CASE STUDY

EndToEndSupplyChainStrategies.pdf

Home >Business & Finance homework help >Operations Management homework help >CASE STUDY

End-To-End Supply Chain Strategies: A Parametric Study of the Apparel Industry

Shardul S. Phadnis* Malaysia Institute for Supply Chain Innovation, 2A, Persiaran Tebar Layar, Seksyen U8, Bukit Jelutong, Shah Alam, 40150, Selangor,

Malaysia, [email protected]

Charles H. Fine MIT Sloan School of Management, 77 Massachusetts Ave, E62-466, Cambridge, Massachusetts 02139, USA Asia School of Business, Sasana Kijang, 2 Jalan Dato Onn, 50480, Kuala Lumpur, Malaysia, [email protected]

T his study examines the tradeoffs in sourcing and sales strategies (i.e., upstream and downstream supply chain strate- gies) by considering them as components of an integral end-to-end supply chain strategy. We evaluate four end-to-

end supply chain strategies under various scenarios using a newsvendor model, and compare the model’s predictions against the prescriptions in Fisher’s (1997) framework, which recommends “cost-efficient” supply chains for “functional” products and “responsive” supply chains for “innovative” products. We considered combinations of offshore vs. near- shore sourcing, and online vs. brick-and-mortar retailing. This study’s key finding is that sourcing and sales strategies are not completely modular: an integral end-to-end strategy may not decompose into an optimal sourcing strategy and a sep- arately computed optimal sales strategy. Our analyses sharpen strategic supply chain thinking by identifying realistic con- ditions in which an end-to-end strategy with cost-efficient components could outperform one with responsive components for innovative products, or when one with responsive components could be more profitable than one with cost-efficient components for functional products.

Key words: supply chain strategy; sourcing strategy; distribution strategy History: Received: June 2016; Accepted: July 2017 by Edward Anderson, after 2 revisions.

1. Introduction

The twin forces of globalization and e-commerce have dramatically changed supply chains in the last two decades. Globalization has provided firms a choice in their sourcing and operations strategies to either pro- duce close to their markets or seek lower costs from distant sources. The growth of the Internet has created powerful new sales channels, where firms can sell goods directly to consumers online instead of through brick-and-mortar stores. However, firms around the world still pursue a range of sourcing and sales strate- gies, so that one cannot (yet) assert the emergence of a dominant design for the 21st century’s supply chains. Nonetheless, firms in Western countries have out-

sourced a large share of production to lower-cost countries in Asia. A McKinsey study (Manyika et al. 2014) showed that one-third of all goods produced in 2012 (36% of global GDP) crossed national borders. The United States alone imported $2.3 trillion worth of goods for local consumption. On the other hand, some Western firms have started bringing some of their offshored production back into the home coun- try. In a Boston Consulting Group (2015) survey, 31%

of senior manufacturing executives from firms with at least $1 billion in annual revenue reported that their companies were most likely to add production capac- ity in the United States in the next five years. On the sales side, e-commerce has grown as an

alternative to brick-and-mortar stores. The $395 bil- lion e-commerce sales in the United States constituted 8.1% of total retail sales in 2016. From 2015 to 2016, the U.S. e-commerce sales grew by 15.1%, while retail sales increased by 2.9% (U.S. Census Bureau 2017). This trend is projected to continue, with e-commerce expected to account for 11% of all U.S. retail sales by 2018 (Forrester Research 2014). Although growth in e- commerce channel has been linked to the demise of established brick-and-mortar stores such as Borders Book Stores and Circuit City, this trend does not spell absolute doom for the brick-and-mortar sales channel. In fact, Amazon.com, a pioneer of online retailing, opened its first physical bookstore in 2015, and announced the eighth bookstore to open in New York City in Spring 2017 (New York Times 2017). The plurality of supply chain strategies can also be

seen within numerous industries. The global apparel industry provides one example of this apparent state

2305

of experimentation in end-to-end supply chain strate- gies. This industry is a posterchild of globalization. In 2016, apparel manufacturers had estimated revenue of $657 billion, 70% of which was generated from exported goods (IBISWorld 2016). International Appa- rel Federation (2017), the leading federation of apparel manufacturers, retailers, and related companies, boasts a membership of 150,000 companies employing over 5 million people in nearly 40 countries. Despite its size, no dominant supply chain design has yet emerged for the apparel industry. Equally successful apparel brands in the West have either outsourced the production to low-cost contract manufacturers in Asia, or tightly integrated with producers located close to the market (e.g., Ghemawat and Nueno 2003, Pisano and Adams 2009). Similarly, numerous clothing brands successfully use brick-and-mortar stores as the primary sales channel (e.g., Zara) while others sell only online (e.g., Amazon). With $63.8 billion online sales, apparel is the largest category of e-commerce sales in the United States (eMarketer 2017). This plu- rality of supply chain configurations observed in prac- tice raises a question: under what circumstances is a particular end-to-end supply chain configuration more profitable than the others? The supply chain literature prescribes ideal supply

chain strategies based on different product and firm attributes. Fisher (1997) suggests choosing a supply chain strategy according to the predictability of the product’s demand: cost-efficient supply chains for products with predictable demand (“functional”) and responsive supply chains for those with unpre- dictable demand (“innovative”). Analytical and empirical examinations partially support Fisher’s guidelines (Li and O’Brien 2001, Selldin and Olhager 2007). The general guidelines help derive specific rec- ommendations for sourcing and sales strategies. An offshore source has lower out-of-pocket costs, but may suffer from higher forecast error and be less responsive due to longer lead time than a nearshore source (de Treville and Trigeorgis 2010, Wu and Zhang 2014). On the sales side, the online channel has lower costs than the brick-and-mortar channel as it has fewer physical assets and centralizes inventory in one location. However, this comes at the expense of a lack of pre-purchase product experience for the con- sumer and results in a higher proportion of returned sales for the online channel (WSJ 2013). Pre-purchase experience may trump lower cost of online channel for novel products, but not for popular products (Balakrishnan et al. 2014, Brynjolfsson et al. 2009). Although the tradeoffs among different sourcing

and sales strategies have been studied in the opera- tions and supply chain management literature, the two strategies have not been studied together as com- ponents of one integral supply chain strategy. This

lack of a systemic approach may be due to the associa- tion of the two strategic decisions with different func- tions in firms or academic disciplines, or due to the academic preference for parsimonious models of managerial decisions. Regardless, a joint exploration of the sourcing and sales strategies is likely to provide novel insights if the two possess a high degree of interdependence (Larsen et al. 2013). The sourcing and sales strategies may have some complementarity, because the firms that choose holistic supply chain strategies (e.g., Zara) seem to outperform their com- petition (Ghemawat and Nueno 2003). This study explores these complementarities by eval-

uating sourcing and sales strategies together, which we call the firm’s end-to-end supply chain strategy. Our model assumes that a firm chooses its end-to-end sup- ply chain strategy, and makes two operational deci- sions in each season: decide the quantity to procure and allocate the procured quantity to store(s). The firm sources either from an offshore or a nearshore supplier, and sells through either an online store or a number of brick-and-mortar stores. We compare the expected profits of four end-to-end supply chain configurations under various scenarios defined by eight factors: con- tribution margin, cost advantage of offshore supplier, forecast accuracy, change in forecast accuracy over time, variation in market demand, number of brick- and-mortar stores, product returns rate, and the online returns penalty. We evaluate the strategies using a numerical exercise and codify the findings in six obser- vations. We present industry executives’ critique of the observations, and conclude by discussing the implica- tions of the study’s findings for theory and practice.

2. Literature Review

Fisher’s (1997) seminal paper argued that products could be categorized as functional or innovative based on the predictability of their demand and rec- ommended that they be served by efficient or respon- sive supply chains, respectively. The efficient supply chains would seek to meet “demand efficiently at the lowest possible cost” through means such as selecting suppliers “primarily for cost and quality” (e.g., using low-cost offshore suppliers), shortening “lead time as long as it doesn’t increase cost,” minimizing “inven- tory throughout the chain” (e.g., by centralizing inventory storage as in the online channel), and so on (ibid, p. 108). Conversely, responsive supply chains would seek to “minimize stockouts, forced mark- downs, and obsolete inventory” through means such as selecting suppliers “for speed, flexibility, and qual- ity” investing “aggressively in ways to reduce lead time,” (e.g., by sourcing nearshore), deploying “sig- nificant buffer stocks of parts or finished goods” (e.g., through brick-and-mortar stores), and so on.

