Data Analysis of Shipment for Textiles and Apparel from Logistics Warehouse to Store Considering Disposal Risk

sustainability

Article Data Analysis of Shipment for Textiles and Apparel from Logistics Warehouse to Store Considering Disposal Risk

Rina Tanaka 1, Aya Ishigaki 1,* , Tomomichi Suzuki 1, Masato Hamada 2 and Wataru Kawai 2

1 Industrial Administration, Tokyo University of Science, Noda, Chiba 278-8510, Japan; [email protected] (R.T.); [email protected] (T.S.) 2 Data-Chef Co., Ltd., Koto-ku, Tokyo 135-0004, Japan; [email protected] (M.H.); [email protected] (W.K.) * Correspondence: [email protected]; Tel.: +81-4-7124-1501

Received: 24 November 2018; Accepted: 2 January 2019; Published: 7 January 2019

Abstract: Given the rapid diversification of products in the textile and apparel industry, manufacturers face significant new challenges in production. The life cycle of apparel products has contracted and is now, generally, a several-week season, during which time a majority of products are supposed to be sold. Products that do not sell well may be sold at a price lower than the fixed price, and products that do not sell at all within the sales period may eventually become forced disposal. This creates long-term management and environmental problems. In practice, shipping personnel determine when to ship products to stores after reviewing product sales information. However, they may not schedule or structure these shipments properly because they cannot effectively monitor sales for a large number of products. In this paper, shipment is considered to reduce the risk of product disposal on the premise of selling at a fixed price. Although shipment quantities are determined by various factors, we only consider the change in inventory at the logistics warehouse, since it is difficult to incorporate all factors into the analysis. From cluster analysis, it is found that shipping personnel should recognize a policy to sell products gradually over time. Furthermore, to reduce the risk of disposal, we forecast the inventory from conditional probability and are able to extract products out of a standard grouping using past data.

Keywords: apparel products; supply chain management; quick response; clustering; forecasting; sale at ﬁxed price

1. Introduction Environmental problems such as global warming and water shortages along with the expansion of social disparities from globalization not only remain unsolved in the world, they continue to escalate. Thus, the need for companies to play signiﬁcant roles in shaping sustainable societies is also increasing [1]. A society can be considered sustainable from three perspectives—environmental, societal, and economic [2]; these are often referred to as the triple bottom line or 3BL [3]. Companies must consider the impact of their business activities on the 3BL and establish long-term corporate strategies to minimize adverse impacts in any of these dimensions. By employing this perspective in supply chain management, in particular, companies are more likely to achieve sustainability [4], which in turn can enhance a company’s competitive advantage [5]. Even in the textile and apparel industry—which is subject to the vagaries of fashion and, thus, characterized by short product life cycles—companies are focusing more on sustainability as globalization progresses [6]. In this industry, the emphasis in sustainable supply chain management is placed on effective risk management.

Sustainability 2019, 11, 259; doi:10.3390/su11010259 www.mdpi.com/journal/sustainability Sustainability 2019, 11, 259 2 of 14

In a textile and apparel supply chain, a company’s factories manufacture products according to a designated production plan. Afterwards, via logistics warehouses, the company ships finished products to stores. At the start of a sales period, a quantity of products is made available at stores based on the production plan. At the end of the sales period, it is hoped that inventories of stores and the logistics warehouse will have harmonized with the production plan as well as with customer demand. In practice, however, customer demand often deviates from what is anticipated in the production plan, and a surplus or shortage results. This is because of the long lead times before a sales period [7] and because apparel products are seasonal, that is, usually only one season. Further, in the textile and apparel industry, the logistics warehouses serve as stock points, and therefore, decisions about shipments to logistics warehouses are crucial [8]. In recent years, as customer needs and preferences have proliferated, textile and apparel products have been made available in an ever-widening range of colors and sizes [9]. Moreover, an increasing number of manufacturers also sell more general apparel products—such as bags, shoes, and accessories—in addition to clothes [10]. Because of these trends, leading to modifications in the design of apparel products every season, demand uncertainty in the industry is large compared to that for industries with more stable products, such as office supplies, or furniture. In other words, every season there are new apparel products with very short life cycles [9]. Therefore, it is difficult to forecast when and which products will be sold at any given time [11]. Many factors are at play in textile and apparel supply chains. When shipping products from the production plants to the logistics warehouse, timing and quantities are determined in advance. On the other hand, when shipping from the logistics warehouse to stores, timing and quantities can be modified in accordance with customer demand by taking into account actual sales information. The earlier that shipments are sent from the logistics warehouse, the higher is the possibility that inventory at the stores will become excessive, such that unsold products will eventually be returned to the logistics warehouse and discarded. Inventory space in a store may also be quite limited, resulting in the need to return or dispose of excess product. In contrast, when shipments from the logistics warehouse are delayed, stores can experience shortages and, as a result, lose opportunities for sales. In addition, because of long lead times before sales seasons, trends may have already have shifted by the time the season arrives, with the result that product designs will have become unsuitable for selling. Shipping personnel are responsible for controlling inventory by determining shipments (timing and quantities) from the logistics warehouse to stores. Normally, these personnel rely on signals from stores when deciding if additional shipments are necessary [12]. Because of the range of products and frequency of shipments, however, it is difficult for them to efficiently and accurately track all sales information. These shortcomings can result in excess product, the proper disposal of which has become a significant environmental problem. The textile and apparel industry is the second largest industry in the world; as such, it can have an enormous impact on the global environment [13], particularly since waste is generated at every stage of in the apparel manufacturing process [14]. In sum, the disposal of excess product within the apparel industry is an increasing threat to the environment [15]. To minimize the quantity of unsold product, manufacturers should more carefully consider customer demand when making production decisions. In this way, they can fulfill their role in helping to build a sustainable society, at least from an environmental perspective. From an economic perspective, manufacturers face another related challenge: anticipating a disposal problem, they may significantly lower the price of products to try to sell more product and, hence, mitigate the challenge of disposal. An increase in the number of unsold products at the end of a sales cycle, then, would lead to an increase in the number of products whose price was discounted by the company. If this cycle were to continue, profits would inevitably decline and sustaining the business over the long term would become more difficult. To prevent problems such as this in the economic dimension, companies would need to increase or at least maintain the number of products sold at a fixed price. Sustainability 2019, 11, 259 3 of 14

On the premise of selling at a ﬁxed price, we consider the timing of shipping as a method to reduce waste risk due to unsold inventory in this study. Initially, the sales scenario of the products that the shipping personnel need to monitor is restricted. Next, in consideration of the need to sell these products at a ﬁxed price, personnel forecast the inventory ratio in the sales period and consider shipping timing in order to reduce the disposal risk.

