ quently pg ðq, v^ g Þ is quasi-concave in q. &
0
Z DM
Result 1. When ðpcÞ=ðpvB Þ o br=d o1, the retailer orders more in
þ ðpbrt M ÞGðpdt M Þax dFðxÞ
DL the first period and obtains larger expected profit under the gift card
Z DH strategy than the discount strategy, i.e., qng Z qnd and png ðqng , v^ g Þ Z
þ ðpbrt L ðxÞÞðqxg xÞ dFðxÞ: ð9Þ pnd ðqnd , v^ d Þ. When 1o br=d op=ðpvB Þ, the retailer orders more in the
DM
first period and obtains larger expected profit under the discount strategy
The derivative of (9) with respect to q is than the gift card strategy, i.e., qnd Zqng and pnd ðqnd , v^ d Þ Z png ðqng , v^ g Þ.

dpg ðq, v^ g Þ Proof. From the expression for dpd ðq, v^ d Þ=dq in Cachon and
¼ p½1FðDH Þc þ ðpbrt H ÞFðDL Þ
dq Swinney [8] and (5), we have
Z DH  
dt L ðxÞ Z Dh
þ ðpbrt L ðxÞÞbr ðqxg xÞ dFðxÞ dpd ðq, v^ d Þ dpg ðq, v^ g Þ
DM dq  ¼ sl FðDl Þ þ ð2sh ðxÞvÞ dFðxÞðpbrt H ÞFðDL Þ
dq dq Dm
¼ pcpFðDH Þ þ ðpbrt H ÞFðDL Þ Z DH  
Z DH   brðpvÞ þ dp
brðpvÞ þ dp  2br t L ðxÞ dFðxÞ: ð12Þ
þ 2br t L ðxÞ dFðxÞ: ð10Þ DM 2brd
DM 2brd
Since v^ d ¼ v^ g , we have xd ¼ xg , pdt L ðxÞ ¼ sh ðxÞ, and Dh ¼ DH . From
Let jðqÞ ¼ dpg ðq, v^ g Þ=dq. Note that jð0Þ ¼ pc 40 and limq-1 the second and fourth terms of the right hand side of (12), we can
jðqÞ ¼ pcðbr=dÞðpvB Þ o0 since br=d 4ðpcÞ=ðpvB Þ. There- simplify the terms inside the integration such that
fore, pg ðq, v^ g Þ has at least one local maximum. To demonstrate    
quasi-concavity of pg ðq, v^ g Þ, we must show that jðqÞ has a unique brðpvÞ þ dp br
ð2sh ðxÞvÞ2br tL ðxÞ ¼ 1 ð2sh ðxÞvpÞ, ð13Þ
2brd d
zero, i.e., that pg ðq, v^ g Þ has a single local maximum. Given the
asymptotic behavior of jðqÞ, a sufficient condition for this to occur
is that jðqÞ itself has at most one local optimum. If this is the case, where 2sh ðxÞvp is negative since sh ðxÞ rv r p. Therefore,
then jðqÞ is either quasi-concave or quasi-convex, and j0 ðqÞ will have if ðpcÞ=ðpvB Þ o br=d o 1, then (13) is negative, i.e., 2sh ðxÞv o
at most one interior zero. Substituting for t L ðxÞ, j0 ðqÞ ¼ d pg
2 2br½ðbrðpvÞ þ dpÞ=2brdt L ðxÞ; whereas if 1 o br=d o p=ðpvB Þ,
^ 2
ðq, v g Þ=dq is given by then (13) is positive, i.e., 2sh ðxÞv 42br½ðbrðpvÞ þ dpÞ=2brd
t L ðxÞ.
brðpvÞ dDH dDL
j0 ðqÞ ¼  f ðDH Þ þ ðpbrt H Þf ðDL Þ Let ðsm =sl ÞGðsm Þ  Ld and ððpbrt M Þ=ðpbrt H ÞÞGðpdt M Þ  Lg .
d dq dq
  Then, Dl ¼ q=ðxd þ Ld aÞ and DL ¼ q=ðxg þ Lg aÞ. Note that sm ¼ max
brðpvÞ þ dp dDM
fv=2, v^ d g and t M ¼ minfðbrðpvÞ þ dpÞ=2brd,ðpv^ g Þ=dg. To find out
2br t M f ðDM Þ
2brd dq
the sign of (12), we need to investigate four cases.