Tests of Fisher’s propositions have produced mixed results. A survey of 128 manufacturing companies, using self-reported performance measures, partially supported Fisher’s dichotomy of products and supply chain strategies: it found that companies using respon- sive supply chains to deliver innovative (functional) products considered their cost performance superior to (worse than) their competitors (Selldin and Olhager 2007). However, the study did not find support for Fisher’s propositions related to cost efficient supply chains. Li and O’Brien’s (2001) test of Fisher’s proposi- tions using an optimization model suggests that the ideal supply chain strategy for products with different levels of demand uncertainty is influenced by condi- tions not considered in Fisher’s model, such as uncer- tainty of material demand, product profit margin, etc. Calvo and Martinez-de-Albeniz (2016) show that for innovative products, contrary to Fisher’s argument, sole sourcing with up-front price commitment is preferable to the more flexible option of dual sourcing because it enforces stronger competition between the suppliers. Broadly, at least three different streams can be iden-

tified in the large body of OM literature examining tradeoffs in sourcing (upstream supply chain) deci- sions. One stream studies supplier selection using mod- els to explore tradeoffs between single and multiple suppliers. A common reason for using multiple sup- pliers is to mitigate the risk of supply disruption due to unpredictable events such as accidents or natural disasters (Treleven and Schweikhart 1988). Dual sour- cing can be optimal when suppliers face positive entry costs (Klotz and Chatterjee 1995). It can also provide better service levels than sole sourcing, except when order costs are high and lead time vari- ability is low (Chiang and Benton 1994). On the other hand, single sourcing with up-front price commit- ment enforces stronger competition and is superior to dual sourcing for procuring products with short life cycles from suppliers choosing prices endogenously (Calvo and Martinez-de-Albeniz 2016). For both sole and dual sourcing, Li (2013) describes when it is opti- mal to commit price ex ante and when it is better to negotiate ex post. A large body of literature studies the properties of

optimal procurement decision in presence of demand uncer- tainty, using the newsvendor model. Several extensions of the basic newsvendor model—such as, the case of price-setting retailer, retailer promotions, retailer com- petition, etc.—are studied in this stream (e.g., Bernstein and Federgruen 2005, Dana and Spier 2001, Taylor 2002). Some recent works extend these models by incorporating the effects of competition for either the buyer or the supplier (Li 2013, Wu and Zhang 2014). A related stream of literature explores mathematical properties of optimal procurement quantities from multiple suppliers (Zhang et al. 2012).

A third stream of research on sourcing strategies examines the strategic reasons and implications of out- sourcing for the buyer firm with a descriptive lens (e.g., Larsen et al. 2013, Reitzig and Wagner 2010). The works in this stream examine the implications of outsourcing in terms of cost, quality, responsiveness, R&D productivity, and so on. One finding from this stream of particular interest to the present work is that the benefits of offshoring, expected to result from the lower cost of goods, may be overestimated in practice due to the difficulty of understanding the interdependence between the tasks offshored and those retained in-house (Larsen et al. 2013). A large body of OM research also examines sales

(downstream supply chain) strategies. One stream of work studies interactions between two sales channels of one firm – one on-line and one in-store. Lenient returns poli- cies that encourage online purchases in the absence of pre-purchase product experience, yield returns rates that are higher for online channels (average 18–35%, depending on product category) than the average retail returns rate of 8.7% (Ofek et al. 2011). Generous returns policies also can signal higher product quality and encourage online purchase (Wood 2001). In multi- channel settings, policies that allow customers to Buy a product Online and Pick it up at the retailer’s brick-and- mortar Store (BOPS) enable the retailer to reach new customers (Gao and Su 2017). The BOPS strategy can also increase the contribution of lowest selling products to the total sales (Gallino et al. 2017). Another stream of literature examines competition

between brick-and-mortar and online retailers. Balasubra- manian (1998) shows that for products well-suited to the online channel (i.e., reliable products with low need for immediate gratification), an online retailer can obtain higher returns by increasing market pres- ence and engaging in greater competition with the brick-and-mortar retailers; conversely, for products poorly suited for the online channel, the online retail- er should lower its market presence and allow compe- tition among brick-and-mortar retailers. Balakrishnan et al. (2014) show that the browse-and-switch behavior —where some customers examine a product in a brick-and-mortar store but buy it from an online retai- ler—is feasible in equilibrium, and it reduces profits for both retailers. Empirical evidence (Brynjolfsson et al. 2009) shows that the competition between brick- and-mortar and online channels is limited to popular products, and an increase in the density of brick-and- mortar stores reduces the demand for popular products from the remote channels. The third stream of literature examines the competi-

tive dynamics between a brick-and-mortar retailer and its manufacturer that introduces its own online sales channel. Under several scenarios, the manufacturer and its brick-and-mortar retailer can both earn higher profits;

this results from a decrease in double-marginalization as manufacturer drops the wholesale price to ensure that the traditional retailer’s demand for its product is not diminished (Arya et al. 2007). This can occur even when sales in the manufacturer’s direct channel are zero (Chiang et al. 2003). Any selfish cost-reducing investments made by the manufacturer could also spill over to reduce the wholesale price paid by the retailer (Yoon 2016). However, manufacturer’s encroachment in retail can lead to loss of profit for both parties, if the retailer’s selling process is more efficient and the prior probability of a large market is low (Li et al. 2014). A common feature of the modeling in all three

streams of this literature is the use of consumer choice models to derive demands at the competing channels of one or multiple firms, endogenously. In contrast, our study assumes that the firm faces an exogenous stochastic demand. We make this choice deliberately. Our model examines four end-to-end strategies of one firm, with only one sales channel (either brick-and- mortar or online). Thus, no inter-channel competitive dynamics are present in our model. This simplifica- tion allows us to draw insights despite the presence of additional complexity introduced by the joint consid- eration of the sourcing and sales strategies. In summary, a review of the vast literature on the

sourcing and sales strategies shows that upstream and downstream supply chain strategies have largely been studied in isolation. The consideration of sales strategy in the sourcing studies is generally limited to the dimensions of price or demand; and the costs and uncertainties associated with physical distribution in the sales channel are not typically considered in the sourcing models. Similarly, the studies of sales strate- gies focus on describing optimal pricing and return policies without incorporating costs and uncertainties encountered in making the sourcing decisions. Against this backdrop, the present study seeks to examine the tradeoffs between different sourcing and retail strategies by considering them jointly as the firm’s integral end-to-end supply chain strategy (Fine 2000, Fisher 1997).

3. Model

This section presents the model of one firm’s end-to- end supply chain strategy choices. The modeling choices are justified with the descriptions of industry practices, obtained through industry publications and interviews with eight executives in six firms (see Table A1 in Appendix S1 for descriptions of the exec- utives interviewed). The firm may source from either an offshore or a nearshore supplier. The key tradeoffs between the two are cost efficiency and responsive- ness. The offshore supplier offers lower total landed cost, whereas the nearshore supplier allows ordering

closer to the selling season when the demand forecast is likely to be more accurate. The firm may sell the products through either brick-and-mortar stores or via an online store. The brick-and-mortar channel “provides quicker gratification and offers an opportu- nity for physical inspection” (Balasubramanian 1998, p. 183). Retail presence makes the firm more respon- sive by deploying “significant buffer stocks of [. . .] finished goods” in order “to reduce lead time” (Fisher 1997, p. 108) between a customer’s order and receipt of goods. Conversely, the online channel is more cost- efficient as it allows pooling of inventory in a few locations, which provides cost savings under a wide range of demand distributions (Berman et al. 2011). To prevent the lack of tactile pre-purchase product experience from deterring shopping, online retailers offer lenient returns policies (Wood 2001) and experi- ence a higher returns rate than their brick-and-mortar competitors (Ofek et al. 2011). Our model assumes that a firm chooses to source

from only one type of supplier (offshore or nearshore) and sells through only one type of channel (online or brick-and-mortar). We do not consider hybrid sour- cing or sales strategies. This assumption allows us to explore the intricate interdependence between sour- cing and sales decisions to draw insights about the end-to-end strategy, without making the model unwieldy. Such simplifying assumptions are com- monly made in analytical models (Wu and Zhang 2014). Furthermore, the sourcing and sales strategies modeled in this manner do exist in the real world. Sole-sourcing is often practiced to minimize the costs associated with developing and maintaining multiple sources, and is superior to dual sourcing when order costs are high or lead time is less variable (Chiang and Benton 1994). R1, the ex-managing director of Aretha (pseudonym, a major global retailer with a few thousand stores), noted that Aretha uses either an off- shore or a nearshore source, but not both, for a given product. He added, “Producing in proximity is not same as producing offshore;” Aretha produces high- fashion items nearshore and cost-sensitive items off- shore. A recent study of chief procurement officers of “leading apparel companies, responsible for a com- bined €28 billion in purchasing volume” shows that apparel companies are beginning to engage more dee- ply with their suppliers to improve productivity and safety (Berg and Hedrich 2014). As a result, they may reduce the number of suppliers. On the sales side, a McKinsey survey of 3000 con-

sumers’ purchases at 17 apparel retailers in the United Kingdom found that only 7% of the purchases at a given retailer were made using both offline and online channels, prompting the authors to conclude that “to- day’s world is still on/offline and not yet omni” (Berg et al. 2015, p. 4). S1, Executive Vice President at GN

Sourcing (pseudonym, sourcing agent used by several brands, has a few thousand suppliers), informed that most of her retail customers “don’t do anything online” and that brands with multiple channels still operate as brick-and-mortar retailers used to. R3, senior manager of global supply chain strategy at Pan- theon (pseudonym, a major global retailer with over a thousand stores), noted that her company set up its online channel as a separate legal entity and could not share inventory with its brick-and-mortar channel; it also had separate inventory planners for the two chan- nels until their recent integration. M1, President & CEO of Lamina (pseudonym, virtual manufacturer, helps brands design products) and M2, Senior Vice President at Lamina, noted in separate interviews that sales of online retailers and online stores of traditional brick-and-mortar retailers are still small, and some of the brick-and-mortar retailers value their online chan- nel more for its ability to predict sales at the stores than for its revenue. R2, co-founder of Ote (pseudonym, ten- year old online retailer), noted that her firm uses its brick-and-mortar stores primarily as showrooms for customers to try the products before making the purchase online. On the other hand, Amazon.com— predicted to be the largest U.S. apparel retailer by 2017— sells clothing only online (Weinswig 2017). Thus, the online and brick-and-mortar channels are still largely independent with regards to apparel sale and inven- tory positioning.