2. Literature Review Unsustainable and unregulated production and consumption are contributing to the degradation of the global environment [16]. Proper monitoring and analysis in the textile and apparel supply chain is necessary to minimize its environmental impact [17]. A literature review on the apparel supply chain is as follows. The sustainability of the textile and apparel supply chain has two aspects: environment sustainability such as reduction of unsold products and economic sustainability that improves return on investment (ROI) [18]. With environment sustainability, manufacturers are required to manage resource procurement, inventory, waste treatment, and transport efficiently. On the other hand, with economic sustainability, manufacturers are required to increase the number of products sold at regular price and have large sales. Caniato et al. investigated large and small companies on the problem of environment sustainability in fashion supply chain [19]. As a result, large companies tend to focus more on improving products and processes, and small and medium companies tend to rebuild supply chains. Choi and Chiu proposed the mean-downside-risk (MDR) and mean-variance (MV) newsvendor model considering both environment sustainability and economic sustainability [20]. Brun and Castelli explored specific factors or product characteristics that promote competition in the fashion industry [21]. Ciarnienˇ e˙ and Vienažindiene˙ identified key features of the modern fashion industry and developed a time efficient supply chain model [22]. Mehrjoo and Pasek described the structure of the apparel supply chain using system dynamics and quantitatively evaluated its risk [23]. Patil et al. developed a mathematical model to evaluate procurement, shipping, and markdown plans for new short-term life cycle products considering discounts associated with large-scale purchase and bulk transport [24]. In order for the shipping personnel to recognize when to send additional shipments, they rely on signals that indicate when products have been sold at stores. This system is known as Quick Response (QR), and it has demand-driven characteristics. Quick Response reduces lead time and allows shippers to quickly respond to market needs [25]. It was first implemented in the US apparel industry in the middle 1980s as a strategy for global response [26] and, since then, its use has spread to other industries. It was designed to be an inventory management strategy that was based on sharing information among supply chain agents [27]. By using a QR strategy, lead time is shorter because the shipping personnel can better forecast customer demand and subsequently modify the production plan [25]. In basic QR, a retailer sends point of sale (POS) data to its supplier, who then uses this information to improve demand forecasting [28]. Tsukagoshi et al. proposed using a QR inventory management strategy of introducing a monitoring period prior to the normal sales period in cases for which replenishment lead time is short compared to the sales period. They found that this approach reduced total shortage costs [29]. Tanaka et al. analyzed shipping data on the premise of selling at a fixed price [8]. They considered the risk of running out of stock but not the risk of disposal of unsold products. In the apparel market, sales forecasting plays an increasingly important role in decision support systems for market competition and globalization [30]. Thomassey explained specific constraints and needs related to supply chain and sales forecasts in the textile apparel industry [31]. In a large set of SKUs and two consecutive sales seasons, Mostard et al. compared the accuracy of three quantitative forecasting methods based on advance demand information [11]. Fan et al. proposed a prediction model combining the Bass/Norton model and the sentiment analysis method [32]. Sustainability 2019, 11, 259 4 of 14

Sustainability 2019, 11 FOR PEER REVIEW 4 3. Status of Products 3. Status of Products InSustainability general, 2019 a manufacturer, 11 FOR PEER REVIEW stores new products in one logistics warehouse for each supply chain.4 Also, a rangeIn general, of products a manufacturer can be stores shipped new fromproducts one in logistics one logistics warehouse warehouse to for multiple each supply stores, chain. but the speciﬁcAlso,3. Status products a range of Products andof products quantities can shippedbe shipped depend from one on thelogistics unique warehouse properties to multiple of stores, stores, such but as clientele,the specific products and quantities shipped depend on the unique properties of stores, such as clientele, sales ﬂoorIn space, general, and a manufacturer geographic location stores new or products climate. in If one one logistics attempts warehouse to consider for each all the supply characteristics chain. sales floor space, and geographic location or climate. If one attempts to consider all the characteristics of storesAlso, in a detail, range theof products scope of can the databe shipped becomes from very one large. logistics For warehouse this reason, to we multiple analyze stores, the composition but the of stores in detail, the scope of the data becomes very large. For this reason, we analyze the of shippingspecific fromproducts one and logistics quantities warehouse shipped depend to one store.on the unique Hence, properties in this paper, of stores, the such textile as clientele, and apparel composition of shipping from one logistics warehouse to one store. Hence, in this paper, the textile supplysales chain floor is space, considered and geographic as shown location in the or Figure climate.1. If This one attempts supply chainto consider includes all the multiple characteristics factories, and apparel supply chain is considered as shown in the Figure 1. This supply chain includes multiple of stores in detail, the scope of the data becomes very large. For this reason, we analyze the one logisticsfactories, warehouse,one logistics onewarehouse, store, and onemany store, customers.and many customers. Products Products arrive in arrive the logistics in the logistics warehouse composition of shipping from one logistics warehouse to one store. Hence, in this paper, the textile fromwarehouse multiple factories, from multiple and then factories, the shipping and then personnelthe shipping ship personnel products ship that products have arrived that have in arrived the logistics and apparel supply chain is considered as shown in the Figure 1. This supply chain includes multiple warehousein the logistics to one store.warehouse to one store. factories, one logistics warehouse, one store, and many customers. Products arrive in the logistics warehouse from multiple factories, and then the shipping personnel ship products that have arrived in the logistics warehouse to one store.

FigureFigure 1. 1.Conceptual Conceptual DiagramDiagram of of Supply Supply Chain Chain Flow. Flow.