Z DH
2br 1 Case1: v=24 v^ d and ðbrðpvÞ þ dpÞ=2brd oðpv^ g Þ=d. Therefore,
 ðvvÞ dFðxÞ:
d DM ax sm ¼ v=2 and t M ¼ ðbrðpvÞ þ dpÞ=2brd.
M. Khouja et al. / Omega 41 (2013) 665–678 677

(i) If ðpcÞ=ðpvB Þ o br=d o 1, we can show that (1) sm 4pdt M , To sell to the bargain-hunters, the retailer choose the high-
which leads to Dm 4 DM , (2) sl o pbrt H , (3) Ld 4 Lg , which value gift card t H such that Uðt H Þ ¼ pvB . Suppose t H 4 T, then
leads to Dl oDL , and (4) 2sh ðxÞv o 2br½ðbrðpvÞ þ dpÞ=2brd Uðt H Þ ¼ U 2 ðt H Þ ¼ dT þ dT ðt H TÞ ¼ pvB , i.e., t H ¼ ððpvB ÞðddT Þ
t L ðxÞ. Therefore, (12) is negative and qnd o qng . The optimal TÞ=dT . Therefore, for t H 4T to hold, ðpvB ðddT ÞTÞ=dT 4T must
profits exhibit the same relationship.
hold resulting in the condition T oðpvB Þ=d. Suppose t M r T, then
(ii) If 1 o br=d op=ðpvB Þ, we can show that (1) sm o pdt M , which
t M will be the same as the t M in Section 4, i.e., t M ¼ minfðbrðp
leads to Dm oDM , (2) sl 4 pbrt H , (3) Ld o Lg , which leads to
Dl 4 DL , and (4) 2sh ðxÞv 42br½ðbrðpvÞ þ dpÞ=2brdt L ðxÞ. vÞ þ drÞ=2brd,ðpv^ g Þ=dg. For t M r T to hold, ðpv^ g Þ=d r T is suffi-
Therefore, (12) is positive and qnd 4qng . The optimal profits exhibit cient. Since v^ g results from the game equilibrium and v o v^ g ,
the same relationship. ðpvÞ=d r T is sufficient for t M rT to hold. Combining the above
two conditions gives
Case2: v=2 o v^ d and ðbrðpvÞ þ dpÞ=2brd 4 ðpv^ g Þ=d. Therefore, pv pvB
rT o
sm ¼ v^ d and t M ¼ ðpv^ g Þ=d. d d
as a sufficient condition for t M r T o t H to hold. &

(i) If ðpcÞ=ðpvB Þ o br=d o1, we can show that (1) sm ¼ pdt M , Proof of Proposition 3. The proof is similar to the proof of
which leads to Dm ¼ DM , (2) sl o pbrt H , (3) Ld 4 Lg , which Propositions 1 and 2. &
leads to Dl oDL , and (4) 2sh ðxÞv o 2br½ðbrðpvÞ þ dpÞ=2brd
t L ðxÞ. Therefore, (12) is negative and qnd o qng . The optimal References
profits exhibit the same relationship.
(ii) If 1 o br=d o p=ðpvB Þ, we can show that (1) sm ¼ pdt M , [1] Accenture. ‘All I Want for Christmas is a Gift Card,’ Say consumers in
which leads to Dm ¼ DM , (2) sl 4 pbrt H , (3) Ld o Lg , which Accenture survey. Last accessed 6/6/2012 from /http://newsroom.accent
leads to Dl 4 DL , and (4) 2sh ðxÞv 4 2br½ðbrðpvÞ þ dpÞ= ure.com/article_display.cfm?article_id=4279S; 2005.