3.1. Model Description Most apparel retailers still plan for seasons. M3, the vice president of the supply chain at Serene (pseudo- nym, global brand of intimate apparel, owns manufac- turing) noted that even its functional products (e.g., undergarments) have two primary selling seasons in the United States: back-to-school (July) and winter holidays (December). On this basis and following the tradition in the OM literature to build on the newsvendor model, we consider a firm that operates in an industry with distinct selling seasons with known beginnings and ends. Before each season, the firm makes a strategic decision about its end-to-end supply chain regarding the source and the sales chan- nel for each product. After choosing its end-to-end strategy (time t0), the firm has one opportunity to order the merchandise to meet the season’s demand: it buys the merchandise either at time t1 (if it chooses an offshore supplier) or at t2 (near-shore supplier). The goods are received at the beginning of the selling season and allocated to the firm’s N stores (t3; N = 1 for online). Finally, the demand is realized (t4); the firm sells the product for price p. A portion q of the sold products are returned for a full refund at the end of the season, with different returns rates for the two channels: qB for brick-and-mortar and qO for online.

At the end of the season, the firm realizes revenue. Figure 1 shows the timeline. If the demand exceeds the purchased quantity, the firm does not have another opportunity to buy additional product and any future demand in the season is not converted into sales. Conversely, if the demand is lower than the procured quantity, the firm has to sell the leftover product for a salvage price s. Let, cx and cm be the total landed costs for the offshore and nearshore suppliers, respectively. We assume, s ≤ cx ≤ cm ≤ p.

3.1.1. Apparel Demand Forecasting. Respondent M2 highlighted two types of apparel forecasting: style forecasting and quantity forecasting. He informed that Lamina uses a proprietary database and algorithms to forecast styles about one year in advance. Lamina’s customer then uses historical sales data to forecast quantity for each item and order the product eight months before the selling season. R1 informed that a team of forecasters in Aretha’s headquarters makes the initial forecast for the season’s demand for each product and the forecasts undergo several changes over time. S1 and R1 both noted that the initial quan- tity forecasts were based on the historic store and aggregate sales data as well as input about fashion trends from store managers and fashion experts. S2, an Executive Director (Fashion) at GN Sourcing, corrobo- rated the retailers’ use of detailed store-level sales data for developing demand forecasts. Retailers are gener- ally unable to change the order placed with the suppli- ers. S2 noted that once a retailer commits to an order quantity, GN Sourcing will buy the raw material for its manufacturers, get the garments made, ship the ordered quantity, and expect to be paid for it. M2 noted that once its retail customers commit to the order quantity, Lamina buys the necessary raw mate- rial and expects the customer to pay for it. This forecast and sourcing process can be modeled

as the additive martingale model of forecast evolu- tion, or a-MMFE (Graves et al. 1986, Heath and Jack- son 1994). The model assumes that the information available to predict a variable grows over time, and the change in its forecast at a given time is uncorre- lated with the information available at the time of the previous change in forecast. These assumptions are valid for the apparel industry where forecasts of the season’s demand are updated based on sales and

Firm chooses its end- to-end strategy

Firm orders goods from offshore supplier at cost

per unit, according to forecast 1

Selling season Price

≤ ≤ ≤

Firm receives and allocates goods to a warehouse for online store ( = 1) or each brick-and-mortar store ( ∈ {10,25})

A portion ( or ) of sales returned by cust- omers; salvaged for 0 ≤ ≤ ≤ 1

Firm orders goods from nearshore supplier at cost

per unit, according to forecast 2

Figure 1 Timeline of Decisions and Events

fashion trends observed, and the forecasts are likely to be conditional expectations of demand given the information available when making the forecast. In fact, “almost all of the existing forecasting techniques are based on assumptions” of a-MMFE (Heath and Jackson 1994, p. 22). Let, Fi be the forecast of the season’s market

demand made at time ti. The initial forecast of the sea- son’s demand, F0, is made based on historic sales and input from fashion experts. Let, l be the expected demand at t0, i.e., E F0½ � ¼ l: The forecast is revised over time as new information becomes available. Let di be the change in the forecast of the season’s demand between times ti-1 and ti, such that Fi = Fi-1 + di, for i = {1, 2, 3, 4}. The forecast updates, Fi, are random variables before time ti with E di½ � ¼ 0 (i.e., unbiased forecast) and VarðdiÞ ¼ Nr2i . Let, F4 = Y be the market demand realized in the season. Assuming a stationary demand process with E Y½ � ¼ l and Var Yð Þ ¼ r2Y ¼ Nr24, the error in the forecast of market demand made at time t1 (i.e., time of confirming the order quantity from the offshore supplier), is E1 = Y � F1. Note, Y ¼ F4 ¼ F3 þ d4 ¼ . . . ¼ F1þðð d2Þ þ d3Þ þ d4. Therefore, the forecast error at t1 is E1 = d2 + d3 + d4. Further, E E1½ � ¼ E d2 þ d3 þ d4½ � ¼ 0 and Var E1ð Þ ¼ Var d2 þ d3 þ d4ð Þ ¼ N r22 þ r23 þ r24

� � . Similarly, error

in the forecast of the season’s demand made at time t2 (i.e., when procuring from a nearshore supplier) has average 0 and variance of N r23 þ r24

� � .

3.1.2 Apparel Sales. At time t0, the firm chooses whether to sell the product through N brick-and-mor- tar stores or an online store (N = 1). For simplicity, we assume that the retail price (p) and the salvage price (s) are identical in the two channels. Since the purpose of this model is to understand the tradeoffs among differ- ent end-to-end supply chain strategies of one firm, we take price and demand as given, instead of making them endogenous to the model. Without loss of gener- ality, we assume that the season’s demands faced by the stores are identically and independently dis- tributed. Therefore, the individual store demand, X,

has expectation E X½ �¼E Y½ � N

¼ l N and variance r2X ¼

r2 Y

N ¼r24.

After defining its end-to-end in strategy, the firm makes two operational decisions every season. First, it decides the quantity of the product to procure from the chosen supplier to satisfy the market demand for the entire season. We assume the firm considers the deterministic product returns rate and uses the newsvendor formulation to determine the purchase quantity. The optimal quantity to buy at time ti (at unit cost ci) for the season is Qi ¼ l þ zðwÞ � rðiÞ;where z(w) is the safety factor corresponding with the newsvendor critical ratio w and r(i) is the standard deviation of error in the forecast of Y made at time ti.

Let, LðqÞ be the expected lost sales when the demand exceeds the quantity available for sale, q.

LEMMA 1. The maximum expected profit for a season from selling goods procured at time ti, where Qi is the quantity procured for the season for per unit cost ci, after all potential returned products are received and salvaged by the firm is given by the following expression.

E P½ � ¼ 1 � qð Þ p � sð Þ l � L Qið Þð Þ � ci � sð ÞQi: ð1Þ

See proof in Appendix S1. For a given end-to-end strategy, the optimal quantity procured from the cho- sen supplier is determined by the supplier’s total landed cost, distribution of errors in the demand fore- casted at the time of procurement, and the product returns rate for the sales channel.

LEMMA 2. The newsvendor critical ratio (w) for products procured for cost c, sold at price p, and salvaged for s when a proportion q of the sold products are returned by customers for full refund is given by

w ¼ 1 1�q

� � p�c p�s

� � � q

1�q � �

. An end-to-end supply chain

strategy is feasible only if p�c p�s � q.

See proof in Appendix S1. Let, k ¼ p�c p�s. Therefore,

w ¼ k�q 1�q. Observe that for 0 < q < k < 1, w ¼

k�q 1�q \k.

Thus, the newsvendor critical ratio given a determinis- tic product returns rate (w) is smaller than the stan- dard newsvendor critical ratio (k). As w is influenced by the product returns rate, sales channels with differ- ent returns rates would have different critical ratios, even when the products are procured, sold, and sal- vaged for identical prices in the two channels. This is one manner in which the sales and sourcing strategies are related to each other. If it has a feasible end-to-end strategy, the firm procures quantity Q* that maximizes the expected profit. Let, Q� ¼ Q�xC if the firm chooses an offshore supplier, and Q� ¼ Q�mC if it chooses a near- shore supplier; the subscript C denotes the sales chan- nel, whose returns rate affects the quantity procured.

LEMMA 3. The optimal quantity procured from the off- shore and the nearshore suppliers for sale in channel C, with newsvendor critical ratio wC, is:

Q�xC ¼ l þ z wCð Þ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi N r22 þ r23 þ r24 � �q

; ð2Þ

Q�mC ¼ l þ z wCð Þ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi N r23 þ r24 � �q

: ð3Þ

See proof in Appendix S1. The demand at each store is assumed to be identically distributed and independent of the demand at the other stores.