Then,Then, we considerwe consider inventory inventory change change inin ourour logisticslogistics warehouse. warehouse. Figure Figure 2 shows2 shows changes changes in the in the Figure 1. Conceptual Diagram of Supply Chain Flow. percentagespercentages of the of the inventories inventories of of products products inin anan actual logistics logistics warehouse. warehouse. As shown As shown in Figure in Figure 2, 2, there are products with inventory ratios of zero in the logistics warehouse at nine weeks, while some there are productsThen, we consider with inventory inventory ratios change of zeroin our in logi thestics logistics warehouse. warehouse Figure at2 shows nine weeks, changes while in the some products remain in high proportions. Figure 3 shows a histogram of the inventory ratios at the ninth productspercentages remain of in the high inventories proportions. of products Figure 3in shows an actual a histogram logistics warehouse. of the inventory As shown ratios in Figure at the 2, ninth week. The ratio is less than 0.10 for most products, but there are still products with considerable week.there The are ratio products is less with than inventory 0.10 for ratios most ofproducts, zero in the butlogistics there warehouse are still productsat nine weeks, with while considerable some inventory remaining. There are two reasons why large inventories for a product could persist. The inventoryproducts remaining. remain in high There proportions. are two reasonsFigure 3 show whys largea histogram inventories of the inventory for a product ratios at could the ninth persist. first is overproduction, that is, a discrepancy between scheduled and actual sales of product. As a The firstweek. is The overproduction, ratio is less than that 0.10 is, for a discrepancymost products, between but there scheduled are still products and actual with salesconsiderable of product. countermeasure, there is the method of discounting and selling inventory or selling it through inventory remaining. There are two reasons why large inventories for a product could persist. The As aanother countermeasure, channel, such there as is an the electronic method commerce of discounting (EC) site. and sellingThe second inventory potential or sellingcause is it that through first is overproduction, that is, a discrepancy between scheduled and actual sales of product. As a anothershipments channel, have such been as antimed electronic poorly. commerceThe shipping (EC) personnel site. The are second the ones potential who determine cause is thatthe timing shipments countermeasure, there is the method of discounting and selling inventory or selling it through haveand been quantities timed poorly. of products The shipping to be shipped personnel based areon sales the onesinformation. who determine However, the due timing to the andchallenge quantities another channel, such as an electronic commerce (EC) site. The second potential cause is that of managing a large number of products, personnel may overlook a signal that a product has sold at of productsshipments to behave shipped been timed based poorly. on sales The information. shipping personnel However, are the due ones to who the challengedetermine ofthe managing timing a its fixed price. If the number of unsold fixed-price products increases, there will be an increase in largeand number quantities of products, of products personnel to be shipped may overlookbased on sales a signal information. that a product However, has due sold to the atits challenge fixed price. inventory in need of disposal at the end of the sales period. After considering sustainability, firms If theof number managing of unsolda large number fixed-price of products, products personnel increases, may there overlook will bea signal an increase that a product in inventory has sold in at need should increase the number of products sold at a fixed price as much as possible. of disposalits fixed at price. the end If the of thenumber sales of period. unsold Afterfixed-price considering products sustainability, increases, there firms will be should an increase increase in the numberinventory of products in1.00 need sold of disposal at a fixed at pricethe end as of much the sa asles possible. period. After considering sustainability, firms should increase0.90 the number of products sold at a fixed price as much as possible. 0.801.00 0.700.90 0.600.80 Product 1 0.500.70 0.400.60 Product 2 0.300.50 Product 1 Inventory Ratio Product 3 0.20 0.40 Product 2 0.100.30

Inventory Ratio Product 3 0.000.20 0.10 0123456789 0.00 Week 0123456789Figure 2. Example of Inventory Ratio Change. Week FigureFigure 2. 2.Example Example of of InventoryInventory Ratio Ratio Change. Change.

Sustainability 2019, 11, 259 5 of 14 Sustainability 2019, 11 FOR PEER REVIEW 5

500 450 400 350 300 250 200 Frequency 150 100 50 0 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 Inventory Ratio FigureFigure 3.3.Histogram Histogram ofof InventoryInventory RatioRatio inin NinthNinth Week.Week. 4. Methods 4. Methods In this paper, we assume the supply chain as shown in the Figure1. Since the sales cycle of the In this paper, we assume the supply chain as shown in the Figure 1. Since the sales cycle of the textile and apparel supply chain is one week [33], we perform our analysis using weekly data. The data textile and apparel supply chain is one week [33], we perform our analysis using weekly data. The used in this paper is collected in the logistics warehouse. data used in this paper is collected in the logistics warehouse. 4.1. Assumptions, Parameters, and Variables 4.1. Assumptions, Parameters, and Variables In this paper, the following assumptions, parameters, and variables are used. The parameters and variablesIn this are paper, described the infollowing subsequent assumptions, sections. parameters, and variables are used. The parameters and variables are described in subsequent sections. Assumptions: Assumptions: • Products are sold in the same period • Products are sold in the same period • The total arrival quantity of each product is known • The total arrival quantity of each product is known • Product movement among stores is not considered • Product movement among stores is not considered • • TheThe period period in in which which products products are are sold sold at at the the ﬁxed fixed price price is is nine nine weeks weeks •• SinceSince products products are are sold sold up up for for nine nine weeks, weeks, the the sales sales period period is is nine nine weeks weeks •• IfIf the the product product did did not not sell sell for for nine nine weeks, weeks, the the product product is is discarded discarded Parameters and Variables: Parameters and Variables: N: number of products N: number of products T: sales period [weeks] T: sales period [weeks] t: elapsed periods [weeks] (t = 1, …, T) t: elapsedi: product periods number [weeks] (i = (t 1,= …, 1,. N . .) , T) i: productdi (t): numbershipping (i amount= 1, . . . ,ofN product) i in week t di (t):D shippingi: total delivery amount amount of product of producti in week i t Di: totalui (t delivery): shipping amount rate in of week product t of iproduct i ui (t):K shipping: number rate of clusters in week t of product i K: numberT’: investigated of clusters sales period

T’: investigatedIit, : inventory sales ratio period of product i in week t Ii,t: inventory ratio of product i in week t X i : random variable representing the realized inventory ratio of product i Xi: random variable representing the realized inventory ratio of product i Y : random variable representing the predicted inventory ratio of product i Y : randomi variable representing the predicted inventory ratio of product i i μ′ µ0: conditional: conditional expectation expectation

Sustainability 2019, 11 FOR PEER REVIEW 6 Sustainability 2019, 11, 259 6 of 14 μ X X : average of i µX: averageμ of Xi Y Y : average of i µ : average of Y Y ρ: correlationi coefficient ρ: correlationσ coefﬁcient x X i σx: variance: variance of Xi of σ : varianceσ of Y Y Y Y : variancei of i σ0: variance in conditional probability σ ′ : variance in conditional probability 4.2. Decision-Making Process for Shipping Personnel 4.2. Decision-Making Process for Shipping Personnel Based on the above parameters and variables, Figure4 shows the decision-making process Based on the above parameters and variables, Figure 4 shows the decision-making process of of shipping personnel. It has seven steps. This paper proposes methods for supporting shipping shipping personnel. It has seven steps. This paper proposes methods for supporting shipping personnel who are in charge of each decision-making process. personnel who are in charge of each decision-making process.