[2] Ali A, Jolson MA, Darmon RY. A model for optimizing the refund value in
2brdt L ðxÞ. Therefore, (12) is positive and qnd 4qng . The opti-
rebate promotions. Journal of Business Research 1994;29(3):239–245.
mal profits exhibit the same relationship. [3] Anderson ET, Song I. Coordinating price reductions and coupon events.
Journal of Marketing Research 2004;41:411–422.
[4] Arcelus FJ, Kumar S, Srinivasan G. Risk tolerance and a retailer’s pricing and
Case3: v=2 4 v^ d and ðbrðpvÞ þ dpÞ=2brd 4ðpv^ g Þ=d. Therefore, ordering policies within a newsvendor framework. Omega 2012;40(2):
sm ¼ v=2 and t M ¼ ðpv^ g Þ=d. In this case, only ðpcÞ=ðpvB Þ o br= 188–198.
[5] Aviv Y, Pazgal A. Optimal pricing of seasonal products in the presence of
d o 1 is possible. We can show that (1) sm 4 pdtM , which leads to forward-looking consumers. Manufacturing & Service Operations Manage-
Dm 4DM , (2) sl opbrt H , (3) Ld 4 Lg , which leads to Dl o DL , and ment 2008;10:339–359.
[6] Bitran GR, Mondschein SV. Periodic pricing of seasonal products in retailing.
(4) 2sh ðxÞv o 2br½ðbrðpvÞ þ dpÞ=2brdt L ðxÞ. Therefore, (12) is Management Science 1997;43(1):64–79.
negative and qnd oqng . The optimal profits exhibit the same relation- [7] Cachon GP, Kök AG. Implementation of the newsvendor model with clearance
pricing: how to (and how not to) estimate a salvage value. Manufacturing &
ship.
Service Operations Management 2007;9(3):276–290.
Case4: v=2 o v^ d and ðbrðpvÞ þ dpÞ=2brd o ðpv^ g Þ=d. Therefore, [8] Cachon GP, Swinney R. Purchasing, pricing, and quick response in the
sm ¼ v^ d and t M ¼ ðbrðpvÞ þ dpÞ=2brd. In this case, only 1 o br= presence of strategic consumers. Management Science 2009;55(3):497–511.
[10] Dhar SK, Morisson DG, Raju JS. The effect of package coupons on brand
d o p=ðpvB Þ is possible. We can show that (1) sm o pdtM , which choice: an epilogue on profits. Marketing Science 1996;(15):192–203.
leads to Dm oDM , (2) sl 4 pbrt H , (3) Ld o Lg , which leads to [11] Folkes V, Wheat RD. Consumers’ price perceptions of promoted products.
Journal of Retailing 1995;71(3):317–328.
Dl 4 DL , and (4) 2sh ðxÞv 42br½ðbrðpvÞ þ dpÞ=2brdt L ðxÞ. There- [12] Frederick S, Loewenstein G, O’Donoghue T. Time discounting and time
fore, (12)is positive and qnd 4 qng . The optimal profits exhibit the same preference: a critical reviewJournal of Economic Literature 2002;40:351–401.
[13] Gaur V, Park Y. Asymmetric consumer learning and inventory competition.
relationship. Management Science 2007;53(2):227–240.
In summary, we have [14] Gilpatric SM. Slippage in rebate programs and present-biased preferences.
Marketing Science 2009;28:229–238.