Therefore, it is optimal to allocate the procured goods among the stores evenly. The quantity available for sale at each store is Q�/N (For Online: N = 1).

3.2. End-to-end Supply Chain Strategies The two sourcing options (offshore vs. nearshore, indicated with subscripts x and m, respectively) and the two sales channel options (online vs. brick-and- mortar, indicated with subscripts O and B, respec- tively) combine to form four end-to-end supply chain strategies. Let ~p ¼ ðp � sÞ denote the effective price. Let ~cv :¼ ðc3 � sÞ ~cx :¼ ðc2 � sÞ and denote the effec- tive total landed costs of the nearshore and offshore suppliers, respectively. Let, LC qð Þ indicate the expected lost demand in sales channel C, given pro- cured quantity q. Since the model assumes that all unsold products can be salvaged, salvage price can be omitted from the analysis without affecting the valid- ity of its qualitative insights. Using equation (1), we can now describe the expected profit for each end-to- end strategy.

PROPOSITION 1. The expected profits of four end-to-end supply chain strategies, denoted by E[Πsource,Channel], are as follows.

E½Px;O�¼ð1�qOÞ~pl�ð1�qOÞ~pLOðQ�xOÞ�~cxQ�xO; ð4Þ

E½Pm;O� ¼ ð1�qOÞ~pl�ð1�qOÞ~pLOðQ�mOÞ�~cmQ�mO; ð5Þ E½Px;B� ¼ ð1�qBÞ~pl�ð1�qBÞ~pLBðQ�xBÞ�~cxQ�xB; ð6Þ E½Pm;B� ¼ ð1 � qBÞ~pl � ð1 � qBÞ~pLBðQ�mBÞ � ~cmQ�mB: ð7Þ

We emphasize two observations from this proposi- tion. In the first observation, the expected profit of a particular end-to-end supply chain strategy depends on three factors: the expected revenue from products sold and not returned, ð1 � qÞ~pl; the expected rev- enue loss from having insufficient goods to meet the demand, ð1 � qÞ~pLðQ�Þ; and the total cost of procur- ing the products, ~cQ�. For given p and s, the first factor is a function of the sales channel alone (i.e., propor- tion of sales returned, qO and qB); the second factor is a function of the sales channel (i.e., the loss function Lð�Þ for the probability distribution of the demand at each sales location and product returns rate qO or qB) as well as the sourcing strategy (i.e., procured quanti- ties, which are determined by the suppliers’ total landed costs—cm and cx—and the increase in variance of the forecast error at the time of offshore sourcing compared to variance at the time of nearshore sourcing; r22); and the third factor is a function of the sourcing strategy (i.e., total landed costs and procured quantity) and one attribute of the sales channel (i.e.,

product returns rate). The second and the third fac- tors embody the interdependence between the sour- cing and sales strategies. In the second observation, it is difficult to state whether an increase in cost would always increase or decrease the expected profit. An increase in cost would decrease the quantity procured and would increase the second factor, which, ceteris paribus, would lower the expected profit. However, it may also decrease the third factor, which, ceteris pari- bus, would increase the expected profit. Therefore, instead of deriving optimality conditions through first- and second-order derivatives, we describe the conditions in which one strategy dominates the others.

3.3. Comparison of End-To-End Supply Chain Strategies Let (S, C) denote an end-to-end strategy, where S and C refer to the chosen source and sales channel, respec- tively. Let �SC :¼ ð1 � qCÞLCðQ�SCÞ=l denote the lost sales (adjusted for product returns rate of the sales channel) as a fraction of the average market demand, and jSC :¼ ~cSQ�SC=~pl denote the effective total landed cost as a fraction of expected revenue before product returns. kSC represents the sales inefficiency of the end- to-end strategy of converting market demand into sales by making the product available for sale through the stores. It is a function of the choice of source (i.e., quantity procured, which depends on the supplier’s total landed cost; and variance in demand forecast at the time of procurement) as well as the choice of sales channel (i.e., product returns rate; lost sales, which depend on the number of store locations over which the procured quantity is distributed for sale and vari- ability of market demand). Similarly, jSC represents the sourcing inefficiency of the end-to-end strategy of procuring the product cost efficiently to generate the expected revenue. It also is a function of the choice of source (i.e., total landed cost, quantity procured) and the choice of sales channel (i.e., quantity procured, as influenced by product returns rate). Smaller values of kSC and jSC are preferable. Three corollaries of Proposition 1 describe the conditions under which one end-to-end strategy dominates another. Proofs of all corollaries are presented in Appendix S1.

COROLLARY 1. When selling products online, the strat- egy of offshore sourcing is superior to nearshore sourcing, i.e., ðx; OÞ � m; Oð Þ, when jxO � jmO < kmO � kxO. Similarly, when selling products through brick-and-mor- tar stores, the strategy of offshore sourcing is superior to nearshore sourcing, i.e., ðx; BÞ � m; Bð Þ, when jxB � jmB < kmB � kxB.

Thus, between two end-to-end strategies that use the same sales channel C, strategy (s1, C) is superior

to strategy (s2, C) if the excess sourcing inefficiency of (s1, C) compared to (s2, C) is smaller than the excess sales inefficiency of strategy (s2, C) compared to (s1, C). Thus, the preference between two end-to-end supply chain strategies that use the same sales chan- nel is determined not only by the characteristics of the two sources but also by the attributes of the sales channel.

COROLLARY 2. When sourcing from an offshore sup- plier, the strategy of online sales is superior to the brick-and-mortar channel, i.e., ðx; OÞ � x; Bð Þ, when jxO � jxB\ �xB � �xOð Þ � qO � qBð Þ. Similarly, when sourcing from a nearshore supplier, the strategy of online sales is superior to brick-and-mortar stores, i.e., ðm; OÞ � m; Bð Þ, when jmO � jmB\ �mB � �mOð Þ� qO � qBð Þ.

Thus, between two end-to-end strategies that use the same source S, strategy (S, c1) is superior to strategy (S, c2) if the excess sourcing inefficiency of (S, c1) compared to (S, c2) is smaller than the excess sales inefficiency of strategy (S, c2) compared to (S, c1) minus the surplus product returns rate for channel c1 compared to channel c2. Thus, the prefer- ence between two end-to-end strategies that use the same sourcing strategy but different sales strategies is determined by characteristics of the two sales channels as expected, but also by the attributes of the sourcing strategy.

COROLLARY 3. The strategy of online sales of products sourced offshore is superior to brick-and-mortar sales of pro- ducts sourced nearshore, i.e., ðx; OÞ � m; Bð Þ, when jxO � jmB\ �mB � �xOð Þ � qO � qBð Þ. Similarly, the strategy of online sales of products sourced nearshore is superior brick-and-mortar sales of products sourced offshore, i.e., ðm; OÞ � x; Bð Þ, when jmO � jxB\ �xB � �mOð Þ� qO � qBð Þ.

Thus, between two end-to-end strategies that use different sourcing as well as sales strategies, strat- egy (s1, c1) is superior to (s2, c2) if the excess sour- cing inefficiency of (s1, c1) compared to (s2, c2) is smaller than the excess sales inefficiency of strategy (s2, c2) compared to (s1, c1) minus the surplus pro- duct returns rate for channel c1 compared to chan- nel c2. Thus, as expected, the preference between two end-to-end supply chain strategies that have no common sources or sales channels is determined by the attributes of the source as well as the sales channel. The three corollaries can jointly specify the condi-

tion in which one end-to-end strategy dominates the remaining three. We illustrate this by stating the con- ditions under which strategy (x, O) is superior to the

remaining three. Similar conditions can be developed for each end-to-end supply chain strategy.

COROLLARY 4. The strategy of online sales of products sourced offshore, (x, O), is the most profitable of the four end-to-end strategies formed by combining the two source options (offshore, nearshore) and the two sales options (online, brick-and-mortar) when the following holds:

ðjxO þ �xOÞ\ min ðjmO þ �mOÞ � ðqO � qOÞ; ðjxB þ �xBÞ � ðqO � qBÞ; ðjmB þ �mBÞ � ðqO � qBÞ

0 B@

1 CA

¼ min ðjmO þ �mOÞ;

ðjxB þ �xBÞ � ðqO � qBÞ; ðjmB þ �mBÞ � ðqO � qBÞ

0 B@

1 CA:

This shows that the dominance of a particular end-to-end strategy over the other three is deter- mined by the sourcing and sales inefficiencies, as defined earlier, of the end-to-end strategies and the difference in product returns rates of the correspond- ing sales channels. The sourcing and sales efficien- cies depend on the quantity procured, which is a nonlinear function of total landed cost, returns rate, and variance of error in the forecast of season’s demand at the time of procurement. The sales effi- ciency also depends on the lost sales, which is also a nonlinear function of procured quantity and the vari- ance in market demand. Due to these complex inter- dependencies, the expected profit for a given supply chain strategy is an intricate nonlinear function that is difficult to analyze analytically. Therefore, we use a numerical exercise to compare the four end-to-end strategies in various scenarios to generate theoretical insights.

