Hindawi Publishing Corporation

Mathematical Problems in Engineering

Volume 2016, Article ID 2017253, 11 pages
http://dx.doi.org/10.1155/2016/2017253

Research Article
The Distribution-Free Newsboy Problem with
Multiple Discounts and Upgrades

Ilkyeong Moon,1 Dong Kyoon Yoo,2 and Subrata Saha1

1
Department of Industrial Engineering, Seoul National University, Seoul 08826, Republic of Korea
2
Quality Management Department, LG Electronics, Seoul 07336, Republic of Korea

Correspondence should be addressed to Subrata Saha; [email protected]

Received 28 March 2016; Accepted 26 July 2016

Academic Editor: Paolo Crippa

Copyright © 2016 Ilkyeong Moon et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Most papers on the newsboy problem assume that excess inventory is either sold after discount or discarded. In the real world,
overstocks are handled with multiple discounts, upgrades, or a combination of these measures. For example, a seller may offer a
series of progressively increasing discounts for units that remain on the shelf, or the seller may use incrementally applied innovations
aimed at stimulating greater product sophistication. Moreover, the normal distribution does not provide better protection than
other distributions with the same mean and variance. In this paper, we find the differences between normal distribution approaches
and distribution-free approaches in four scenarios with mean and variance of demand as the only available data to decision-makers.
First, we solve the newsboy problem by considering multiple discounts. Second, we formulate and solve the newsboy problem by
considering multiple upgrades. Third, we formulate and solve a mixed newsboy problem characterized with multiple discounts
and upgrades. Finally, we extend the model to solve a multiproduct newsboy problem with a storage or a budget constraint and
develop an algorithm to find the solutions of the models. Concavity of the models is proved analytically. Extensive computational
experiments are presented to verify the robustness of the distribution-free approach. The results show that the distribution-free
approach is robust.

1. Introduction and developed the min-max distribution-free approach as a

closed-form expression for deriving the optimal order quan-
In general selling situations, the information on demand tity.
distribution is limited. Often, sellers are only provided with Gallego and Moon [2] simplified the proof of Scarf ’s [1]
an educated guess of the mean and the variance. Then, they ordering rule for the newsboy problem. Moon and Gallego [3]
will use the normal distribution of the demand to determine applied the approach to several inventory models including
the proper inventory levels. However, the normal distribution a periodic review model. Moon and Choi [4] solved the
for demand does not provide the best protection for the distribution-free continuous-review inventory model with
occurrence of other distributions with the same mean and a service-level constraint. Moon and Choi [5] provided
the same variance. In a classic paper, Scarf [1] addressed the the make-to-order and make-in-advance models using the
newsboy problem in which the mean 𝜇 and the variance 𝜎2 distribution-free procedure. Moon and Choi [6] improved
of demand were given, whereas assumptions on the distri- the continuous-review inventory model by simultaneously
bution of demand were not available. Taking a conservative optimizing the order quantity and reorder point. Ouyang and
approach, Scarf modeled the problem by finding the order Wu [7] developed models for reduction of the continuous-
quantity that maximizes the expected profit and compared review inventory system. They also applied the distribution-
the result with the worst possible distribution of demand as free approach to each model and proved its robustness.
characterized with the mean 𝜇 and the variance 𝜎2 . Through a Hariga and Ben-Daya [8] developed optimal reduction pro-
beautiful, but lengthy, mathematical argument, Scarf showed cedures in the procurement lead time duration and applied
that the worst demand distribution is positive at two points the distribution-free approach to several stochastic inventory
2 Mathematical Problems in Engineering

models such as the continuous- and periodic-review models the traditional and distribution-free approaches and consider
with a mixture of backorders and lost sales and the base stock. the cost of improving the quality of the product. Nonprice
Ouyang and Chang [9] applied a minimax distribution-free competition is found in many industries, and, as a key
procedure for mixed inventory models with variable lead nonprice competitive feature in the majority of industries,
time and fuzzy lost sales. Tajbakhsh [10] derived closed- improved product quality has received intensive attention,
form expressions for the model of Moon and Choi [4]. especially in terms of investment [33]. Quality upgrading
Numerous models have been studied using the distribution- improves firm’s ability to serve more consumers because
free approach [11–16]. For more details, one can refer to an quality products appeal to wealthier consumers. We refer the
extensive review of the literature provided by Qin et al. [17]. reader to Garvin [34] for an excellent summary of quality
In this study, we verify the robustness of the distribution-free characterization. Also, since 1970, Toyota Motor Co. Ltd.
approach where multiple discounts or upgrades are used to began cooperating with its supplier to improve product qual-
sell products and study the impact of the distribution-free ity [35]. In an increasingly competitive global environment,
approach under a storage or a budget constraint. there is growing evidence that investment in quality improve-
The classical newsboy problem is designed to find the ments, such as applying, renewing, updating, accumulating,
product order quantity that maximizes the expected profit and redeeming, enhances customer satisfaction and builds
in a single-period, probabilistic demand framework. Hadley long-term relationships between sellers and customers [36].
and Whitin [18] addressed the newsboy model under the However, most of the studies on quality improvement fall
assumption that an excess inventory from an order quantity under the supply chain framework [37, 38]. Due to these
larger than the realized demand is either sold with a single extensions we make, our model is applicable not only to the
discount or discarded. Following the work, various versions apparel industry, but also to the computer, telecommunica-
of pricing and discount strategies have been discussed to tion machine, college textbooks, and video game industries
explore the characteristics of the newsboy problem. Readers among others. In each case, we are not able to estimate the
can refer to some major research performances by Carrizosa exact price because an upgrade cost may be incurred due to
et al. [19], Li et al. [20], and Ye and Sun [21]. Recently, Petruzzi growing technology. However, one must note that upgrade
and Dada [22] addressed a comprehensive review and mean- cost does not necessarily translate into increased demand.
ingful extensions for the pricing decision in the newsboy Therefore, we keep the price fixed and formulate the problem
problem. However, multiple progressive discounts in a news- with multiple upgrades. To the best of our knowledge, this
boy framework have received only minor attention despite type of quality improvement strategy has received little atten-
their common appearances in apparel industry where dis- tion in studies of the newsboy problem. Moreover, we con-
counts get steeper as the season draws to an end [23]. Khouja sider the effect of both multiple discounts and upgrades on
[24] developed a model in which multiple discounts are profitability.
used progressively to sell excess inventory with normally dis- In this study, we extend the newsboy problem to consider
tributed demand. Later, Khouja and Mehrez [25] considered the situation in which the seller applies progressive multiple
an extension in which a space or budget constraint is placed discounts, upgrades, or their combination, to enhance prod-
on a group of items. Khouja [26] extended the newsboy uct flow and profit. We elongate the newsboy problem by con-
problem by considering price-dependent demand and mul- sidering four different scenarios and compare the effects of
tiple discounts to sell excess inventory. Such extensions have the distribution-free approach with the traditional distribu-
been thoroughly studied for deterministic economic-order tion approach in single- and multiple-product environments.
quantity models. In this paper, we apply the distribution-free We also examine the effect of a budget or a storage constraint.
approach to the basic model of Khouja [24] and consider the The remainder of the paper is organized as follows. Section 2
situation in which more than one type of stock-keeping unit analyzes the behavior of newsboy problems for demand dis-
is stocked for a single-period demand. tribution of a single product in three different scenarios. We
Considerable research has been conducted to analyze the present the theoretical background for obtaining the order
multiproduct newsboy problems with constraints [18, 27–29]. quantities under the worst- and best-case scenarios. Section 3
Among those works, Hadley and Whitin [18] were the first to discusses the newsboy problem with progressive multiple
consider a single constraint to a multiproduct newsboy prob- discounts or upgrades under a budget or a storage constraint.
lem. Nahmias and Schmidt [30] proposed a multiproduct Section 4 presents computational examples to illustrate all the
newsboy problem with stochastic demand subject to linear models. Finally, Section 5 addresses final remarks as well as
and deterministic constraints on space or budget. H.-S. Lau suggestions for future research and extensions of the obtained
and A. H.-L. Lau [31] extended the constrained newsboy results.
problem to general demand distributions. Erlebacher [32]
developed optimal and heuristic solutions for the capacitated 2. Mathematical Models
newsboy problem. Although budget and storage constraints
as well as price discounts are common features in numerous The following notations are used to establish mathematical
problems, few researchers have considered both a constraint models:
and a discount in a newsboy model. In this paper, we 𝑗: the discounted (or upgraded) index.
extend the multiproduct newsboy problem to the situation
in which the newsboy problem is characterized by a budget 𝑛: the number of discounts (or upgrades).
constraint and multiple discounts to stimulate demand under 𝑃: the unit selling price.
Mathematical Problems in Engineering 3