Figure 4. Decision-Making Process for Shipping Personnel. Figure 4. Decision-Making Process for Shipping Personnel. These steps are for product i. These steps are for product i. Step 1. Product i arrives at the logistics warehouse from a plant, and its quantity is Di. Step 1. Product i arrives at the logistics warehouse from a plant, and its quantity is Di. i StepStep 2. 2. ShippingShipping personnel personnel practice practice first first shipment shipment of product of productbased i based on asales on apolicy sales policy which iswhich given is by manufactures and its quantity is di (1). given by manufactures and its quantity is di (1). StepStep 3. 3. ShippingShipping personnel personnel recognize recognize sales sales information information of productof producti for i for one one week. week. StepStep 4. 4. ShippingShipping personnel personnel consider consider and and decide decide whether whether to to ship ship additional additional product product during during the the next next week.week. At thisAt this time, time, they referthey notrefer only not to only sales to quantity sales quantity of product ofi butproduct also toi salesbut also information to sales of similarinformation products of similar sold last products year and sold current last year trends. and current trends. StepStep 5. 5. BasedBased on on shipping shipping personnel personnel decision, decision, product producti is shippedi is shipped whose whose quantity quantity is di (ist) d ati (tt)-th at t-th week.week. In theIn the determination determination of the of quantity,the quantity, the inventorythe inventory of the of store the andstore the and scale the ofscale the storeof the arestore taken are into taken consideration. into consideration. Then, if personnel Then, if decidepersonnel not decide to make not an additionalto make an shipment, additional shippingshipment, quantity shipping of the quantity product ofd ithe(t) product is zero. di (t) is zero. StepStep 6. 6. If theIf the next next week, week,t + t 1, + is1, endis end of salesof sales period periodT, goT, go to Stepto Step 7. Otherwise,7. Otherwise, shipping shipping personnel personnel returnreturn to Stepto Step 3 and 3 and repeat repeat from from Step Step 3 to 3 Stepto Step 5 until 5 until they they fulfill fulfill Step Step 6. 6. StepStep 7. 7. ShippingShipping personnel personnel finish finish their their decision. decision.

Sustainability 2019, 11, 259 7 of 14

4.3. Product Data Extraction Products in the apparel and textile industry come in many categories, colors, and sizes; thus, the amountSustainability of 2019 available, 11 FOR productPEER REVIEW data enormous. Further, customer demand is uncertain, and7 sales area affected by fashion and other industry trends that are in a state of constant flux. In light of these 4.3. Product Data Extraction circumstances, the data that shipping personnel have access to at any given time is limited in scope. Therefore,Products when analyzing in the apparel our and own textile data, industry we divide come thein many products categories, into severalcolors, and groups, sizes; thus, extract the data fromamount products of likelyavailable to requireproduct shippingdata enormous. plan revision,Further, customer and analyze demand them. is uncertain, and sales area affected by fashion and other industry trends that are in a state of constant flux. In light of these An example of the actual change in inventory in a logistics warehouse is shown in Table1. There is circumstances, the data that shipping personnel have access to at any given time is limited in scope. a significantTherefore, difference when analyzing in product our quantitiesown data, we even divide in the the first products week, into making several it difficult groups, toextract see inventorydata transitionsfrom products of all products likely to on require one scale shipping of inventory plan revision, quantity. and analyze Therefore, them. the shipping ratio is calculated as the incomingAn example inventory of the actual in each change week in inventorydi(t), normalized in a logistics by warehouse the total is arrival shown quantityin Table 1.D Therei of each productis ai :significant difference in product quantities even in the first week, making it difficult to see inventory transitions of all products on one scale of dinventoryi(t) quantity. Therefore, the shipping ratio ui(t) = . (1) is calculated as the incoming inventory in each weekD dii(t), normalized by the total arrival quantity Di of each product i: After calculating the shipping rate for each product, a k-means cluster analysis of u (t) is conducted dt() i for t = 1, ... , 9, with a sample of N = 796 in() order= i to divide products into groups. Since the number uti . (1) of pieces of data is as large as N = 796, we decidedDi to use nonhierarchical instead of hierarchical clustering.After To calculatecalculating the the distance shipping between rate for samples, each product, Euclidean a k-means distance cluster is used. analysis One of disadvantage ui(t) is of theconductedk-means for method t = 1, …, is the9, with dependence a sample of on N = the 796 initial in order value. to divide Therefore, products some into groups. initial values Since the are set and clusternumber analysis of pieces is of performed, data is as andlarge initial as N values= 796, we are decided chosen to minimizeuse nonhierarchical the distance instead within of the averagehierarchical cluster. clustering. The results To ofcalculate the cluster the distance analysis between leads samples, to patterns, Euclidean here calleddistance sales is used. policies One (i.e., disadvantage of the k-means method is the dependence on the initial value. Therefore, some initial policies involving selling by advertising products that will be recognized by customers vs. policies to values are set and cluster analysis is performed, and initial values are chosen to minimize the distance sell normally,within the as average standard cluster. products). The results of the cluster analysis leads to patterns, here called sales policies (i.e., policies involving selling by advertising products that will be recognized by customers Table 1. Examples of Changes in Inventory. vs. policies to sell normally, as standard products). Change in Inventory in Each Week Table 1. Examples of Changes in Inventory. Product 0 1 2 3 4 5 6 7 8 9 Change in Inventory in Each Week A 2884 291 266 240 198 154 123 80 44 36 Product 0 1 2 3 4 5 6 7 8 9 B 2078 1151 1038 347 221 171 122 0 0 0 A C 2884 293 291 249 266 248 240 220 198 220 220154 191 123 116 80 94 44 32 36 B 2078 1151 1038 347 221 171 122 0 0 0 C 293 249 248 220 220 220 191 116 94 32 The result of the k-means method suggested that it was reasonable to divide the analyzed data into two groups,The result which of the were k-means then method roughly suggested divided that into it was five reasonable groups to to provide divide morethe analyzed detail. data Figure 5 showsinto the two average groups, value which of were inventory then roughly ratios divided in each in weekto five for groups each ofto theprovide five more clusters. detail. Depending Figure 5 on the shipmentshows the percentages average value in of the inventory first week, ratios five in each clusters week can for beeach broadly of the five categorized. clusters. Depending on the shipment percentages in the first week, five clusters can be broadly categorized. 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 Inventory Ratio Inventory 0.1 0 0123456789 Week

1 2 3 4 5

FigureFigure 5. Average5. Average Inventory Inventory RatiosRatios in Each Each Week Week for for Each Each Cluster. Cluster.