3.4. Numerical Exercise We compare the four end-to-end strategies by varying eight attributes: type of product (functional vs. inno- vative) as modeled by varying the coefficient of varia- tion of forecast error at the time of offshore

procurement (CV2þ3 ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi N r22 þ r23 � �q

=l), proportion

of variance of forecast errors resolved between off- shore and nearshore procurement opportunities

r2 2

r2 2 þr2

� � , coefficient of variation of market demand

(CV4 ¼ ffiffiffiffiffiffiffiffiffi Nr24

q =l), contribution margin if product is

sourced nearshore p�cm p

� � , proportion of brick-and-

mortar sales returned by customers (qB), returns

penalty for online channel qO�qB qB

� � , number of brick-

and-mortar stores in the market (N), and offshore cost advantage (modeled as cx

cm Þ. Exhibits 1 and 2 present

the results of the numerical exercise for 32 cases. These cases are combinations of two extreme values of each of the first five attributes above, and are sum- marized in Table 1. Two extreme values each of the

next two attributes — qO�qB qB

� � and N—are the data

series in the charts showing results of the 32 cases. The last attribute, cx/cm, is varied along the X axis in each chart. The Y axis in each chart shows the

High Contribution Margin: ( − )/ = . Low Contribution Margin: ( − )/ = .

2 2 = 0.9( 2

2 + 3 2) 2

2 = 0.1( 2 2 + 3

2) 2 2 = 0.9( 2

2 + 3 2) 2

2 = 0.1( 2 2 + 3

H ig

h M

ar ke

t D em

an d

V ar

ia ti

on :

= .

L ow

M ar

ke t D

em an

d V

ar ia

ti on

: =

= .

Note: represents the product returns rate for the brick-and-mortar channel (low=0.1, high=0.3). B&M (N=10) Online (10% pen.) Online (30% pen.)B&M (N=25)Offshore +…

Nearshore +… B&M (N=10) B&M (N=25) Online (10% pen.) Online (30% pen.)

Exhibit 1 Expected Profits (in 000s) for Functional Products (CV2+3 = 0.1) [Color figure can be viewed at wileyonlinelibrary.com]

expected profit of the end-to-end strategies. The numeric values used in the exercise for several attrib- uted are based on the practices in the apparel indus- try. However, the qualitative insights obtained from

the exercise are generalizable to other industries with similar attributes as well. IBISWorld (2014a,b) reports that clothing retailers

spend an average of 57.4% and 61.9% of the revenue

High Contribution Margin: ( − )/ = . Low Contribution Margin: ( − )/ = .

2 2 = 0.9( 2

2 + 3 2) 2

2 = 0.1( 2 2 + 3

2) 2 2 = 0.9( 2

2 + 3 2) 2

2 = 0.1( 2 2 + 3

H ig

h M

ar ke

t D em

an d

V ar

ia ti

on :

= .

L ow

M ar

ke t D

em an

d V

ar ia

ti on

: =

= .

Note: represents the product returns rate for the brick-and-mortar channel (low=0.1, high=0.3). B&M (N=10) Online (10% pen.) Online (30% pen.)B&M (N=25)Offshore +…

Nearshore +… B&M (N=10) B&M (N=25) Online (10% pen.) Online (30% pen.)

Exhibit 2 Expected Profits (in 000s) for Innovative Products (CV2+3 = 1) [Color figure can be viewed at wileyonlinelibrary.com]

for procuring men’s and women’s clothing items, respectively. This amounts to about 40% contribution margin. Fisher and Raman (2010) note that the typical gross margins for retailers are between 30% and 50%. In the numerical exercise, we use two values for con- tribution margin, 40% and 40%, for the goods pro- cured from a nearshore source (i.e., cm = 0.6p and cm = 0.3p). We then consider offshore cost advantage by evaluating the strategies for a range of offshore- cost-to-nearshore-cost-ratios (cx/cm) between 0.2 and 1.2, in increments of 0.05. We could not find statistics about forecast accuracy in apparel sourcing. There- fore, we use the statistic “average margin of error in the forecast at the time production is committed” from Fisher (1997) instead. This statistic is equivalent to the uncertainty resolved by demand signals d2 and d3 in our model. We model functional products as hav- ing CV2+3 = 0.1, and innovative products with CV2+3 = 1, according to the statistics by Fisher (1997, p. 107). We tested the model with CV2+3 = 0.4, the lower end of the range of margin of error for innova- tive products suggested by Fisher (1997); the results lie between those for CV2+3 = 0.1 and CV2+3 = 1, and do not provide additional insights. Therefore, the results for CV2+3 = 0.4 are not included in the study. We also could not find statistics related to the deterio- ration of forecast accuracy over procurement lead time. Therefore, we consider two cases of proportion of uncertainty about season’s demand resolved between offshore and nearshore sourcing opportuni- ties, modeled by a ¼ r22=ðr22 þ r23Þ. In the first case, most of this uncertainty is resolved before the near- shore procurement (but after the offshore procure- ment opportunity) by demand signal d2; we assume a = 0.9. In the second case, most of the uncertainty still remains unresolved at the time of nearshore

procurement; for this we assume a = 0.1. The values of a, although somewhat arbitrary, are chosen to test the performance of end-to-end supply chains strategy between two fairly extreme cases. Greater the value of a, the more advantageous it is to source from a near- shore source.1

On the sales side, a study using six years of data at a large national catalog retailer of apparel and acces- sories found that, on average, 16% of products sold were returned by the customers (Petersen and Kumar 2009). A similar returns rate was found for L.L. Bean —a retailer known for one of the most generous pro- duct return policies—where eight out of 48 million items it shipped in 2006 were returned (Loudin 2007). Other studies of product returns have reported returns rates of five-to-nine percent for hard goods and up to 35% for high-end apparel (Guide et al. 2006). We test the end-to-end strategies for the returns rates of 10% and 30% at the brick-and-mortar stores (qB), with an addition penalty of 10% or 30% for the same product sold online (i.e., qO = 1.1qB or = 1.3qB). Thus, in the model, the returns rates for the products sold online vary between 11% and 39%. We also vary the number of brick-and-mortar stores in a market (N = 10 and 25, for the brick-and-mortar channel). Finally, we consider two values of variation of market demand (CV4 ¼ r4l ¼ 0:01 and 0.1). These values are chosen so that the coefficient of variation of demand at a single brick-and-mortar store (CVS) does not exceed 0.5, which is considered to be a satisfactory threshold to ensure that the assumption of normal distribution of store-level demand is valid (Berman et al. 2011). For these values, the highest value of

CVS ¼ r4= ffiffiffi N

p l=N ¼

ffiffiffiffi N

p r4 l

� � in the numerical exercise is 0.5

when N = 25 and r4l ¼ 0:1. Assuming that the demand

Table 1 Summary of Cases

Functional products Innovative products

Case CV2+3 p�cm p CV4 qB

r2 2

r2 2 þr2

3 Case CV2+3 p�cm p CV4 qB

r2 2

r2 2 þr2

1 0.1 0.7 0.01 0.1 0.9 17 1 0.7 0.01 0.1 0.9 2 0.1 0.7 0.01 0.1 0.1 18 1 0.7 0.01 0.1 0.1 3 0.1 0.7 0.01 0.3 0.9 19 1 0.7 0.01 0.3 0.9 4 0.1 0.7 0.01 0.3 0.1 20 1 0.7 0.01 0.3 0.1 5 0.1 0.7 0.1 0.1 0.9 21 1 0.7 0.1 0.1 0.9 6 0.1 0.7 0.1 0.1 0.1 22 1 0.7 0.1 0.1 0.1 7 0.1 0.7 0.1 0.3 0.9 23 1 0.7 0.1 0.3 0.9 8 0.1 0.7 0.1 0.3 0.1 24 1 0.7 0.1 0.3 0.1 9 0.1 0.4 0.01 0.1 0.9 25 1 0.4 0.01 0.1 0.9 10 0.1 0.4 0.01 0.1 0.1 26 1 0.4 0.01 0.1 0.1 11 0.1 0.4 0.01 0.3 0.9 27 1 0.4 0.01 0.3 0.9 12 0.1 0.4 0.01 0.3 0.1 28 1 0.4 0.01 0.3 0.1 13 0.1 0.4 0.1 0.1 0.9 29 1 0.4 0.1 0.1 0.9 14 0.1 0.4 0.1 0.1 0.1 30 1 0.4 0.1 0.1 0.1 15 0.1 0.4 0.1 0.3 0.9 31 1 0.4 0.1 0.3 0.9 16 0.1 0.4 0.1 0.3 0.1 32 1 0.4 0.1 0.3 0.1

at individual stores is normally distributed, the high- est likelihood of experiencing negative demand is only 2.3%.

4. Results

The expected profits for the four end-to-end strate- gies in a given scenario are calculated using equa- tions (4)–(7). Exhibits 1 and 2 present the results graphically. The chart for each case shows the expected profits (Y-axis) for a range of offshore cost discounts, calculated as the ratios of offshore-cost- to-nearshore-cost, cx/cm (X-axis). They are calculated by keeping cm constant and varying cx from 0.2cm to 1.2cm. Four sales strategies—namely, brick-and-mor- tar channel with 10 and 25 stores each, and online channel with 10% and 30% returns penalty (i.e., qO�qB qB

¼ 0:1 and = 0.3) each—are plotted as four data series in each chart. The expected profit of an end- to-end strategy involving a nearshore source is pro- jected along the X-axis with a dotted line to show its relation to the expected profits for various off- shore cost conditions. We compare results from multiple cases to draw theoretical insights about the attractiveness of different end-to-end strategies under various scenarios. These insights are pre- sented as a series of observations in the following sections.