𝐶: the unit cost. zero. Here, the profit function (𝜋𝑑𝑁(𝑄)) can be represented as
follows:
𝑆: the salvage value.
𝜋𝑑𝑁 (𝑄)
𝑋: the quantity demanded, a random variable (𝑋 > Q
0). {
{ 𝛼0 𝑡0 𝑄 if ≤𝑋
{
{ 𝑉0
{
{
{
{ 𝑄 𝑄
{
{ 𝛼0 𝑡0 𝑋 + 𝛼1 (𝑄 − 𝑉0 𝑋) ≤𝑋<
𝑓(𝑋): the probability density function of 𝑋. {
{ if
{
{ 𝑉1 𝑉0
{
{ (1)
{ 𝑄 𝑄
𝐹(𝑋): the cumulative distribution function of 𝑋. = {𝛼 𝑡 𝑋 + 𝛼 𝑡 𝑋 + 𝛼 (𝑄 − 𝑉 𝑋) if ≤𝑋<
{
{ 0 0 1 1 2 1
𝑉2 𝑉1
{
{
{
{ . ..
𝑄: the order quantity. {
{ .. .
{
{
{
{ 𝑛−1
{
{ 𝑄
{
{𝑋 ∑ 𝛼𝑗 𝑡𝑗 + 𝛼𝑛 (𝑄 − 𝑉𝑛−1 𝑋) if 0 ≤ 𝑋 < ,
𝜇: the expected demand. 𝑉𝑛−1
{ 𝑗=0
𝑗
where 𝛼𝑗 = 𝑃𝑗 − 𝐶, 𝑉𝑗 = 1 + ∑𝑘=1 𝑡𝑘 , 𝑡0 = 1, and 𝑉0 = 1. The
𝜎: the standard deviation of demand.
expected value of 𝜋𝑑𝑁(𝑄) is
𝑑𝑗 : the 𝑗th discounted rate (0 ≤ 𝑑𝑗 ≤ 1), 0 = 𝑑0 <
𝑑1 < 𝑑2 < ⋅ ⋅ ⋅ < 𝑑𝑛 . Π𝑁
𝑑 (𝑄)
∞
=∫ 𝛼0 𝑡0 𝑄𝑓 (𝑋) 𝑑𝑋
𝑃𝑗 : the 𝑗th discounted selling price, 𝑃𝑗 = (1 − 𝑑𝑗 )𝑃. 𝑄/𝑉0

𝑄/𝑉0
𝑢𝑗 : the 𝑗th upgraded rate (0 ≤ 𝑢𝑗 ≤ 1), 0 = 𝑢0 < 𝑢1 < +∫ [𝛼0 𝑡0 𝑋 + 𝛼1 (𝑄 − 𝑉0 𝑋)] 𝑓 (𝑋) 𝑑𝑋 + ⋅ ⋅ ⋅ (2)
𝑢2 < ⋅ ⋅ ⋅ < 𝑢𝑛 . 𝑄/𝑉1

𝑄/𝑉𝑛−1 𝑛−1
𝐶𝑗 : the 𝑗th upgraded cost, 𝐶𝑗 = (1 + 𝑢𝑗 )𝐶. +∫ [𝑋 ∑ 𝛼𝑗 𝑡𝑗 + 𝛼𝑛 (𝑄 − 𝑉𝑛−1 𝑋)] 𝑓 (𝑋) 𝑑𝑋.
0
[ 𝑗=0 ]
𝑡𝑗 : the fraction of the realized demand at the original
price (or cost) that can additionally be sold by The necessary condition for 𝑄 to be optimal is given by
discounting (or upgrading) the product to price 𝑃𝑗 (or
cost 𝐶𝑗 ) (0 < 𝑡𝑗 < 1). 𝑑Π𝑁
𝑑 (𝑄)
𝑛−1
𝑄
= 𝛼0 − ∑ (𝑃𝑗 − 𝑃𝑗+1 ) 𝐹 ( ) = 0. (3)
𝑑𝑄 𝑗=0 𝑉𝑗
2.1. The Newsboy Problem with Progressive Multiple Discounts.
We focus on the special case where the additional quantity The second-order derivative of Π𝑁 𝑑 (𝑄) is given by
that can be sold at the 𝑗th discount price is directly propor- 𝑑2 Π𝑁 (𝑄)/𝑑𝑄 = ∑ 𝑛−1
(1/𝑉 )(𝑃 − 𝑃
𝑑 2 𝑗=0 𝑗 𝑗+1 𝑗 )𝑓(𝑄/𝑉𝑗 ). Because
tional to the quantity realized at the regular price. In other 2 𝑁
words, if the product has high demand at the regular price, 𝑓(𝑋) ≥ 0 and 𝑃𝑗+1 − 𝑃𝑗 ≤ 0, 𝑑 Π𝑑 (𝑄)/𝑑𝑄2 ≤ 0 for 𝑄 ≥ 0
then discounting the product should result in proportionally and hence Π𝑁 𝑑 (𝑄) is concave.
additional demand and vice versa. If we refer to the time of The normal distribution does not provide the best pro-
the 𝑗th discount as the 𝑗th discount period, then statistically tection against the occurrence of other distributions with
the demand in each discount period is perfectly and positively the same mean and variance. Therefore, we formulate and
correlated with the demand in the first (nondiscount) period. solve the distribution-free newsboy problem with multiple
We assume that the salvage value of the product is the 𝑛th discounts. In this case, the profit function (𝜋𝑑𝑊(𝑄)) can be
discount price, and the penalty for not satisfying demand is obtained as

𝑄
{
{ 𝛼0 𝑡0 𝑄 if ≤𝑋
{
{ 𝑉0
{
{
{
{ 𝑄 𝑄
{
{ 𝛼0 𝑡0 𝑄 − 𝑑1 𝑃 (𝑄 − 𝑉0 𝑋) if ≤𝑋<
{
{ 𝑉1 𝑉0
{
{
{
𝜋𝑑𝑊 (𝑄) = {𝛼 𝑡 𝑄 − 𝑑 𝑃 (𝑄 − 𝑉 𝑋) − (𝑑 − 𝑑 ) 𝑃 (𝑄 − 𝑉 𝑋) 𝑄 𝑄 (4)
{ 0 0 1 0 2 1 1 if ≤𝑋<
{
{ 𝑉2 𝑉1
{
{ . ..
{
{ ..
{
{ .
{
{ 𝑛
𝑄
{
{ 𝛼 𝑡 𝑄 − 𝑃 ∑ (𝑑𝑗 − 𝑑𝑗−1 ) (𝑄 − 𝑉𝑗−1 𝑋) if 0 ≤ .
{ 00
𝑉𝑛−1
{ 𝑗=1
4 Mathematical Problems in Engineering

The expected value of 𝜋𝑑𝑊(𝑄) is given by The necessary condition for 𝑄 to be optimal is

𝑛
Π𝑊
+ 𝑑𝐶𝑑𝑊 (𝑄)
𝑑 (𝑄) = 𝛼0 𝑡0 𝑄 − 𝑃 ∑ (𝑑𝑗 − 𝑑𝑗−1 ) 𝐸 (𝑄 − 𝑉𝑗−1 𝑋)
𝑗=1 𝑑𝑄

= 𝛼0 𝑡0 𝑄 1
(5) = 𝑃𝑑𝑛 − 𝛼0 𝑡0
2
𝑛 + (10)
𝑄
− 𝑃∑ (𝑑𝑗 − 𝑑𝑗−1 ) 𝑉𝑗−1 𝐸 ( − 𝑋) . 1 𝑛 (𝑄/𝑉𝑗−1 − 𝜇)
𝑗=1 𝑉𝑗−1 + ∑ (𝑑𝑗 − 𝑑𝑗−1 )
2 𝑗=1 2 1/2
[𝜎2 + (𝑄/𝑉𝑗−1 − 𝜇) ]
Observing that (𝑄 − 𝑋)+ = (𝑄 − 𝑋) + (𝑋 − 𝑄)+ , we can write
the expected profit as = 0.