Sustainability 2019, 11, 259 8 of 14

For clusters 1 and 2, the shipment ratio in the ﬁrst week is high. On the other hand, in clusters 3, 4, and 5, the shipment ratio of the ﬁrst week is low. Investigating the characteristics of each cluster further, products of clusters 1 and 2 represent the policy that a manufacturer wants to sell featured products. Also, the trend of shipment ratios from the second to the ninth week shows that products remaining in the logistics warehouse are gradually shipped until the end of the sale period. Therefore, the shipping personnel will not revise the shipping plan later. On the contrary, products in clusters 3, 4, and 5 demonstrate a policy whereby the shipping personnel sell the product in stages, conditional on the level of sales. Accordingly, products belonging to these classes are highly likely to have been subject to revised shipping plans. In other words, shipping personnel should have recognized products exhibiting this kind of sales information.

4.4. Forecasting Inventory Ratios The cluster analysis identifies products that shipping personnel should recognize as undergoing significant changes in inventory. However, the shipping personnel do not know the optimal time or other factors about the shipment. Therefore, the reference lines obtained by cluster analysis (Figure5) predict the inventory ratio at certain weeks. Changes in inventory that deviate significantly from the basic line are signals that should be transmitted to shipping personnel to highlight important sales information. In this paper, to predict how far the inventory transition will be from the basic line, the conditional probability of the bivariate normal distribution is used for each product. In this way, it is possible to predict the subsequent inventory transition when the change in inventory is recognized until a specific week. The conditional expectation and variance in the conditional probability of the bivariate normal distribution are obtained by the following Equations (2) and (3), respectively.

0 Ii,t − µX µ = µY + ρσY , (2) σx q 0 2 σ = σY 1 − ρ (≤ σY). (3) Based on the above conditional expectation and the conditional variance of the conditional probability, a 95% confidence interval is generated, which is used as the range of prediction. However, since the actual inventory ratio may be smaller than the inventory ratio considering 95% confidence intervals for the conditional expectation obtained by the conditional probability of the bivariate normal distribution, this paper proposes minimum values for each week. Since the inventory ratio cannot be smaller than zero, it is considered to be the minimum. Hence, the upper and lower limits of the range from the inventory ratio considering the 95% confidence interval and the actual inventory ratio are as follows: 0 0 Upper limit : Min µ + 1.96σ , Ii,t , (4) Lower limit : Maxµ0 − 1.96σ0, 0 . (5)

Here, we predict the inventory ratio of product A using information on all clusters of products to be sold in stages, namely, clusters 3, 4, and 5, and for the cluster to which it belongs, 3. The prediction for product A is shown below. The gray line of each ﬁgure is a line connecting the conditional expectation values at each week calculated by the Equation (2). The blue line represents change of the upper limit value in each week obtained from Equation (4). The orange line shows change of the lower limit derived from Equation (5). Finally, in order to compare how the actual inventory ratio changes to changes of gray, blue, orange lines, the change of the actual inventory ratio is indicated by yellow line. Here, the calculation results are shown assuming that the sales period is T = 9. Therefore, based on the assumption, the products will not be sold after nine weeks and will be discarded. Sustainability 2019, 11, 259 9 of 14 Sustainability 2019, 11 FOR PEER REVIEW 9

TheThe shipmentshipment of thethe ﬁrstfirst weekweek isis determineddetermined byby thethe manufacturer’smanufacturer’s sales policy, andand shippersshippers whowho overseeoversee shippingshipping decisionsdecisions beginbegin shippingshipping afterafter thethe secondsecond week.week. Therefore, this paper predicts inventoryinventory levels levels for for all all weeks, weeks, conditional conditional on on the the realized realized level level of sales of sales up to up any to given any givenweek. week.Form FormFigure Figure 6a,b, since6a,b, sincethe shipping the shipping ratio at ratio the atthird the week third weekis large, is large,the range the rangeof prediction of prediction after the after fourth the fourthweek can week be can narrowed be narrowed significan signiﬁcantly.tly. Furthermore, Furthermore, with with the the expected expected value value of of the the conditionalconditional probability,probability, whichwhich isis thethe graygray line,line, atat thethe timetime whenwhen thethe inventoryinventory ratioratio upup toto thethe secondsecond weekweek isis known,known, the inventory ratio of the ninth week has alreadyalready come close to the actual inventory ratio. Therefore,Therefore, inin thethe casecase of of product product A, A, the the inventory inventory ratio ratio that that will will result result in in disposal disposal could could be be predicted predicted at theat the second second week. week. Shipping Shipping personnel personnel who knowwho know this can this then can reduce then reduce the amount the amount of product of disposalproduct bydisposal considering by considering various countermeasures various countermeasures and taking and action. taking action.

1 1

0.8 0.8

0.6 0.6

0.4 0.4 Inventory Ratio 0.2 Inventory Ratio 0.2

0 0 02468 02468 Week Week (a) (b) 1 1

0.8 0.8

0.6 0.6

0.4 0.4 Inventory Ratio 0.2 Inventory Ratio0.2

0 0 02468 02468 Week Week (c) (d)

FigureFigure 6.6. TheThe ExpectationExpectation for ProductProduct A (conditional(conditional on datadata of clustersclusters 3, 4,4, andand 5).5). Note: graph (a) representsrepresents whenwhen the the actual actual inventory inventory ratio ratio is is known known by by the the second second week, week, (b) ( isb) when is when it is it known is known by the by thirdthe third week, week, (c) is (c when) is when it is knownit is known by the by fourth the fourth week, week, and ( dand) is ( whend) is when it is known it is known by the by ﬁfth the week. fifth Theweek. gray The line gray shows line shows the conditional the conditional expectation, expectatio the bluen, the and blue orange and orange lines show lines theshow upper the upper and lower and limitslower oflimits the range,of the respectively,range, respectively, and the and yellow the lineyello showsw line theshows change the inchange the actual in the inventory actual inventory ratio of productratio of product A. A.

In both Figures6 and7, the more weeks in which the inventory ratio is known, the narrower the In both Figures 6 and 7, the more weeks in which the inventory ratio is known, the narrower rangethe range of prediction. of prediction. In theIn the case case of productof product A, A, the the prediction prediction range range is narrowedis narrowed signiﬁcantly significantly in the thirdin the week. third Thisweek. is This because is because the actual the inventoryactual inventory ratio of ratio product of product A is drastically A is drastically reduced reduced in the third week.in the third Moreover, week. when Moreover, predicting when withpredicting only the with cluster only the to which cluster product to which A product belongs, A the belongs, prediction rangethe prediction is narrowed range as ais whole.narrowed However, as a whole. product Howe A isver, shipped product at aA high is shippe rate ind theat a third high week.rate in Then, thethe inventorythird week. ratio Then, at thethe thirdinventory week ratio is much at the smaller third thanweek the is much expected smaller inventory than the ratio expected of the third weekinventory at the ratio second of the week. third Theweek actual at the sales second are week. not known The actual at that sales time, are butnot shippingknown at morethat time, than the expectedbut shipping value more may than result the in expected more waste. value Hence, may result excess in shippingmore waste. can Hence, be reduced excess by shipping shipping the productcan be reduced in the fourth by shipping week after the receivingproduct in more the detailedfourth week sales after information. receiving In more other detailed words, sales since this productinformation. is highly In other likely words, to result since in some this disposal,product is it ishighly possible likely to reduceto result extra in some shipping disposal, by lengthening it is possible to reduce extra shipping by lengthening the period for recognizing sales information.