4.1. End-To-End Supply Chain Strategies for Functional Products We first evaluate the four end-to-end strategies for functional products (Exhibit 1). For all 16 cases pre- sented in Exhibit 1, the expected profit for a given sales channel is always higher when the products are procured from an offshore source with a lower cost than the nearshore source. Thus, consistent with Fisher’s (1997) argument, the optimal sourcing strategy for functional products is to choose a low- cost offshore supplier, for a given sales strategy. Furthermore, when the product returns rate is low (qB = 0.1) and the market demand has low variabil- ity (CV4 = 0.01; cases 1, 2, 9, and 10), the expected profits in the online and the brick-and-mortar sales channels for a given sourcing option are similar. These cases represent the products in familiar cate- gories (per hypotheses 3 and 4 in Petersen and Kumar (2009)). Thus, the best supply chain strategy for delivering familiar functional products is to source them from an offshore supplier; they could be sold either online or through brick-and-mortar stores with similar level of profitability. The advan- tage of an offshore source over a nearshore source increases if the contribution margin of the product is low (as in cases 9, 10 vs. cases 1, 2). Also, as expected, an increase in the variation of market

demand makes the online channel more profitable compared to the brick-and-mortar channel (cases 5–8 vs. cases 1–4, and cases 13–16 vs. 9–12) because of the inventory pooling benefit of the online channel.

OBSERVATION 1. For functional products, (a) offshore source is the dominant sourcing choice for a given sales channel, and (b) brick-and-mortar and online channels have similar profitability, for a given source, if the product returns rate is low and the market demand is predictable (i.e., low CV4).

However, for functional products with market demand of low variability, if the offshore cost dis- count is small, an appropriate end-to-end strategy involving the responsive nearshore sourcing can dominate a strategy involving the low-cost offshore sourcing. If the functional products have high returns rate (qB = 0.3) (cases 3, 4, 11, 12)—which may be the case for the products in new categories (Petersen and Kumar 2009)—it could be more profitable to source them nearshore and sell through brick-and-mortar stores, instead of sourcing offshore for selling online. This is especially the case if the online returns penalty is high. This counterintuitive result is due to the effect of product returns rate on newsvendor critical ratio, as stated in Lemma 2. In our model, the returns rate of the online channel is higher than that of the brick- and-mortar channel by a certain proportion (10% and 30% higher, in the numerical cases). When the returns rate for the brick-and-mortar channel is high (qB = 0.3), the (multiplicative) online returns rate is even higher (qO = 0.33 or 0.39). When the online chan- nel is paired with an offshore source whose total landed cost is only marginally lower than the cost of the nearshore source, the drop in the critical ratio due to higher returns rate of the online channel is not com- pensated adequately by the lower cost of the offshore source. For instance, for a product with nearshore

contribution margin p�cm p

� � of 0.4, the critical ratio for

the strategy of brick-and-mortar selling (qB = 0.3) is w ¼ 1

1�0:3 � �

0:4 � 0:3 1�0:3 � �

¼ 0:143. If that product was procured from an offshore source with 10% price dis-

count (i.e., p�cx p

¼ 0:36) and sold online with a returns penalty of 10% (i.e., qO = 0.33), the critical ratio is w ¼ 1

1�0:33 � �

0:36 � 0:33 1�0:33 � �

¼ 0:045. As a result, the firm with an end-to-end strategy of ‘offshore sourcing and online sales’ will procure fewer pieces of the apparel item compared to the firm with an end-to-end strat- egy of ‘nearshore sourcing and brick-and-mortar sales.’ Since the variability of market demand is low, the risk-pooling benefit of the online channel over the brick-and-mortar channel is small. The net result of these effects is that the strategy of ‘nearshore sourcing

with brick-and-mortar sales’ can outperform the strat- egy of ‘offshore sourcing and online sales.’ Thus, an end-to-end strategy involving a responsive source (i.e., nearshore) and a responsive sales channel (i.e., brick-and-mortar) can dominate the one involving a cost-efficient source and a cost-efficient sales channel, for functional products.

OBSERVATION 2. For functional products with pre- dictable market demand (i.e., low CV4) but high product returns rates, the responsive end-to-end strategy of nearshore sourcing and brick-and-mortar retail can outperform the cost-efficient strategy of offshore sourcing and online retail, if the offshore price discount is low (i.e., cx/cm close to 1) and the online returns penalty (qO/qB) is high.

4.2. End-to-end Supply Chain Strategies for Innovative Products We next consider end-to-end strategies for innovative products, which are likely to have high uncertainty in the aggregate season’s demand (CV2+3 = 1; results in Exhibit 2). If a large portion of uncertainty in the sea- son’s demand gets resolved between the times for making the purchase decisions for offshore (t2) and nearshore sourcing (t3)—as represented in the odd- numbered cases in Exhibit 2 —a responsive nearshore source can be more profitable than an offshore source for a given sales strategy despite high offshore cost discount. This is consistent with Fisher’s (1997) rec- ommendation. The advantage of nearshore sourcing is enhanced if the product has high contribution mar- gin, as is likely for innovative products. Compared to their equivalent cases of functional products (i.e., odd-numbered cases in 1–16), their expected profit also rises more slowly with the increase in margin resulting from an increase in the offshore cost advan- tage. The range of offshore cost advantage for which the nearshore source is more profitable than the lower-cost offshore source varies for different sales strategies. For the products with low returns rates and low variability in market demand (cases 17–18, 25–26)—which are typically the products in familiar categories (Petersen and Kumar 2009)—the online and brick-and-mortar sales channels have similar profitability for a given sourcing option, as was the case with their equivalent functional products (cases 1–2 and 9–10). Thus, the best end-to-end supply chain strategy for the innovative functional products is to source them from a nearshore supplier; either sales channel can be used with similar profitability. If the variability of market demand is low, the end-to-end strategy with a nearshore source and brick-and-mor- tar stores is superior despite fairly high offshore cost advantages. This is the supply chain strategy

famously associated with the Zara model (Ghemawat and Nueno 2003).

OBSERVATION 3. For innovative products, (a) nearshore source is the dominant sourcing choice for a given sales channel for a wide range of offshore cost discounts, (b) brick-and-mortar and online channels have similar profitability, for a given source, if the product returns rates are low and the market demand is predictable, but (c) the brick-and-mortar channel can be more profitable than the online channel if the product returns rates are high.

Obviously, the nearshore source is more prof- itable when a large portion of uncertainty in the season’s aggregate demand gets resolved between the times when procurement decisions for offshore and nearshore sources are made. Comparison of the odd-numbered cases in Exhibit 2 (in which 90% of the variance in forecast error at the time of pro- curement gets resolved between the timings of offshore and nearshore procurement) with the even-numbered cases to their right (in which only 10% of this variance is resolved) shows that near- shore procurement is advantageous only if delay- ing the purchase decision to buy from the nearshore supplier, instead of the offshore supplier, helps resolve most of the uncertainty associated with the season’s demand. However, an end-to-end strategy with nearshore

sourcing can get dominated by one with offshore sourcing even when a majority of uncertainty at the time of procurement gets resolved between the times when procurement decisions for offshore and nearshore sources are made. Comparison of the end- to-end strategy of ‘nearshore sourcing and brick-and- mortar selling (N = 25)’ with ‘offshore sourcing and online sales’ between cases 17 and 21, and cases 19 and 23 shows that the former strategy gets dominated by the latter as the variation of market demand increases from CV4 = 0.01 (cases 17 and 19) to CV4 = 0.1 (cases 21 and 23). This superiority of the offshore-online supply chain strategy is valid even when the offshore discount is fairly low (cx/cm close to 1). Admittedly, the best performing end-to-end strategy in case 21 still maintains the responsive near- shore source (but with the cost-efficient online chan- nel) if the offshore cost discount is moderate (approximately, 0.7 ≤ cx/p ≤ 1). However, the impor- tant point to note here is that the supposedly ideal responsive strategy of ‘nearshore sourcing with brick- and-mortar sales’ can now get dominated by an entirely cost-efficient strategy of ‘offshore sourcing with online sales,’ and the best performing strategy also has one cost-efficient element (i.e., the online channel).

The reason for this result is as follows. At low off- shore cost discounts, the newsvendor critical ratio for the offshore source is only moderately higher than that for the nearshore source. As a result, the quantity procured from an offshore source is only moderately higher than that from a nearshore source. However, this quantity is then placed at a single location (ware- house for the online store) instead of being distributed among multiple brick-and-mortar stores. The inven- tory pooling benefit of the online channel exceeds the benefit obtained from reducing uncertainty in the sea- son’s demand through nearshore sourcing, when variability of the market demand is high.

OBSERVATION 4. For innovative products, if the store- level demand has high uncertainty (i.e., large CV4), (a) a cost-efficient end-to-end strategy of offshore sourcing and online retail can outperform the responsive end-to-end strategy of nearshore sourcing and brick-and-mortar retail, (b) nearshore sourcing and online retail can be the optimal strategy with low offshore cost discount, and (c) offshore sourcing and online retail can be the optimal strategy if the offshore cost discount is moderate.