𝑛 On simplification,
Π𝑊
𝑑 (𝑄) = 𝜇𝑃 ∑ (𝑑𝑗 − 𝑑𝑗−1 ) 𝑉𝑗−1
𝑗=1
2𝐶
2− − 𝑑𝑛
𝑛 𝑃
− [𝑃 ∑ (𝑑𝑗 − 𝑑𝑗−1 ) − 𝛼0 𝑡0 ] 𝑄 (6) (11)
𝑛 (𝑄/𝑉𝑗−1 − 𝜇)
[ 𝑗=1 ] = ∑ (𝑑𝑗 − 𝑑𝑗−1 ) .
2 1/2
𝑛
𝑄
+ 𝑗=1 [𝜎2 + (𝑄/𝑉𝑗−1 − 𝜇) ]
− 𝑃 ∑ (𝑑𝑗 − 𝑑𝑗−1 ) 𝑉𝑗−1 𝐸 (𝑋 − ) .
𝑗=1 𝑉𝑗−1
The second-order derivative of 𝐶𝑑𝑊(𝑄) is given by 𝑑2 𝐶𝑑𝑊(𝑄)/
Because maximizing Π𝑊 𝑑𝑄2 = 𝑃 ∑𝑛𝑗=1 ((𝑑𝑗 − 𝑑𝑗−1 )𝜎2 /𝑉𝑗−1 [𝜎2 + (𝑄/𝑉𝑗−1 − 𝜇)2 ]3/2 ).
𝑑 (𝑄) is equivalent to minimizing
𝐶𝑑𝑊(𝑄), we concentrate on minimizing 𝐶𝑑𝑊(𝑄), where Because 𝑑𝑗 − 𝑑𝑗−1 ≥ 0, 𝑑2 𝐶𝑑𝑊(𝑄)/𝑑𝑄2 ≥ 0 for 𝑄 ≥ 0 and
hence 𝐶𝑑𝑊(𝑄) is convex. Therefore, Π𝑊 𝑑 (𝑄) is concave.
𝑛
𝐶𝑑𝑊 (𝑄) = [𝑃∑ (𝑑𝑗 − 𝑑𝑗−1 ) − 𝛼0 𝑡0 ] 𝑄 2.2. The Newsboy Problem with Progressive Multiple Upgrades.
[ 𝑗=1 ] (7)
We focus on the special case where the additional quality
+
upgraded is directly proportional to the realized demand for
𝑛
𝑄 the initial product. In other words, if the product has high
+ 𝑃∑ (𝑑𝑗 − 𝑑𝑗−1 ) 𝑉𝑗−1 𝐸 (𝑋 − ) .
𝑗=1 𝑉𝑗−1 demand at the regular price, then upgrading the product
should result in proportionally additional demand and vice
versa. If we refer to the time of the 𝑗th upgrade as the 𝑗th
Here, ∑𝑛𝑗=1 (𝑑𝑗 − 𝑑𝑗−1 ) = 𝑑𝑛 . Because the distribution 𝐹 of 𝑋 upgrade period, then statistically the demand in each upgrade
is unknown, we want to minimize 𝐶𝑑𝑊(𝑄) against the worst period is perfectly and positively correlated with the demand
possible distribution. For this, we need to use the following in the first period. We assume that the salvage cost of the
lemma as in Gallego and Moon [2]. product is the 𝑛th upgrade cost. Here, we maximize the
expected profit to compare the normal with the worst-case
Lemma 1. For any 𝐹 ∈ F, demand distribution. The profit function (𝜋𝑢𝑁(𝑄)) in this case
can be written as
1 2 1/2
𝐸 [𝑥 − 𝑄]+ = {[𝜎2 + (𝑄 − 𝜇) ] − (𝑄 − 𝜇)} . (8)
2 𝜋𝑢𝑁 (𝑄)

Moreover, the upper bound is tight. 𝑄

{
{ 𝛼0 𝑡0 𝑄 if ≤𝑋
{
{ 𝑉0
{
{
Using Lemma 1, our problem now reduces to minimizing {
{
{
{ 𝑄 𝑄
the upper bound {
{ 𝛼0 𝑡0 𝑋 + 𝛼1 (𝑄 − 𝑉0 𝑋) if ≤𝑋<
{
{ 𝑉1 𝑉0
{
{
{
{ (12)
𝑃 𝑛 = {𝛼0 𝑡0 𝑋 + 𝛼1 𝑡1 𝑋 + 𝛼2 (𝑄 − 𝑉1 𝑋) 𝑄 𝑄
𝐶𝑑𝑊 (𝑄) = (𝑃𝑑𝑛 − 𝛼0 𝑡0 ) 𝑄 + ∑ (𝑑𝑗 − 𝑑𝑗−1 ) { if ≤𝑋<
{
{ 𝑉2 𝑉1
2 𝑗=1 {
{ .. ..
{
{
{
{ . .
1/2
(9) {
{
{ 2
} {
{
𝑄 𝑄 {
{
𝑛−1
𝑄
⋅ 𝑉𝑗−1 {[𝜎2 + ( − 𝜇) ] −( − 𝜇)} . {
{𝑋 ∑ 𝛼𝑗 𝑡𝑗 + 𝛼𝑛 (𝑄 − 𝑉𝑛−1 𝑋) if 0 ≤ 𝑋 < ,
𝑉𝑗−1 𝑉𝑗−1 𝑉𝑛−1
{ } { 𝑗=0
Mathematical Problems in Engineering 5

𝑗
where 𝛼𝑗 = 𝑃 − 𝐶𝑗 , 𝑉𝑗 = 1 + ∑𝑘=1 𝑡𝑘 , 𝑡0 = 1, and 𝑉0 = 1. The The necessary condition for 𝑄 to be optimal is given by
expected value of 𝜋𝑢𝑁(𝑄) is
𝑑Π𝑁
𝑢 (𝑄)
𝑛−1
𝑄
∞ = 𝛼0 − ∑ (𝐶𝑗+1 − 𝐶𝑗 ) 𝐹 ( ) = 0. (14)
Π𝑁 𝑑𝑄 𝑉
𝑢 (𝑄) = ∫ 𝛼0 𝑡0 𝑄𝑓 (𝑋) 𝑑𝑋 𝑗=0 𝑗
𝑄/𝑉0

𝑄/𝑉0
The second-order derivative of Π𝑁 2 𝑁
𝑢 (𝑄) is given by 𝑑 Π𝑢 (𝑄)/
𝑛−1
+∫ [𝛼0 𝑡0 𝑋 + 𝛼1 (𝑄 − 𝑉0 𝑋)] 𝑓 (𝑋) 𝑑𝑋 𝑑𝑄2 = − ∑𝑗=0 (1/𝑉𝑗 )(𝐶𝑗+1 − 𝐶𝑗 )𝑓(𝑄/𝑉𝑗 ). Because 𝑓(𝑋) ≥ 0
𝑄/𝑉1
(13) and 𝐶𝑗+1 − 𝐶𝑗 ≥ 0, 𝑑2 Π𝑁𝑢 (𝑄)/𝑑𝑄2 ≤ 0 for 𝑄 ≥ 0 and hence
𝑄/𝑉𝑛−1 𝑛−1 Π𝑁𝑢 (𝑄) is concave.
+ ⋅⋅⋅ + ∫ [𝑋 ∑ 𝛼𝑗 𝑡𝑗 + 𝛼𝑛 (𝑄 − 𝑉𝑛−1 𝑋)] Similar to the previous section, we formulate and
0
[ 𝑗=0 ] solve the newsboy problem with multiple upgrades by the
distribution-free approach. In this case, the profit function
⋅ 𝑓 (𝑋) 𝑑𝑋. (𝜋𝑢𝑊(𝑄)) can be represented as follows:

𝑄
{
{ 𝛼0 𝑡0 𝑄 if ≤𝑋
{
{ 𝑉0
{
{
{
{
{
{ 𝑄 𝑄
{
{ 𝛼0 𝑡0 𝑄 − 𝑢1 𝐶 (𝑄 − 𝑉0 𝑋) if ≤𝑋<
{
{ 𝑉1 𝑉0
{
{
{
{
𝜋𝑢𝑊 (𝑄) = {𝛼 𝑡 𝑄 − 𝑢 𝐶 (𝑄 − 𝑉 𝑋) − (𝑢 − 𝑢 ) 𝐶 (𝑄 − 𝑉 𝑋) 𝑄 𝑄 (15)
{ 0 0 1 0 2 1 1 if ≤𝑋<
{
{ 𝑉2 𝑉1
{
{ . ..
{
{ ..
{
{ .
{
{
{
{ 𝑛
{
{ 𝑄
{
{𝛼0 𝑡0 𝑄 − 𝐶∑ (𝑢𝑗 − 𝑢𝑗−1 ) (𝑄 − 𝑉𝑗−1 𝑋) if 0 ≤ .
{ 𝑗=1 𝑉𝑛−1

The expected value of 𝜋𝑢𝑊(𝑄) is given by Because maximizing Π𝑊 𝑢 (𝑄) is equivalent to minimizing
𝐶𝑢𝑊(𝑄), we concentrate on minimizing 𝐶𝑢𝑊(𝑄), where

𝑛
+
Π𝑊
𝑢 (𝑄) = 𝛼0 𝑡0 𝑄 − 𝐶 ∑ (𝑢𝑗 − 𝑢𝑗−1 ) 𝐸 (𝑄 − 𝑉𝑗−1 𝑋)
𝑗=1 𝑛
𝐶𝑢𝑊 (𝑄) = [𝐶∑ (𝑢𝑗 − 𝑢𝑗−1 ) − 𝛼0 𝑡0 ] 𝑄
= 𝛼0 𝑡0 𝑄 (16) [ 𝑗=1 ] (18)
𝑛 + 𝑛 +
𝑄 𝑄
− 𝐶∑ (𝑢𝑗 − 𝑢𝑗−1 ) 𝑉𝑗−1 𝐸 ( − 𝑋) . + 𝐶∑ (𝑢𝑗 − 𝑢𝑗−1 ) 𝑉𝑗−1 𝐸 (𝑋 − ) .
𝑗=1 𝑉𝑗−1 𝑗=1 𝑉𝑗−1

Observing that (𝑄 − 𝑋)+ = (𝑄 − 𝑋) + (𝑋 − 𝑄)+ , we can write Here, ∑𝑛𝑗=1 (𝑢𝑗 − 𝑢𝑗−1 ) = 𝑢𝑛 . Because the distribution 𝐹
the expected profit as
of 𝑋 is unknown, we want to minimize 𝐶𝑢𝑊(𝑄) against the
worst possible distribution. Using Lemma 1, our problem now
reduces to minimizing the upper bound
𝑛
Π𝑊
𝑢 (𝑄) = 𝜇𝐶 ∑ (𝑢𝑗 − 𝑢𝑗−1 ) 𝑉𝑗−1
𝑗=1

1 𝑛
𝑛
𝐶𝑢𝑊 (𝑄) = (𝐶𝑢𝑛 − 𝛼0 𝑡0 ) 𝑄 + 𝐶∑ (𝑢𝑗 − 𝑢𝑗−1 )
− [𝐶∑ (𝑢𝑗 − 𝑢𝑗−1 ) − 𝛼0 𝑡0 ] 𝑄 (17) 2 𝑗=1
[ 𝑗=1 ] (19)
2 1/2
𝑛 + { 𝑄 𝑄 }
𝑄 ⋅ 𝑉𝑗−1 {[𝜎2 + ( − 𝜇) ] −( − 𝜇)} .
− 𝐶∑ (𝑢𝑗 − 𝑢𝑗−1 ) 𝑉𝑗−1 𝐸 (𝑋 − ) . 𝑉𝑗−1 𝑉𝑗−1
𝑗=1 𝑉𝑗−1 { }
6 Mathematical Problems in Engineering

The necessary condition for 𝑄 to be optimal is The second-order derivative of 𝐶𝑢𝑊(𝑄) is given by 𝑑2 𝐶𝑢𝑊(𝑄)/
𝑑𝑄2 = 𝐶 ∑𝑛𝑗=1 ((𝑢𝑗 − 𝑢𝑗−1 )𝜎2 /𝑉𝑗−1 [𝜎2 + (𝑄/𝑉𝑗−1 − 𝜇)2 ]3/2 ).
𝑑𝐶𝑢𝑊 (𝑄) Because 𝑢𝑗 − 𝑢𝑗−1 ≥ 0, 𝑑2 𝐶𝑢𝑊(𝑄)/𝑑𝑄2 ≥ 0 for 𝑄 ≥ 0 and
𝑑𝑄 hence 𝐶𝑢𝑊(𝑄) is convex. Therefore, Π𝑊 𝑢 (𝑄) is concave.
1
= 𝐶𝑢𝑛 − 𝛼0 𝑡0 2.3. The Mixed Newsboy Problem with Progressive Multiple
2
(20) Discounts and Upgrades. Multiple discounts lead to higher
1 𝑛 (𝑄/𝑉𝑗−1 − 𝜇) order quantities compared to the one found in the classi-
+ 𝐶∑ (𝑢𝑗 − 𝑢𝑗−1 ) cal newsboy problems because multiple discounts result in
2 𝑗=1 2 1/2
[𝜎2 + (𝑄/𝑉𝑗−1 − 𝜇) ] increased demand at prices that are higher than the salvage
value in the classical newsboy problem. More practically, the
=0 discount policy is not the only way to sell excess products.
Sellers make additional investment in marketing to attract
which gives more customers to stores through progressive discounts.
Bridson et al. [39] suggested that an appropriate mix of hard
2𝑃 attributes (discounts and coupons) and soft attributes (better
− 2 − 𝑢𝑛
𝐶 service, upgrades, and special attention) can affect cus-
tomer satisfaction with stores. Therefore, when considering
𝑛 (𝑄/𝑉𝑗−1 − 𝜇) (21)
the mixed model with progressive multiple discounts and
= ∑ (𝑢𝑗 − 𝑢𝑗−1 ) .
2 1/2 upgrades (i.e., multiple discounts with progressively increas-
𝑗=1 2
[𝜎 + (𝑄/𝑉𝑗−1 − 𝜇) ] ing costs), the profit function of the model is given by

𝑊
𝜋𝑑𝑢 (𝑄)
𝑄
{
{ 𝛼0 𝑡0 𝑄 if ≤𝑋
{
{ 𝑉0
{
{
{
{
{
{ 𝑄 𝑄
{
{ 𝛼0 𝑡0 𝑄 − 𝑑1 𝑃 (𝑄 − 𝑉0 𝑋) if ≤𝑋<
{
{ 𝑉1 𝑉0
{
{
{
{
{
{ 𝑄 𝑄
{
{ 𝛼0 𝑡0 𝑄 − 𝑑1 𝑃 (𝑄 − 𝑉0 𝑋) − 𝑢1 𝐶 (𝑄 − 𝑉1 𝑋) if ≤𝑋<
{
{ 𝑉2 𝑉1
{
{
{
{ (22)
= {𝛼 𝑡 𝑄 − 𝑑 𝑃 (𝑄 − 𝑉 𝑋) − 𝑢 𝐶 (𝑄 − 𝑉 𝑋) − (𝑑 − 𝑑 ) 𝑃 (𝑄 − 𝑉 𝑋) 𝑄 𝑄
{
{ 0 0 1 0 1 1 2 1 2 if ≤𝑋<
{
{ 𝑉3 𝑉2
{
{
{
{ 𝑄 𝑄
{
{
{
{ 𝛼0 𝑡0 𝑄 − 𝑑1 𝑃 (𝑄 − 𝑉0 𝑋) − 𝑢1 𝐶 (𝑄 − 𝑉1 𝑋) − (𝑑2 − 𝑑1 ) 𝑃 (𝑄 − 𝑉2 𝑋) − (𝑢2 − 𝑢1 ) 𝐶 (𝑄 − 𝑉3 𝑋) if ≤𝑋<
{
{ 𝑉4 𝑉3
{
{ .. ..
{
{
{
{ . .
{
{
{
{ 𝑛 𝑛
𝑄
{𝛼 𝑡 𝑄 − 𝑃 ∑ (𝑑 − 𝑑 ) (𝑄 − 𝑉
{
{ 00 𝑗 𝑗−1 2(𝑗−1) 𝑋) − 𝐶 ∑ (𝑢𝑗 − 𝑢𝑗−1 ) (𝑄 − 𝑉2𝑗−1 𝑋) if 0 ≤ 𝑋 < ,
𝑉𝑛−1
{ 𝑗=1 𝑗=1