Sustainability 2019, 11, 259 10 of 14 Sustainability 2019, 11 FOR PEER REVIEW 10

Asthe in period the other for recognizing method, when sales shipping information. at the As third in the week, other it method,is better whento adjust shipping the shipping at the third amountweek, it of is betterproduct to A. adjust For theexample, shipping product amount A is of shipped product at A. proportions For example, according product to A the isshipped basic at lineproportions (gray line according of Figure to 6 theor Figure basic line 7). (gray line of Figure6 or Figure7).

1 1

0.8 0.8

0.6 0.6

0.4 0.4 Inventory Ratio 0.2 Inventory Ratio 0.2

0 0 02468 02468 Week Week (a) (b) 1 1

0.8 0.8

0.6 0.6

0.4 0.4 Inventory Ratio

Inventory Ratio 0.2 0.2

0 0 02468 02468 Week Week (c) (d)

Figure 7. TheThe Conditional Conditional Probability Probability Expectation Expectation of of the the Bivariate Bivariate Normal Normal Distribution Distribution of ofProduct Product A forA for Data Data of ofCluster Cluster 3. 3.Note: Note: graph graph (a ()a represents) represents when when the the actual actual in inventoryventory ratio ratio is is known known by by the second week, week, ( (bb)) is is when when it it is is known known by by the the third third week, week, ( (cc)) is is when when it it is is known known by by the the fourth fourth week, week, and ( d) is is when when it it is is known known by by the the 5th 5th week. week. The The gray gray line line shows shows the the conditional conditional expectation, expectation, the the blue blue and orange lines show the the upper upper an andd lower lower limits limits of of the the range, range, respectively, respectively, and the yellow line shows shows thethe change in the actual inventory ratio of product A.

As a second example, the prediction results for product B are shown in Figures 8 8 andand9 9.. Product Product B is is also also sold sold in in stages, stages, and and the the detailed detailed cluster cluster nu numbermber is is5. 5.As Aswith with product product A, we A, wepredict predict cases cases for whichfor which the actual the actual inventory inventory ratio ratio is known is known in the in se thecond second week. week. The figures The figures show showthat product that product B has Ba highhas a inventory high inventory ratio even ratio at even the ninth at the week. ninth Also, week. in Also, the case in the of caseproduct of product B, the predicted B, the predicted ninth week’s ninth inventoryweek’s inventory ratio at ratio the atsecond the second week weekdoes doesnot diff noter differ much much from from the the prediction prediction at atthe the fifth fifth week. week. Therefore, some some degree degree of of prediction prediction is is possible possible in in the the second second week. week. Furthermore, Furthermore, as ascan can be beseen seen in Figurein Figure 9d,9 d,the the inventory inventory ratio ratio at at the the sixth sixth week week is is lower lower than the expected value,value, suggestingsuggesting thatthat shipments were made to avoid avoid being being left left with with a a large large inventory inventory in in the the logistics logistics warehouse. warehouse. However, However, since the product is unlikely to sell much after being shipped in the sixth week, perhaps it should have been shipped earlier to increase sales. By By sh shippingipping in in this this manner, manner, the product product would have been placed in the store for a longer period, and opportun opportunitiesities to sell at the fixed fixed price would also have increased.increased. This This would would have have reduced the amount of waste due to unsold product. When using the results of prediction, first, the shipping personnel would compare the conditional expectation with the inventory ratio of the product at the present moment. This shows how the inventory ratio of the product changes with respect to the average. Then, the shipping personnel would plan the shipping for the next week. Finally, they would determine the amounts of the shipment while also taking into account the sales policy.

Sustainability 2019, 11, 259 11 of 14 Sustainability 2019, 11 FOR PEER REVIEW 11

1 1

0.8 0.8

0.6 0.6

0.4 0.4 Inventory Ratio 0.2 Inventory Ratio 0.2

0 0 02468 02468 Week Week (a) (b) 1 1

0.8 0.8

0.6 0.6

0.4 0.4 Inventory Ratio Inventory Ratio0.2 0.2

0 0 02468 02468 Week Week (c) (d)

Figure 8. TheFigure Expectation 8. The Expectation of Product of Product B (conditionalB (conditional on on da datata forfor clusters clusters 3, 4, and 3, 4,5).and Note: 5). graph Note: (a) graph (a) represents when the actual inventory ratio is known by the second week, (b) is when it is known by represents when the actual inventory ratio is known by the second week, (b) is when it is known by the the third week, (c) is when it is known by the fourth week, and (d) is when it is known by the fifth third week,week. (c) isThe when gray line it isshows known the conditional by the fourth expectatio week,n, the and blue ( dand) is orange when lines it is show known the upper by the and ﬁfth week. The gray linelower shows limits of the the conditional range, respectively, expectation, and the yello thew blue line andshows orange the change lines in showthe actual the inventory upper and lower limits of theratio range, of product respectively, B. and the yellow line shows the change in the actual inventory ratio of productSustainability B. 2019, 11 FOR PEER REVIEW 12 Also, the results of this prediction lead to different recommendations depending on the type of

product.1 For example, in the case of a standard product,1 manufacturers sell for a longer term, so if the inventory runs out earlier than forecasted, it is possible to produce additional products to meet 0.8 customer0.8 demand. On the other hand, in the case of a product sold for only one season, when the

expected0.6 value of the stock ratio at the ninth week is large,0.6 this is a signal that products are not selling well that should be heeded by shipping personnel. 0.4In QR strategy in the logistics warehouse, product0.4 shipping quantities are decided at the time Inventory Ratio whenInventory Ratio shipping personnel receive sales information on each product during the monitoring period. 0.2 0.2 However, in this paper, we give a signal to shipping personnel that is based on the comparison of previous0 inventory transition for a similar product to the0 current inventory transition for the current product.02468 Therefore, even if there is no arrival of sales information,02468 shipping personnel can receive Week signals about shipping. In addition, this method is effective for all sales periods.Week (a) (b) 1 1