4.3. Effect of Contribution Margin The above observations compare different end-to-end strategies for functional and innovative products. Next, we examine how the performance of end-to- end strategies is affected by changes in the product’s contribution margins. We begin with the analysis of functional products. Comparison of cases 1–8 (high margin) with the corresponding cases 9–16 (low mar- gin) shows that contribution margin does not affect the preference ordering of the four end-to-end strate- gies. The only noticeable change is that the benefit of offshore sourcing over nearshore sourcing rises more rapidly with an increase in offshore cost discount when the products have low contribution margin.

OBSERVATION 5. For functional products, a change in the contribution margin does not alter the preference ordering of end-to-end supply chain strategies.

For innovative products, the preference ordering between end-to-end strategies does not change if the store-level demand is fairly predictable (CV4 = 0.01), as observed by comparing cases 17–20 with the corresponding cases 25–28. However, when the store- level demand has higher variability, a drop in the contribution margin can make the end-to-end strategy of ‘offshore sourcing and online sales’ more profitable than the strategy of ‘offshore sourcing and brick-and- mortar sales’ for moderate level of offshore cost discounts, especially when online returns penalty is small (as observed by comparing cases 21–22 with

cases 29–30). This change in preference ordering of end-to-end strategies can be explained by the newsvendor ratio. At low contribution margin, the newsvendor critical ratio is small, and as a result, pro- cured quantity is small compared to the cases with high contribution margin. In presence of higher vari- ability in the store-level demand, it is more beneficial to pool the procured quantity for sale though one (or fewer) locations, as achieved in online selling, instead of in the brick-and-mortar channel.

OBSERVATION 6. For innovative products, a change in the contribution margin has the following effect on the preference ordering of end to end strategies: (a) if the store-level demand is fairly predictable (i.e., small CV4), a change in the contribution margin does not alter the preference ordering of end-to-end supply chain strategies, (b) if the store-level demand has high variability (i.e., large CV4), a decrease in the contribution margin can make online selling more attractive than brick-and-mortar selling, when paired with offshore sourcing with moderate-to-high levels of discount.

4.4. Practitioners’ Critique of the Study’s Observations We sought post hoc critique of the study’s observations from eight executives in the apparel industry. Below, we present a gist of their responses to the observation. A more detailed version of the critique is presented in the paper’s Appendix S1. Five executives (R1, R3, S1, M1 and M3) commenting on Observation 1 all con- curred with part (a), and agreed with part (b) with the qualification that other factors—such as channel fixed costs, sales volumes, etc.—also influence the respec- tive channel’s profitability. Our respondents did not have a strong reaction to Observation 2: those com- menting (R1, R3, S1, and S2) tended to agree with the observation but did not provide any corroborative argument. Five executives (R1, R2, R3, S2, and M2) agreed with Observation 3 that retailers need to use nearshore sources for fashion products. All executives commenting on Observation 4 (M1, R1, R2, and R3) mentioned that Zara’s knowledgeable store managers and its creation of impulse-purchase environment made store demand more predictable to enable Zara’s ‘nearshore-sourcing, brick-and-mortar sales’ strategy. Two executives stressed that creating this experience online was difficult (M1, R2). All executives com- menting on Observations 5 and 6 (R1, S2, M3) agreed with them. They also cautioned that the retailer needs to consider whether it should sell through the brick- and-mortar channel if it does not understand the store demand (R1, S2). The executives also mentioned a few factors not considered in the model as being rele- vant, such as channel fixed costs, channel’s ability to

generate additional sales, sources of raw materials used for making garments, and use of brick-and-mor- tar outlets as showrooms for the retailer’s online store. These factors could be considered in future models of apparel supply chains.

5. Discussion

This study advances the supply chain strategy litera- ture (Fine 2000, Fisher 1997) by exploring the trade- offs between different sourcing and sales strategies of one firm in an end-to-end model of its supply chain strategy. We compare the model’s predictions against Fisher’s (1997) benchmark recommendations of ideal supply chain strategies for delivering a firm’s func- tional and innovative products under various scenar- ios. Our analysis supports the recommendations in several cases, but also highlights the scenarios in which the recommended strategies may get domi- nated by those considered to be inferior. The benefit of the integral end-to-end approach is seen in the intriguing insights and the corresponding practical implications yielded by the study. They may explain the “significant gap between expected and achieved performance” observed in real-world offshoring pro- jects (Larsen et al. 2013, p. 534) and other strategic supply chain decisions. Observation 1 supports Fisher’s (1997) recommen-

dations by showing that a cost-efficient offshore supplier is the dominant source for functional prod- ucts. It also shows that brick-and-mortar and online channels may be employed equally effectively, in conjunction with offshore sourcing, if the market demand is predictable and product returns rate is low. Familiar functional products typically have low returns rate and predictable market demand (Petersen and Kumar 2009). Based on observation 1, the most effective end-to-end supply chain for such products would pair cost-efficient sourcing with either online or brick-and-mortar channel. Observation 2 describes the instances in which Fisher’s recommendations for functional products may not hold. It suggests that functional products with predictable market demand but high returns rate—such as, the products that need to be examined physically to determine their fit with the customer need before making the purchase deci- sion—could be served more profitably by a respon- sive supply chain involving a nearshore source and brick-and-mortar stores, instead of a cost-efficient supply chain consisting an offshore source with low- to-moderate discount and the online channel. However, the superiority of brick-and-mortar channel diminishes and online selling becomes more attrac- tive, even in presence of high online returns penalty, when the demand at individual stores becomes less predictable. These results suggest that online clothing

and shoe retailers (like Zappos) could become more profitable if the demand at physical stores becomes more variable. In such a scenario, the firms using brick-and-mortar stores will need to change their end- to-end strategies to consolidate the inventory at fewer physical locations (not necessarily at one warehouse, as for the online channel). Observation 3 supports Fisher’s recommendations

for innovative products by highlighting the superior- ity of responsive nearshore sourcing over the cost- efficient offshore sourcing. The obvious caveat for the nearshore source to be more effective is that signifi- cant amount of uncertainty about the aggregate market demand experienced at the time of procuring goods from the offshore source needs to be resolved by the time of ordering goods from the nearshore source. This justifies the practice of sourcing products with high demand uncertainty, such as fashion items and new products, close to the market. Observation 3 also shows that the brick-and-mortar channel is supe- rior to the online channel when the product returns rates are high. This economic advantage of ‘nearshore source with brick-and-mortar sales channel’ for fash- ion products justifies the Zara model. However, the superiority of this end-to-end model requires that the uncertainty in the store-level demand is low. The cele- brated Zara model minimizes uncertainty about the store-level demand by actively involving its store managers in design and ordering of the merchandise. This is such an important aspect of Zara’s strategy that its CEO called availability of store managers who understand the store demand “the single most impor- tant constraint on the rate of store additions” (Ghe- mawat and Nueno 2003, p. 14). In the case of high uncertainty in the store-level demand, as noted in Observation 4, we see a complete reversal of the pref- erence for end-to-end strategies: the cost-effective end-to-end supply chain consisting of ‘offshore sour- cing and online selling’ can become more profitable than the responsive supply chain involving ‘near- shore sourcing and brick-and-mortar selling.’ This is particularly true when the total landed cost of the offshore supplier is close to that of the nearshore supplier! Two aspects of this reversal highlight the need for taking an end-to-end perspective. First, the reversal occurs when offshore cost is high and closer to that of the nearshore source, which is counter-intui- tive. The reason for this effect is that the quantity pro- cured from the offshore source is not much higher than that procured from the nearshore supplier, but is now centralized in one location for online sales, where the benefits of inventory pooling is high due to the high uncertainty of store-level demand. Second, the reversal shows how change in one attribute of the business environment—namely, the predictability of market demand in this case—can cause a eulogized

and emulated business model to get dethroned by its polar opposite. If the increase in mobile e-commerce results into an increased uncertainty of the store-level demand, the revered Zara business model could get deposed by the one using the end-to-end strategy of ‘offshore sourcing and online sales.’ Observations 5 and 6 relate to the robustness of

end-to-end strategies to changes in contribution mar- gin. For functional products, it is always preferable to source from a cost-efficient offshore supplier, and sell via brick-and-mortar stores (online) if the store-level demand is (not) predictable (Observation 5). For inno- vative products, the end-to-end strategy is robust to changes in contribution margin if the store-level demand is fairly predictable; however, if the store- level demand is unpredictable, a decrease in contribu- tion margin can rapidly make online selling more attractive than the brick-and-mortar channel (Obser- vation 6). This suggests that firms using the Zara model of brick-and-mortar selling need to ensure that their products enjoy a high margin to ensure the viability of the responsive nearshore source.