𝑗 𝑛 +
where 𝛼𝑗 = 𝑃𝑗 − 𝐶𝑗 and 𝑉𝑗 = 1 + ∑𝑘=1 𝑡𝑘 , 𝑡0 = 1, and 𝑉0 = 1. 𝑄
𝑊 − 𝑃 ∑ (𝑑𝑗 − 𝑑𝑗−1 ) 𝑉2(𝑗−1) 𝐸 ( − 𝑋)
The expected value of 𝜋𝑑𝑢 (𝑄) is 𝑗=1 𝑉2(𝑗−1)
𝑛 +
𝑄
− 𝐶∑ (𝑢𝑗 − 𝑢𝑗−1 ) 𝑉2𝑗−1 𝐸 ( − 𝑋) .
Π𝑊
𝑑𝑢 (𝑄) 𝑗=1 𝑉2𝑗−1
𝑛
+
(23)
= 𝛼0 𝑡0 𝑄 − 𝑃∑ (𝑑𝑗 − 𝑑𝑗−1 ) 𝐸 (𝑄 − 𝑉2(𝑗−1) 𝑋) + +
𝑗=1
Observing that (𝑄 − 𝑋) = (𝑄 − 𝑋) + (𝑋 − 𝑄) , we can write
the expected profit as
𝑛
+
− 𝐶∑ (𝑢𝑗 − 𝑢𝑗−1 ) 𝐸 (𝑄 − 𝑉2𝑗−1 𝑋) Π𝑊
𝑑𝑢 (𝑄)
𝑗=1 𝑛
= 𝜇𝑃 ∑ (𝑑𝑗 − 𝑑𝑗−1 ) 𝑉2(𝑗−1)
= 𝛼0 𝑡0 𝑄 𝑗=1
Mathematical Problems in Engineering 7

𝑛 𝑛 𝑊
The necessary condition for 𝑄 to be optimal is 𝑑𝐶𝑑𝑢 (𝑄)/𝑑𝑄 =
+ 𝜇𝐶∑ (𝑢𝑗 − 𝑢𝑗−1 ) 𝑉2𝑗−1 − 𝑃𝑄∑ (𝑑𝑗 − 𝑑𝑗−1 )
𝑗=1 𝑗=1
0, which gives

𝑛 2𝛼0 𝑡0 − 𝑃𝑑𝑛 − 𝐶𝑢𝑛

− 𝐶𝑄∑ (𝑢𝑗 − 𝑢𝑗−1 )
𝑗=1 1 𝑛 (𝑄/𝑉2𝑗−1 − 𝜇)
= 𝐶∑ (𝑢𝑗 − 𝑢𝑗−1 )
𝑛 + 2 𝑗=1 2 1/2
𝑄 [𝜎2 + (𝑄/𝑉2𝑗−1 − 𝜇) ]
− 𝑃 ∑ (𝑑𝑗 − 𝑑𝑗−1 ) 𝑉2(𝑗−1) 𝐸 (𝑋 − ) (27)
𝑗=1 𝑉2(𝑗−1)
1 𝑛 (𝑄/𝑉2(𝑗−1) − 𝜇)
𝑛
𝑄
+ + 𝑃 ∑ (𝑑𝑗 − 𝑑𝑗−1 ) .
2 𝑗=1 2 1/2
− 𝐶∑ (𝑢𝑗 − 𝑢𝑗−1 ) 𝑉2𝑗−1 𝐸 (𝑋 − ) [𝜎2 + (𝑄/𝑉2(𝑗−1) − 𝜇) ]
𝑗=1 𝑉2𝑗−1
𝑊
+ 𝛼0 𝑡0 𝑄. The second-order derivative of 𝐶𝑑𝑢 (𝑄) is given by
(24)
𝑑2 𝐶𝑑𝑢
𝑊
(𝑄) 𝑛
1
Π𝑊 = 𝐶∑ ( ) (𝑢𝑗 − 𝑢𝑗−1 )
In the distribution-free approach, maximizing is 𝑑𝑢 (𝑄) 𝑑𝑄2 𝑉2𝑗−1
𝑊 𝑗=1
equivalent to minimizing 𝐶𝑑𝑢 (𝑄); we concentrate on mini-
𝑊 𝑛
mizing 𝐶𝑑𝑢 (𝑄), where 𝜎2 1
⋅ + 𝑃∑ ( )
𝑊 2 3/2 𝑉2(𝑗−1) (28)
𝐶𝑑𝑢 (𝑄) [𝜎2 + (𝑄/𝑉2𝑗−1 − 𝜇) ] 𝑗=1

𝑛 𝑛
= 𝑃𝑄∑ (𝑑𝑗 − 𝑑𝑗−1 ) + 𝐶𝑄∑ (𝑢𝑗 − 𝑢𝑗−1 ) 𝜎2
⋅ (𝑑𝑗 − 𝑑𝑗−1 ) .
𝑗=1 𝑗=1 2 3/2
[𝜎2 + (𝑄/𝑉2(𝑗−1) − 𝜇) ]
𝑛 +
𝑄
+ 𝑃 ∑ (𝑑𝑗 − 𝑑𝑗−1 ) 𝑉2(𝑗−1) 𝐸 (𝑋 − ) (25) Because 𝑢𝑗 − 𝑢𝑗−1 ≥ 0 and 𝑑𝑗 − 𝑑𝑗−1 ≥ 0, 𝑑2 𝐶𝑑𝑢𝑊
(𝑄)/𝑑𝑄2 ≥ 0
𝑗=1 𝑉2(𝑗−1)
for 𝑄 ≥ 0 and hence 𝐶𝑑𝑢 (𝑄) is convex. Therefore, Π𝑊
𝑊
𝑑𝑢 (𝑄) is
− 𝛼0 𝑡0 𝑄 concave.
The optimality of the solution for the normal distribution
𝑛 +
𝑄 expected-profit model can be verified by the same method
+ 𝐶∑ (𝑢𝑗 − 𝑢𝑗−1 ) 𝑉2𝑗−1 𝐸 (𝑋 − ) .
𝑗=1 𝑉2𝑗−1 mentioned above. Hence, the detailed discussion is omitted.

Because the distribution 𝐹 of 𝑋 is unknown, we want to 3. The Newsboy Problem with

𝑊
minimize 𝐶𝑑𝑢 (𝑄) against the worst possible distribution. Progressive Multiple Discounts with
Using Lemma 1, our problem now reduces to minimizing the
upper bound a Budget or a Storage Constraint
𝑛 𝑛 The purpose of this section is to investigate the effect of a bud-
𝑊
𝐶𝑑𝑢 (𝑄) = 𝑃𝑄∑ (𝑑𝑗 − 𝑑𝑗−1 ) + 𝐶𝑄∑ (𝑢𝑗 − 𝑢𝑗−1 ) get or a storage constraint on the optimal order quantities in a
𝑗=1 𝑗=1 multiproduct newsboy problem in which multiple discounts
or upgrades are used to sell excess inventory. The following
1 𝑛
− 𝛼0 𝑡0 𝑄 + 𝑃∑ (𝑑𝑗 − 𝑑𝑗−1 ) additional notations are used in this section to develop the
2 𝑗=1 newsboy problem applied to multiple products:

2 1/2 𝑖: the product index.