0.8 0.8

0.6 0.6

0.4 0.4 Inventory Ratio Inventory Ratio0.2 0.2

0 0 02468 02468 Week Week (c) (d)

Figure 9. FigureThe Conditional 9. The Conditional Probability Probability Expectation Expectation of ofthe theBivariate Bivariate Normal Normal Distribution Distribution of Product A of Product for Data of Cluster 5. Note: graph (a) represents when the actual inventory ratio is known by the A for Data of Cluster 5. Note: graph (a) represents when the actual inventory ratio is known by the second week, (b) is when it is known by the third week, (c) is when it is known by the fourth week, second week,and (d ()b is) iswhen when it is itknown is known by the byfifth the week. third The week,gray line (c shows) is when the conditional it is known expectation, by the the fourth week, and (d) isblue when and it orange is known lines show by the the ﬁfth upper week. and lower The limit grays of line the showsrange, respectively, the conditional and the expectation, yellow line the blue and orangeshows lines the show change the in the upper actual and inventory lower ratio limits of product of the range,A. respectively, and the yellow line shows the change in the actual inventory ratio of product A. 5. Conclusions In this study, based on the premise of selling at fixed price, we tested an approach for reducing waste from unsold products in the textile and apparel industry and, thereby, mitigating adverse effects on the environment. The purpose of shipping personnel in the study is to recognize relevant data about sales situations and make appropriate judgments (i.e., predictions) about whether to issue alerts. The alerts in this paper are made from the result of data analysis to the product that shipping personnel should watch sales situation carefully. Although there are many kinds of data that can be analyzed in this particular industry, we chose to focus on logistics warehouse data, examining stock points to assess changes in the warehouse inventory ratios. Since the total quantities of apparel products varied considerably, the data was normalized in our study. Furthermore, because it was difficult to monitor the status of each individual product, we categorized products into several groups using cluster analysis. We found that the products whose sales information required the most careful monitoring were products that were sold in stages. For these products, we also forecasted inventory ratios using cluster results, and predictions showed that some products were more likely to remain unsold. Given this, we considered the products whose shipment patterns (i.e., timing and quantities) should be modified. This study has practical value for shipping personnel. By using conditional probability in the manner we have proposed, shipping personnel could accurately predict inventory ratios at the ninth week from changes in the inventory ratios up to the third week. Therefore, at the third week, shipping personnel could take remedial measures for products predicted to have excess inventory in the logistics warehouse.

Sustainability 2019, 11, 259 12 of 14

When using the results of prediction, ﬁrst, the shipping personnel would compare the conditional expectation with the inventory ratio of the product at the present moment. This shows how the inventory ratio of the product changes with respect to the average. Then, the shipping personnel would plan the shipping for the next week. Finally, they would determine the amounts of the shipment while also taking into account the sales policy. Also, the results of this prediction lead to different recommendations depending on the type of product. For example, in the case of a standard product, manufacturers sell for a longer term, so if the inventory runs out earlier than forecasted, it is possible to produce additional products to meet customer demand. On the other hand, in the case of a product sold for only one season, when the expected value of the stock ratio at the ninth week is large, this is a signal that products are not selling well that should be heeded by shipping personnel. In QR strategy in the logistics warehouse, product shipping quantities are decided at the time when shipping personnel receive sales information on each product during the monitoring period. However, in this paper, we give a signal to shipping personnel that is based on the comparison of previous inventory transition for a similar product to the current inventory transition for the current product. Therefore, even if there is no arrival of sales information, shipping personnel can receive signals about shipping. In addition, this method is effective for all sales periods.

5. Conclusions In this study, based on the premise of selling at fixed price, we tested an approach for reducing waste from unsold products in the textile and apparel industry and, thereby, mitigating adverse effects on the environment. The purpose of shipping personnel in the study is to recognize relevant data about sales situations and make appropriate judgments (i.e., predictions) about whether to issue alerts. The alerts in this paper are made from the result of data analysis to the product that shipping personnel should watch sales situation carefully. Although there are many kinds of data that can be analyzed in this particular industry, we chose to focus on logistics warehouse data, examining stock points to assess changes in the warehouse inventory ratios. Since the total quantities of apparel products varied considerably, the data was normalized in our study. Furthermore, because it was difficult to monitor the status of each individual product, we categorized products into several groups using cluster analysis. We found that the products whose sales information required the most careful monitoring were products that were sold in stages. For these products, we also forecasted inventory ratios using cluster results, and predictions showed that some products were more likely to remain unsold. Given this, we considered the products whose shipment patterns (i.e., timing and quantities) should be modified. This study has practical value for shipping personnel. By using conditional probability in the manner we have proposed, shipping personnel could accurately predict inventory ratios at the ninth week from changes in the inventory ratios up to the third week. Therefore, at the third week, shipping personnel could take remedial measures for products predicted to have excess inventory in the logistics warehouse. In light of the trade-off between the risk of disposal from unsold product and the risk of opportunity loss from product shortage, future research should investigate optimal shipment timing when disposal and shortage are simultaneously taken into consideration. Also, the optimal shipping quantities should be obtained at the optimum shipment timing. In practice, however, the timing or quantities of products shortage cannot be obtained as data. Therefore, under the assumption of a product shortage situation, the risk of opportunity loss will be evaluated. In addition, future research should verify the effectiveness of the proposed method by performing analyses in several different scenarios. In sum, using past data as well as current sales data can allow for better detection of market trends. Although the larger the amount of data, the easier it is to detect market trends in past and present, to do so, many textile and apparel companies are needed to share and analyze data. Apparel and textile companies can use our findings to improve the timing and quantity of shipments, Sustainability 2019, 11, 259 13 of 14 more effectively manage supply chains, and protect profits. In so doing, they can also reduce the burden of disposal, minimize adverse effects on the environment, and contribute to the development of more sustainable societies.

Author Contributions: R.T. performed the research and wrote the manuscript. A.I. coordinated and supervised the overall research. T.S. assisted in developing mathematical methods. M.H. and W.K. acquired funds and collected data. All authors read and approved the final manuscript. Conflicts of Interest: The authors declare no conflict of interest.