[15] Gupta D, Hill AV, Bouzdine-Chameeva T. Pricing model for clearing end-of-
(i) If ðpcÞ=ðpvB Þ o br=d o1, the retailer orders more in the season retail inventory. European Journal of Operational Research 2006;170(2):
first period and obtains higher expected profit under the gift 518–540.
card strategy than the discount strategy, i.e., qng Zqnd and [16] Horne DR. Unredeemed gift cards and the problem of not providing
customers with value. Journal of Consumer Marketing 2007;24(4):192–193.
png ðqng , v^ g Þ Z pnd ðqnd , v^ d Þ. [17] Horne DR. Gift cards: disclosure one step removed. Journal of Consumer
(ii) If 1o br=d op=ðpvB Þ, the retailer orders more in the first Affairs 2007;41(2):341–350.
period and obtains higher expected profit under the discount [18] Jerath K, Netessine S, Veeraraghavan SK. Revenue management with strategic
customers: last-minute selling and opaque selling. Management Science
strategy than the gift card strategy, i.e., qnd Z qng and pnd
2010;56(3):430–448.
ðqnd , v^ d Þ Z png ðqng , v^ g Þ. & [19] Khouja M, Pan J, Ratchford BT, Zhou J. Analysis of free gift cards program
effectiveness. Journal of Retailing 2011;87(4):444–461.
[20] Khouja M, Zhou J. The effect of delayed incentives on supply chain profits and
Proof of Corollary 1. Substituting for b in br=d o 1 in Proposition
consumer surplus. Production and Operations Management 2010;19:
1 and simplifying gives mð1ZÞy 4 mð1ZÞ. If Z o1, then 1Z 4 172–197.
0. Dividing mð1ZÞy 4 mð1ZÞ by mð1ZÞ gives y 4 1 which is [21] Lai G, Debo LG, Sycara K. Buy now and match later: the impact of posterior
satisfied if y Z0. & price matching on profit with strategic consumers. Manufacturing & Service
Operations Management 2010;12(1):33–55.
[22] Levin Y, McGill J, Nediak M. Dynamic pricing in the presence of strategic
Proof of Lemma 2. Based on the utility function in (7), the consumers and oligopolistic competition. Management Science 2009;55:32–46.
retailer’s expected net revenue in Eq. (4) becomes [23] Loewenstein G, Prelec D. Anomalies in intertemporal choice: evidence and an
interpretation. Quarterly Journal of Economics 1992;57:573–598.
[24] Netessine S, Tang C. Operations management models with consumer-driven
demand. In: Hiller, Frederick S, Series editor. Springer International Series in
8 Operations Research & Management Science; 2009.
>
> ðpbrtÞminðaDGðpUðtÞÞ,Ig Þ if pv rUðtÞ rpv^ g ,
< [25] Offenberg JP. Markets: gift cardsJournal of Economic Perspectives 2007;21(2):

NRðt,I gÞ ¼ ðpbrtÞminðaDGðv^ g Þ,Ig Þ if pv^ g oUðtÞo pvB , ð14Þ 227–238.
>
> [27] Shu SB, Gneezy A. Procrastination of enjoyable experiences. Journal of
: ðpbrtÞIg if UðtÞ ZpvB :
Marketing Research 2010;47:933–944.
678 M. Khouja et al. / Omega 41 (2013) 665–678

[28] Su X. Intertemporal pricing with strategic customer behavior. Management [31] Yang S, Shi CV, Zhao X. Optimal ordering and pricing decisions for a target
Science 2007;53(5):726–741. oriented newsvendor. Omega 2011;39(1):110–115.
[29] Thaler R. Some empirical evidence on dynamic inconsistency. Economics [32] Zhao L, Tian P, Li X. Dynamic pricing in the presence of consumer inertia.
Letters 1981;8:201–207. Omega 2011;40(2):137–148.
[30] Thomas LM, Dillenbeck MS. Best used by expiration date. Marketing Manage- [33] Zhao W, Zheng YS. Optimal dynamic pricing for perishable assets with
ment 2004;13(1):53–55. nonhomogeneous demand. Management Science 2000;46(3):375–388.

You might also like