{ 𝑄
⋅ 𝑉2(𝑗−1) {[𝜎2 + ( − 𝜇) ] 𝐼: the number of products.
𝑉2(𝑗−1)
{ 𝑗: the discounted index.
(26)
𝑄 } 1 𝑛 𝑛: the number of discounts offered.
+( − 𝜇)} + 𝐶∑ (𝑢𝑗 − 𝑢𝑗−1 )
𝑉2(𝑗−1) 2 𝑗=1 𝑃𝑖 : the unit selling price offered for product 𝑖.
}
𝐶𝑖 : the unit cost offered for product 𝑖.
2 1/2
{ 𝑄 𝑆𝑖 : the salvage value offered for product 𝑖.
⋅ 𝑉2𝑗−1 {[𝜎2 + ( − 𝜇) ]
𝑉2𝑗−1 𝑑𝑖,𝑗 : the 𝑗th discounted rate for product 𝑖
{
(0 ≤ 𝑑𝑖,𝑗 ≤ 1), 0 = 𝑑𝑖,0 < 𝑑𝑖,1 < 𝑑𝑖,2 < ⋅ ⋅ ⋅ < 𝑑𝑖,𝑛 .
𝑄 }
+( − 𝜇)} . 𝑃𝑖,𝑗 : the 𝑗th discounted selling price for product 𝑖,
𝑉2𝑗−1 𝑃𝑖,𝑗 = (1 − 𝑑𝑖,𝑗 )𝑃𝑖 .
}
8 Mathematical Problems in Engineering

𝜇𝑖 : the expected demand for product 𝑖. where 𝜆 1 (> 0) is the Lagrange multiplier and 𝐶𝑑 (𝑍) is the cost
𝜎𝑖 : the standard deviation of demand for product 𝑖. function under the distribution-free approach. Algorithm 1 is
used to find the optimal 𝑄𝑖 , 𝑖 = 1, 2, . . . , 𝐼.
𝑘𝑖 : the storage space needed per unit of product 𝑖.
𝑄𝑖 : the order quantity for product 𝑖. Algorithm 1.
𝑡𝑖,𝑗 : the fraction of the realized demand of product 𝑖
at the original price that can additionally be sold by Step 1. Find 𝑄𝑖 , 𝑖 = 1, 2, . . . , 𝐼, using an unconstrained
discounting the product to price 𝑃𝑖,𝑗 . solution.
If ∑𝐼𝑗=1 𝑘𝑖 𝑄𝑖 ≤ 𝐾, the solution is optimal.
𝐾: the total storage capacity.
Otherwise, set 𝜆 1𝑈 = Max[(𝑃𝑖,0 − 𝐶𝑖 )/𝑘𝑖 ] and 𝜆 1𝐿 = 0
𝐵: the budget for the period. and go to Step 2.
For the multiproduct problem, the total expected profit
(𝐸𝑑 (𝑍)) under progressive multiple discount is obtained by Step 2. Set a sufficiently small computational accuracy
considering the sum of the expected profits for all products parameter 𝜖 > 0.
as follows: Set 𝜆 1𝑀 = (𝜆 1𝑈 + 𝜆 1𝐿 )/2. Using 𝜆 1𝐿 , solve (30) or (31) for
𝐼
𝑖 = 1, 2, . . . , 𝐼.
𝐸𝑑 (𝑍) = ∑𝐸𝑑 (𝜋𝑖 ) , If no solution exists for product 𝑖, set 𝑄𝑖 = 0.
(29)
𝑗=1
Otherwise, find 𝑄𝑖 and go to Step 3.

where 𝐸𝑑 (𝜋𝑖 ) is the expected profit from the 𝑖th product. Step 3. If ∑𝐼𝑗=1 𝑘𝑖 𝑄𝑖 > 𝐾, set 𝜆 1𝐿 = 𝜆 1𝑀 and go to Step 2.
Under the multiproduct problem, we have the following
If ∑𝐼𝑗=1 𝑘𝑖 𝑄𝑖 < 𝐾, 𝜆 1𝑈 − 𝜆 1𝑀 > 𝜖 are satisfied.
constraint optimization problems.
Set 𝜆 𝑈 = 𝜆 𝑀.
Problem 1. Maximizing the expected profit under a storage Go to Step 2.
constraint, Otherwise, the solution is optimal.
Max 𝐸𝑑 (𝑍) Similarly, we can formulate the newsboy problem with
progressive multiple upgrades with a budget or storage
𝐼 (P1)
Subject to ∑𝑘𝑖 𝑄𝑖 ≤ 𝐾, 𝑄𝑖 ≥ 0, 𝑖 = 1, 2, . . . , 𝐼. constraint. The necessary conditions for 𝑄𝑖 (𝑖 = 1, 2, . . . , 𝐼)
𝑗=1 to be optimal under progressive multiple upgrades are as
follows.
Problem 2. Maximizing the expected profit under a budget In the traditional approach,
constraint,
Max 𝐸𝑑 (𝑍) 𝜕𝐸𝑢 (𝑍) 𝑛−1
𝑄
= 𝛼0 − ∑ (𝐶𝑖,𝑗+1 − 𝐶𝑖,𝑗 ) 𝐹 ( 𝑖 ) − 𝜆 1 𝑘𝑖
𝐼 (P2) 𝜕𝑄𝑖 𝑗=0 𝑉𝑖,𝑗 (32)
Subject to ∑𝐶𝑖 𝑄𝑖 ≤ 𝐵, 𝑄𝑖 ≥ 0, 𝑖 = 1, 2, . . . , 𝐼.
𝑗=1 = 0, ∀𝑖.
We concentrate on problem 1 (P1) due to the heavy
similarities between the formulations of (P1) and (P2). All the In the distribution-free approach,
following results apply to problem 2 (P2) by replacing 𝑘𝑖 and
𝐾 with 𝐶𝑖 and 𝐵, respectively. The necessary conditions for 𝜕𝐶𝑢 (𝑍) 2𝑃𝑖
= − 𝑢𝑖,𝑛 − 2
𝑄𝑖 (𝑖 = 1, 2, . . . , 𝐼) to be optimal are as follows. 𝜕𝑄𝑖 𝐶𝑖
In the traditional approach,
𝑛 (𝑄𝑖 /𝑉𝑖,𝑗−1 − 𝜇𝑖 )
𝜕𝐸𝑑 (𝑍) 𝑛−1
𝑄 + ∑ (𝑢𝑖,𝑗 − 𝑢𝑖,𝑗−1 ) (33)
= 𝛼0 − ∑ (𝑃𝑖,𝑗 − 𝑃𝑖,𝑗+1 ) 𝐹 ( 𝑖 ) − 𝜆 1 𝑘𝑖 2 1/2
𝜕𝑄𝑖 𝑉𝑖,𝑗
𝑗=1 [𝜎𝑖2 + (𝑄𝑖 /𝑉𝑖,𝑗−1 − 𝜇𝑖 ) ]
𝑗=0 (30)
= 0, ∀𝑖. 2𝜆 1 𝑘𝑖
− = 0, ∀𝑖.
𝐶𝑖
In the distribution-free approach,
𝜕𝐶𝑑 (𝑍) 2𝐶𝑖 The computational results for all four scenarios are shown in
= + 𝑑𝑖,𝑛 − 2 the following section.
𝜕𝑄𝑖 𝑃𝑖
𝑛 (𝑄𝑖 /𝑉𝑖,𝑗−1 − 𝜇𝑖 ) 4. Numerical Examples
+ ∑ (𝑑𝑖,𝑗 − 𝑑𝑖,𝑗−1 ) (31)
2 1/2
𝑗=1 [𝜎𝑖2 + (𝑄𝑖 /𝑉𝑖,𝑗−1 − 𝜇𝑖 ) ] In this section, we provide numerical illustration on how the
expected profits under the general and the distribution-free
2𝜆 1 𝑘𝑖 approaches are affected by various cases by using the models
− = 0, ∀𝑖,
𝑃𝑖 developed in Sections 2 and 3.
Mathematical Problems in Engineering 9