References

1. Seuring, S.; Müller, M. From a literature review to a conceptual framework for sustainable supply chain management. J. Clean. Prod. 2008, 16, 1699–1710. [CrossRef] 2. Baumgartner, R.J.; Ebner, D. Corporate sustainability strategies: Sustainability proﬁles and maturity levels. Sustain. Dev. 2010, 18, 76–89. [CrossRef] 3. Norman, W.; MacDonald, C. Getting to the bottom of “triple bottom line”. Bus. Ethics Q. 2004, 14, 243–262. [CrossRef] 4. Ahi, P.; Searcy, C. Measuring social issues in sustainable supply chains. Meas. Bus. Excell. 2015, 19, 33–45. [CrossRef] 5. Markley, M.J.; Davis, L. Exploring future competitive advantage through sustainable supply chains. Int. J. Phys. Distrib. Logist. Manag. 2007, 37, 763–774. [CrossRef] 6. Bruce, M.; Daly, L.; Towers, N. Lean or agile: A solution for supply chain management in the textiles and clothing industry? Int. J. Oper. Prod. Manag. 2004, 24, 151–170. [CrossRef] 7. Iyer, A.V.; Bergen, M.E. Quick response in manufacturer-retailer channels. Manag. Sci. 1997, 43, 559–570. [CrossRef] 8. Tanaka, R.; Ishigaki, A.; Suzuki, T.; Hamada, M.; Kawai, W. Shipping plan for apparel products using shipping record and just-in-time inventory at a logistics warehouse. In Proceedings of the 7th International Congress on Advanced Applied Informatics, Yonago, Tottori, Japan, 8–13 July 2018. 9. Kim, H.S. A bayesian analysis on the effect of multiple supply options in a quick response environment. Nav. Res. Logist. (NRL) 2003, 50, 937–952. [CrossRef] 10. Liu, N.; Ren, S.; Choi, T.M.; Hui, C.L.; Ng, S.F. Sales forecasting for fashion retailing service industry: A review. Math. Probl. Eng. 2013, 2013, 1–9. [CrossRef] 11. Mostard, J.; Teunter, R.; De Koster, R. Forecasting demand for single-period products: A case study in the apparel industry. Eur. J. Oper. Res. 2011, 211, 139–147. [CrossRef] 12. Gurnani, H.; Tang, C.S. Note: Optimal ordering decisions with uncertain cost and demand forecast updating. Manag. Sci. 1999, 45, 1456–1462. [CrossRef] 13. Moore, S.B.; Ausley, L.W. Systems thinking and green chemistry in the textile industry: Concepts, technologies and beneﬁts. J. Clean. Prod. 2004, 12, 585–601. [CrossRef] 14. Larney, M.; Van Aardt, A.M. Case study: Apparel industry waste management: A focus on recycling in South Africa. Waste Manag. Res. 2010, 28, 36–43. [CrossRef][PubMed] 15. Briga-Sa, A.; Nascimento, D.; Teixeira, N.; Pinto, J.; Caldeira, F.; Varum, H.; Paiva, A. Textile waste as an alternative thermal insulation building material solution. Constr. Build. Mater. 2013, 38, 155–160. [CrossRef] 16. Tseng, M.L.; Tan, R.R.; Siriban-Manalang, A.B. Sustainable consumption and production for Asia: Sustainability through green design and practice. J. Clean. Prod. 2013, 40, 1–5. [CrossRef] 17. Kuo, T.C.; Hsu, C.W.; Huang, S.H.; Gong, D.C. Data sharing: A collaborative model for a green textile/clothing supply chain. Int. J. Comput. Integr. Manuf. 2014, 27, 266–280. [CrossRef] 18. Wang, F.; Zhuo, X.; Niu, B. Sustainability analysis and buy-back coordination in a fashion supply chain with price competition and demand uncertainty. Sustainability 2016, 9, 25. [CrossRef] 19. Caniato, F.; Caridi, M.; Crippa, L.; Moretto, A. Environmental sustainability in fashion supply chains: An exploratory case based research. Int. J. Prod. Econ. 2012, 135, 659–670. [CrossRef] 20. Choi, T.M.; Chiu, C.H. Mean-downside-risk and mean-variance newsvendor models: Implications for sustainable fashion retailing. Int. J. Prod. Econ. 2012, 135, 552–560. [CrossRef] Sustainability 2019, 11, 259 14 of 14

21. Brun, A.; Castelli, C. Supply chain strategy in the fashion industry: Developing a portfolio model depending on product, retail channel and brand. Int. J. Prod. Econ. 2008, 116, 169–181. [CrossRef] 22. Ciarnienˇ e,˙ R.; Vienažindiene,˙ M. Management of contemporary fashion industry: Characteristics and challenges. Procedia-Soc. Behav. Sci. 2014, 156, 63–68. [CrossRef] 23. Mehrjoo, M.; Pasek, Z.J. Risk assessment for the supply chain of fast fashion apparel industry: A system dynamics framework. Int. J. Prod. Res. 2016, 54, 28–48. [CrossRef] 24. Patil, R.; Avittathur, B.; Shah, J. Supply chain strategies based on recourse model for very short life cycle products. Int. J. Prod. Econ. 2010, 128, 3–10. [CrossRef] 25. Choi, T.M.; Sethi, S. Innovative quick response programs: A review. Int. J. Prod. Econ. 2010, 127, 1–12. [CrossRef] 26. Fisher, M.; Raman, A. Reducing the cost of demand uncertainty through accurate response to early sales. Oper. Res. 1996, 44, 87–99. [CrossRef] 27. Choi, T.M. Quick response in fashion supply chains with retailers having boundedly rational managers. Int. Trans. Oper. Res. 2017, 24, 891–905. [CrossRef] 28. Choi, T.M.; Zhang, J.; Cheng, T.E. Quick response in supply chains with stochastically risk sensitive retailers. Decis. Sci. 2018, 49, 932–957. [CrossRef] 29. Tsukagoshi, Y.; Ishigaki, A.; Sandoh, H. Effective Quick Response in a Supply Chain. In Proceedings of the 17th Asia Paciﬁc Industrial Engineering and Management Systems, Taipei, Taiwan, 7–10 December 2016. 30. Sun, Z.L.; Choi, T.M.; Au, K.F.; Yu, Y. Sales forecasting using extreme learning machine with applications in fashion retailing. Decis. Support Syst. 2008, 46, 411–419. [CrossRef] 31. Thomassey, S. Sales forecasts in clothing industry: The key success factor of the supply chain management. Int. J. Prod. Econ. 2010, 128, 470–483. [CrossRef] 32. Fan, Z.P.; Che, Y.J.; Chen, Z.Y. Product sales forecasting using online reviews and historical sales data: A method combining the Bass model and sentiment analysis. J. Bus. Res. 2017, 74, 90–100. [CrossRef] 33. Frank, C.; Garg, A.; Sztandera, L.; Raheja, A. Forecasting women’s apparel sales using mathematical modeling. Int. J. Cloth. Sci. Technol. 2003, 15, 107–125. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).