Example 1. We consider a product with a regular unit price quantity can be sold at the salvage value. The results show
𝑊 𝑁
and a constant unit cost. Discounts of 10%, 20%, 30%, 40%, (normal distribution in parenthesis) 𝑄𝑑𝑢 = 121.5615 (𝑄𝑑𝑢 =
𝑊
and 50% off the regular price are progressively used to sell 122.2547) and a worst-case expected profit of Π𝑑𝑢 (𝑄) =
excess inventory. We also assume that a fraction of the $258.4478 (Π𝑁 𝑑𝑢 = $258.4653). The value of EVAI is $0.0174.
original demand will be realized as additional sales at each Examples 1, 2, and 3 show that the expected profit is
of the first four discounted prices: 𝑡1 = 0.1, 𝑡2 = 0.1, 𝑡3 = 0.2, increased and the order quantity is decreased when the
and 𝑡4 = 0.3. Furthermore, we assume that an unlimited seller applies both discounts and upgrades. Therefore, we
quantity can be sold at the fifth discount price. In addition, can conclude that simultaneous discounts and upgrades may
𝑃 = $10 per unit, 𝐶 = $7.5 per unit, 𝑆 = $5 per unit, 𝜇 = 100 profit the seller.
unit, and 𝜎 = 15 unit. We compare the performance of 𝑄𝑑𝑁
with 𝑄𝑑𝑊. The results are (normal distribution in parenthesis) Example 4. We consider a newsboy problem with five prod-
ucts. The total space available for the products is 7000 ft3 . We
𝑄𝑑𝑊 = 122.0732 units (𝑄𝑛𝑁 = 122.5361 units) and a worst-case
assume that, for each product, the management offers two
expected profit of Π𝑊 𝑁
𝑑 (𝑄) = $257.4775 (Π𝑑 = $257.4845). The discounts before offering the product at salvage value. The
Expected Value of Additional Information (EVAI) is $0.007. first discount is 10% off the original price and results in 10%
Additionally, if the management did not consider multiple additional sales (𝑡𝑖,1 = 0.1), and the second discount is 25%
discounts and immediately discounted the product’s price off the original price and results in 20% additional sales (𝑡𝑖,2 =
from $10 to $5, the optimal quantity would be the same 0.2). Unlimited quantity can be sold at the third discount
compared to the result from the classical newsboy problem of price. The third discount on items 1 and 5 is 50% off the
100 units with the corresponding expected profit of $154.89. original price, while it is 35% off on item 2, 45.5% off on item 3,
Therefore, we show that progressive multiple discounts are and approximately 53.1% off on item 4. The optimal results for
profitable for the seller. the normally distributed demand and unknown distributed
demand are shown in Table 1.
Example 2. We use the same data as in Example 1 for
comparison. In this example, we consider a product with First, we solve the unconstrained problem. The total space
a regular unit cost and a constant unit price. Upgrades of required for this problem is 14, 251 ft3 which exceeds the
13.33%, 26.67%, 40%, 53.33%, and 66.67%, added without total available space. The total expected profit for the worst
increasing the cost, are progressively used to sell excess case and the normal distribution is $34111.05 and $34111.26,
inventory. We also assume that a fraction of the original respectively. The value of EVAI is $0.21. Using the proposed
demand will be realized as additional sales at each of the algorithm (Algorithm 1), we start with 𝜆 1𝑈 = 6.47, which is
first four upgrade costs: 𝑡1 = 0.1, 𝑡2 = 0.1, 𝑡3 = 0.2, and Max[(𝑃𝑖,0 − 𝐶𝑖 )/𝑘𝑖 ] for Product 4. Under the normal distri-
𝑡4 = 0.3. An unlimited quantity can be sold at the salvage bution of demand, the optimal value of 𝜆 1 is 𝜆∗1 = 1.8916,
value. The results show (normal distribution in parenthesis) and, under the distribution-free approach, the optimal value
𝑄𝑢𝑊 = 122.0732 units (𝑄𝑢𝑁 = 122.5361 units) and a worst- is 𝜆∗1 = 1.8186. The new optimal solution demonstrates
case expected profit of Π𝑊 𝑁
𝑢 (𝑄) = $257.4775 (Π𝑢 = $257.4845). that, under the normal distribution demand, total storage
The value of EVAI is $0.007. The model with discounts is of 7000 ft3 (sum of the 11th row) is needed, and, under
equivalent to the model with upgrades. Therefore, if the the distribution-free approach, total storage of 6999.82 ft3 is
impacts of discounts and upgrades are uniform, then the needed. The total expected profit for the normal distribution
seller can use either strategy to increase profit. and the worst case is $24439.74 and $24436.73, respectively.
The value of EVAI is $3.01. Moreover, from Example 4,
Example 3. We use the same data as in Example 1 for we conclude that the ordering decision determined by the
comparison. We consider discounts of 10%, 20%, 30%, 40%, newsboy problem in the multiple-item environment is highly
and 50% off the regular price and upgrades of 5%, 10%, 15%, influenced by an additional constraint. Finally, all of the above
and 20% are progressively used, with no increase in price, to examples demonstrate the robustness of the distribution-free
sell excess inventory. We assume that management offers four approach.
discounts and four upgrades before offering the product at
salvage value. The first discount is 10% off the original price 5. Concluding Remarks
and results in 10% additional sales (𝑡1 = 0.1), and the first
upgrade is 5% of the original cost and results in 5% additional Motivated by the discounts and the quality upgrade strategy
sales (𝑡2 = 0.05). The second discount is 20% off the original practices in the real world, we extend and derive the optimal
price and results in 10% additional sales (𝑡3 = 0.1), and the ordering rule for the newsboy problem with multiple dis-
second upgrade is 10% of the original cost and results in 5% counts, multiple upgrades, and the combined uses of both
additional sales (𝑡4 = 0.05). The third discount is 30% off strategies in single- and multiple-item environments. We
the original price and results in 20% additional sales (𝑡5 = adopt the distribution-free approach where only available
0.2). The third upgrade is 15% of the original cost and results information is the mean and standard deviation of demand
in 10% additional sales (𝑡6 = 0.1). The fourth discount is and compare the results with those found through the
40% off the original price and results in 30% additional sales traditional approach. We conclude that the distribution-free
(𝑡7 = 0.3). The fourth upgrade is 20% of the original cost and approach is robust from computational results. Experiments
results in 15% additional sales (𝑡8 = 0.15), while unlimited with numerical data under an identical environment indicate
10 Mathematical Problems in Engineering

Table 1: Data and the optimal results for the basic and distribution-free model with a storage constraint.

Parameter 1 2 3 4 5
𝑃𝑖 120 100 220 160 130
𝐶𝑖 80 75 170 105 100
𝑆𝑖 60 (0.5) 65 (0.35) 120 (0.455) 75 (0.531) 65 (0.5)
𝜇𝑖 200 250 120 150 180
𝜎𝑖 40 50 15 30 40
𝑘𝑖 21 7 12 8.5 16.25
𝑄𝑖𝑁 257.57 315.46 138.82 192.18 205.20
Without
𝑄𝑖𝑊 257.40 314.33 138.80 192.15 205.84
storage
constraint 𝜋𝑖 (𝑄𝑖𝑁 )
𝑁
8246.49 6277.35 6071.14 8495.51 5020.77
𝜋𝑖𝑊(𝑄𝑖𝑊) 8246.49 6277.24 6071.14 8495.51 5020.67
𝑄𝑖𝑁 107.94 253.05 124.89 172.14 0
With
𝑄𝑖𝑊 105.86 257.18 125.14 173.50 0
storage
constraint 𝜋𝑖𝑁 (𝑄𝑖𝑁 ) 4313.66 5876.84 5911.79 8337.45 0
𝜋𝑖𝑊(𝑄𝑖𝑊) 4231.19 5929.64 5917.35 8358.55 0

that a greater number of discounts and upgrades are associ- [4] I. Moon and S. Choi, “The distribution free continuous review
ated with greater expected profits. The result offers a novel inventory system with a service level constraint,” Computers and
addition to the literature on the newsboy problem. Industrial Engineering, vol. 27, no. 1–4, pp. 209–212, 1994.
The study proposes several directions for future research. [5] I. Moon and S. Choi, “Distribution free procedures for make-
The developed model under a multiple-item scenario to-order (MTO), make-in-advance (MIA), and composite poli-
assumes the presence of independent demand and the influ- cies,” International Journal of Production Economics, vol. 48, no.
ence of discounts or upgrades for various products. Therefore, 1, pp. 21–28, 1997.
we encourage the development of a new approach where [6] I. Moon and S. Choi, “A note on lead time and distributional
researchers also consider demand substitution effect and assumptions in continuous review inventory models,” Comput-
price-dependent demand as well as bundle discounts under ers and Operations Research, vol. 25, no. 11, pp. 1007–1012, 1998.
various distribution functions. In addition, researchers may [7] L.-Y. Ouyang and K.-S. Wu, “A minimax distribution free
extend all the models in the supply chain framework to procedure for mixed inventory model with variable lead time,”
study the effect of bargaining power of channel members on International Journal of Production Economics, vol. 56-57, pp.
discounts or upgrades. 511–516, 1998.
[8] M. Hariga and M. Ben-Daya, “Some stochastic inventory
models with deterministic variable lead time,” European Journal
Competing Interests of Operational Research, vol. 113, no. 1, pp. 42–51, 1999.
The authors declare that they have no competing interests. [9] L.-Y. Ouyang and H.-C. Chang, “A minimax distribution free
procedure for mixed inventory models involving variable lead
time with fuzzy lost sales,” International Journal of Production
Acknowledgments Economics, vol. 76, no. 1, pp. 1–12, 2002.
[10] M. M. Tajbakhsh, “On the distribution free continuous-review
This research was supported by the National Research Foun-
inventory model with a service level constraint,” Computers &
dation of Korea (NRF) funded by the Ministry of Science, ICT Industrial Engineering, vol. 59, no. 4, pp. 1022–1024, 2010.
& Future Planning (2015R1A2A1A15053948).
[11] I. Moon, E. Shin, and B. Sarkar, “Min-max distribution free
continuous-review model with a service level constraint and
References variable lead time,” Applied Mathematics and Computation, vol.
229, pp. 310–315, 2014.
[1] H. Scarf, “A min max solution of an inventory problem,” in Stud-
[12] J. Kamburowski, “On the distribution-free newsboy problem
ies in the Mathematical Theory of Inventory and Production, pp.
with some non-skewed demands,” Operations Research Letters,
201–209, Stanford University Press, Stanford, Calif, USA, 1958.
vol. 43, no. 2, pp. 165–171, 2015.
[2] G. Gallego and I. Moon, “The distribution free newsboy [13] M.-H. Shu, C.-W. Yeh, and Y.-C. Fu, “Impacts of transportation
problem: review and extensions,” Journal of the Operational cost on distribution-free newsboy problems,” Mathematical
Research Society, vol. 44, no. 8, pp. 825–834, 1993. Problems in Engineering, vol. 2014, Article ID 307935, 10 pages,
[3] I. Moon and G. Gallego, “Distribution free procedures for some 2014.
inventory models,” Journal of the Operational Research Society, [14] W.-M. Ma and B.-B. Qiu, “Distribution-free continuous review
vol. 45, no. 6, pp. 651–658, 1994. inventory model with controllable lead time and setup cost in
Mathematical Problems in Engineering 11

the presence of a service level constraint,” Mathematical Prob- [33] M. Mussa and S. Rosen, “Monopoly and product quality,”
lems in Engineering, vol. 2012, Article ID 867847, 16 pages, 2012. Journal of Economic Theory, vol. 18, no. 2, pp. 301–317, 1978.
[15] H. Bidkhori, D. Simchi-Levi, and Y. Wei, “Analyzing process [34] D. A. Garvin, “What does product quality really mean,” MIT
flexibility: a distribution-free approach with partial expecta- Sloan Management Review, vol. 26, pp. 25–43, 1984.
tions,” Operations Research Letters, vol. 44, no. 3, pp. 291–296, [35] Toyota, “New initiatives for quality improvement,” Toyko, 2012,
2016. http://www.toyota-global.com/company/history of toyota/
[16] K. Kwon and T. Cheong, “A minimax distribution-free proce- 75years/text/entering the automotive business/chapter2/sec-
dure for a newsvendor problem with free shipping,” European tion1/item3.html.
Journal of Operational Research, vol. 232, no. 1, pp. 234–240, [36] L. A. Crosby, K. R. Evans, and D. Cowles, “Relationship quality
2014. in services selling: an interpersonal influence perspective,”
[17] Y. Qin, R. Wang, A. J. Vakharia, Y. Chen, and M. M. H. Seref, Journal of Marketing, vol. 54, no. 3, pp. 68–81, 1990.
“The newsvendor problem: review and directions for future
[37] X. Xu, “Optimal price and product quality decisions in a distri-
research,” European Journal of Operational Research, vol. 213, no.
bution channel,” Management Science, vol. 55, no. 8, pp. 1347–
2, pp. 361–374, 2011.
1352, 2009.
[18] G. Hadley and T. M. Whitin, Analysis of Inventory Systems,
[38] P. Ma, H. Wang, and J. Shang, “Supply chain channel strategies
Prentice-Hall, Englewood Cliffs, NJ, USA, 1963.
with quality and marketing effort-dependent demand,” Interna-
[19] E. Carrizosa, A. V. Olivares-Nadal, and P. Ramı́rez-Cobo, tional Journal of Production Economics, vol. 144, no. 2, pp. 572–
“Robust newsvendor problem with autoregressive demand,” 581, 2013.
Computers and Operations Research, vol. 68, pp. 123–133, 2016.
[39] K. Bridson, J. Evans, and M. Hickman, “Assessing the relation-
[20] Q. Li, C. Dong, and R. Zhuang, “Managing the newsvendor
ship between loyalty program attributes, store satisfaction and
modeled product system with random capacity and capacity-
store loyalty,” Journal of Retailing and Consumer Services, vol. 15,
dependent price,” Mathematical Problems in Engineering, vol.
no. 5, pp. 364–374, 2008.
2015, Article ID 296132, 10 pages, 2015.
[21] T. Ye and H. Sun, “Price-setting newsvendor with strategic
consumers,” Omega, vol. 63, pp. 103–110, 2016.
[22] N. C. Petruzzi and M. Dada, “Pricing and the newsvendor prob-
lem: a review with extensions,” Operations Research, vol. 47, no.
2, pp. 183–194, 1999.
[23] H. Chen and A. R. Rao, “When two plus two is not equal to
four: errors in processing multiple percentage changes,” Journal
of Consumer Research, vol. 34, no. 3, pp. 327–340, 2007.
[24] M. Khouja, “The newsboy problem under progressive multiple
discounts,” European Journal of Operational Research, vol. 84,
no. 2, pp. 458–466, 1995.
[25] M. Khouja and A. Mehrez, “A multi-product constrained news-
boy problem with progressive multiple discounts,” Computers &
Industrial Engineering, vol. 30, no. 1, pp. 95–101, 1996.
[26] M. J. Khouja, “Optimal ordering, discounting, and pricing in
the single-period problem,” International Journal of Production
Economics, vol. 65, no. 2, pp. 201–216, 2000.
[27] I. Moon and E. A. Silver, “The multi-item newsvendor problem
with a budget constraint and fixed ordering costs,” Journal of the
Operational Research Society, vol. 51, no. 5, pp. 602–608, 2000.
[28] L. L. Abdel-Malek and R. Montanari, “An analysis of the
multi-product newsboy problem with a budget constraint,”
International Journal of Production Economics, vol. 97, no. 3, pp.
296–307, 2005.
[29] G. Zhang, “The multi-product newsboy problem with supplier
quantity discounts and a budget constraint,” European Journal
of Operational Research, vol. 206, no. 2, pp. 350–360, 2010.
[30] S. Nahmias and C. P. Schmidt, “An efficient heuristic for the
multi-item newsboy problem with a single constraint,” Naval
Research Logistics, vol. 31, no. 3, pp. 463–474, 1984.
[31] H.-S. Lau and A. H.-L. Lau, “The newsstand problem: a capaci-
tated multiple-product single-period inventory problem,” Euro-
pean Journal of Operational Research, vol. 94, no. 1, pp. 29–42,
1996.
[32] S. J. Erlebacher, “Optimal and heuristic solutions for the multi-
item newsvendor problem with a single capacity constraint,”
Production and Operations Management, vol. 9, no. 3, pp. 303–
318, 2000.
Advances in Advances in Journal of Journal of
Operations Research
Hindawi Publishing Corporation
Decision Sciences
Hindawi Publishing Corporation
Applied Mathematics
Hindawi Publishing Corporation
Algebra
Hindawi Publishing Corporation
Probability and Statistics
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

The Scientific International Journal of

World Journal
Hindawi Publishing Corporation
Differential Equations
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Submit your manuscripts at

http://www.hindawi.com

International Journal of Advances in

Combinatorics
Hindawi Publishing Corporation
Mathematical Physics
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Journal of Journal of Mathematical Problems Abstract and Discrete Dynamics in

Complex Analysis
Hindawi Publishing Corporation
Mathematics
Hindawi Publishing Corporation
in Engineering
Hindawi Publishing Corporation
Applied Analysis
Hindawi Publishing Corporation
Nature and Society
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

International
Journal of Journal of
Mathematics and
Mathematical
Discrete Mathematics
Sciences

Journal of International Journal of Journal of

Hindawi Publishing Corporation Hindawi Publishing Corporation Volume 2014

Function Spaces
Hindawi Publishing Corporation
Stochastic Analysis
Hindawi Publishing Corporation
Optimization
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014