5.1. Generalization to Other Industries Some of this study’s conclusions may differ in other industries if one of the underlying premises, peculiar to the apparel industry, does not hold. In 2015, 54.6% of the $340 billion e-commerce revenue in the U.S. was generated by four product categories (eMarketer 2017): apparel & accessories, computers & computer electronics, auto & parts, and books/music/video. Apparel products differ from those in the other three industries in that buyers can judge the fit of apparel items completely only after their physical examina- tion. As a result, apparel products may experience much higher returns rate for online sales than the brick-and-mortar sales. Observation 2 states that func- tional products with predictable market demand but high returns rates could be sold more profitably using a responsive ‘nearshore sourcing and brick-and-mor- tar selling’ than a cost-efficient offshore-online strat- egy, if the online returns penalty is high. This contradiction of Fisher’s (1997) prescription of the ideal supply chain strategy may not be observed if a product’s fit to a buyer’s needs can be assessed with- out experiencing the product physically, such as based on technical specifications and online reviews (e.g., electronics, auto parts, or books). The contradiction of Fisher’s (1997) recommenda-

tion for innovative products (Observation 4) is possible only when the store-level demand is difficult to pre- dict. Apparel retailers, like Zara, use vast data of store sales and store managers’ knowledge of the local trends to estimate the store-level demand more accu- rately for allocating inventory to the brick-and-mortar stores (Ghemawat and Nueno 2003). However, the

industries that experience highly unpredictable store- level demands for the innovative products (e.g., new movies at the movie halls, new trade books in book stores, etc.) could be served more profitably by the online channel than the brick-and-mortar stores.

5.2. Limitations and Future Research Our analysis focused on the tradeoffs among four ‘pure’ end-to-end strategies. In reality, different end- to-end strategies may be ideal for a firm’s different functional and innovative products based on their returns rates, predictability of market demand, off- shore cost discounts, and so on. Whether a firm adopts the strategy best suited for each product depends on the tradeoff between the marginal benefit obtained by choosing that strategy and the fixed costs associated with the implementation of that strategy. The model presented in this work does not consider these financial tradeoffs involving the fixed costs. A second limitation of this work is the assumption

that the firm uses the newsvendor model to determine its purchase quantity for each selling season. Empirical research shows that managers making pro- curement decisions in newsvendor settings exhibit pull-to-center bias, i.e., the quantity they order is between the average demand and the newsvendor optimum (Schweitzer and Cachon 2000). Thus, the quantity procured in all four end-to-end configura- tions would be lower than the standard newsvendor optimal when the contribution margin is high (cases 1–8, 17–24), and higher when the margin is low (cases 9–16, 25–32). The newsvendor critical ratio with a deterministic product returns rate, as used in this study, is lower than the standard newsvendor critical ratio (lemma 2). Therefore, in presence of the pull-to- center bias, the procured quantity will be closer to (farther from) optimal for high-margin (low-margin) products. However, in absence of field data of how decision makers consider product returns rate when deciding procurement quantities, this remains specula- tive and provides an opportunity for future research. Another limitation of the newsvendor assumption is

that some items may be replenished using a multi- period periodic-review inventory policy, in which the leftover inventory is not salvaged but available for sale in the following period. Such a policy can be modeled as newsvendor, with the cost of carrying inventory to the next period as the cost of excess inventory. To model replenishment using a multiperiod inventory policy, additional parameters, such as inventory carry- ing charge, will need to be considered. A limitation related to the model of the sales chan-

nel is the assumption that the retailer sells through either brick-and-mortar stores or an online store, but not both. Many current industry practices justify this assumption (Berg et al. 2015). However, the model in

this study can be extended to examine end-to-end strategies with omnichannel retailing. At least two forms of omnichannel retail should be considered: showrooming, in which shoppers examine products at a brick-and-mortar store and complete the purchase online (Balakrishnan et al. 2014) and webrooming or buy-online-pick-at-store, in which shoppers read pro- duct reviews and check store availability online, and complete the purchase at a brick-and-mortar store (Gao and Su 2017). To compare end-to-end strategies with brick-and-mortar, online, and omnichannel sales, a demand model to specify product demand from different types of consumers—such as, those who shop exclusively in either brick-and-mortar or online channels, as well as the omnichannel shoppers —may need to be used. Additionally, product returns rates for omnichannel shopping may be assumed to be between the returns rates for the brick-and-mortar and online channels. The model presented in this study could be

extended to include a profit multiple for the brick-and- mortar channel to reflect that customers may make additional purchases after visiting a store (Gao and Su 2017). Future modeling works could also evaluate mixed strategies, in which a firm uses more than one type of source or sales channel (Balakrishnan et al. 2014, Klotz and Chatterjee 1995). The end-to-end model for one firm developed here could be extended to study competition between sales channels (Balasub- ramanian 1998) or the choice of sourcing strategy under competition (Wu and Zhang 2014). Future empirical studies could test the observations in this study. Even though the model was evaluated for the parametric values observed in the apparel industry, the observations from the numerical analysis are gen- eral and could be tested in other industry contexts. Case studies of end-to-end supply chain strategies of different firms in the same industry could be used to validate or challenge the observations. In summary, this study illustrates the interdepen-

dence of two important decisions in apparel supply chains: sourcing and sales strategies. This interdepen- dence is demonstrated by revealing the conditions in which the ideal sourcing and sales strategies for func- tional and innovative products (Fisher 1997) can be outperformed by those considered inferior, when used as the components of an integral end-to-end strategy. Such interdependencies could explain the unexpected results, such as the hidden costs of out- sourcing resulting from an increase in the firm’s ‘con- figuration complexity’ (Larsen et al. 2013).

Acknowledgments

We are grateful to Fred Abernathy, Inga-Lena Darkow, David Gonsalvez, Steve Graves, John Gray, Nitin Joglekar,

Sang Jo Kim, Leonard Lane, Angel Poyato, and the seminar participants at the Asia School of Business, Industry Studies Conference (2015), and Academy of Management Annual Meeting (2015) for many helpful comments and for connect- ing us with industry practitioners who helpfully critiqued the study’s propositions. We sincerely thank the editor, the senior editor, and the anonymous reviewers for their insightful suggestions.

Note

1We thank one of the anonymous referees for pointing this out.

References Arya, A., B. Mittendorf, D. Sappington. 2007. The bright side of

supplier encroachment. Market. Sci. 26(5): 651–659.

Balakrishnan, A., S. Sundaresan, B. Zhang. 2014. Browse-and- switch: Retail-online competition under value uncertainty. Prod. Oper. Manag. 23(7): 1129–1145.

Balasubramanian, S. 1998. Mail versus mall: A strategic analysis of competition between direct marketers and conventional retailers. Market. Sci. 17(3): 181–195.

Berg, A., S. Hedrich. 2014. What’s next in apparel sourcing. Per- spectives on retail and consumer goods, edited by McKinsey & Company, (3).

Berg, A., L. Brantberg, L. Herring, P. Sil�en. 2015. Mind the Gap: What Really Matters for Apparel Retailers in Omnichannel. McKinsey & Company.

Berman, O., D. Krass, M. Tajbakhsh. 2011. On the benefits of risk pooling in inventory management. Prod. Oper. Manag. 20(1): 57–71.

Bernstein, F., A. Federgruen. 2005. Decentralized supply chains with competing retailers under demand uncertainty. Manage- ment Sci. 51(1): 18–29.

Boston Consulting Group. 2015. Reshoring of Manufacturing to the US Gains Momentum. Boston Consulting Group.

Brynjolfsson, E., Y. Hu, M. Rahman. 2009. Battle of the retail chan- nels: How product selection and geography drive cross- channel competition. Management Sci. 55(11): 1755–1765.

Calvo, E., V. Martinez-de-Albeniz. 2016. Sourcing strategies and supplier incentives for short-life-cycle goods. Management Sci. 62(2): 436–455.

Chiang, C., W. Benton. 1994. Sole sourcing versus dual sourcing under stochastic demands and lead times. Nav. Res. Logisti. 41(5): 609–624.

Chiang, W., D. Chhajed, J. Hess. 2003. Direct marketing, indirect profits: A strategic analysis of dual-channel supply-chain design. Management Sci. 49(1): 1–20.

Dana, J., K. Spier. 2001. Revenue sharing and vertical control in the video rental industry. J. Indus. Econ. 49(3): 223–245.

Fine, C. 2000. Clockspeed-based strategies for supply chain design. Prod. Oper. Manag. 9(3): 213–221.

Fisher, M. 1997. What is the right supply chain for your product? Harv. Bus. Rev. 75(2): 105–116.

Fisher, M., A. Raman. 2010. The New Science of Retailing: How Ana- lytics Are Transforming the Supply Chain and Improving Perfor- mance. Harvard Business Press, Boston, MA.

Forrester Research. 2014. U.S. e-commerce forecast: 2013 to 2018. Forrester Research.

Gallino, S., A. Moreno, I. Stamatopoulous. 2017. Channel integra- tion, sales dispersion, and inventory management. Manage- ment Sci. 63(9): 2813–2831.

Gao, F., X. Su. 2017. Omnichannel retail operations with buy-online-and-pick-up-in-store. Management Sci. 63(8): 2478–2492.

Ghemawat, P., J. Nueno. 2003. Zara: Fast Fashion. Harvard Busi- ness School, Boston, MA.

Graves, S., H. Meal, S. Dasu, Y. Qiu. 1986. Two-stage production planning in a dynamic environment. S. Axater, C. Sch- neeweiss, E. Silver, eds. Multi-Stage Production Planning and Control. Springer-Verlag, Berlin, 9–43.

Guide Jr, V., G. Souza, L. Van Wassenhove, J. Blackburn. 2006. Time value of commercial product returns. Management Sci. 52(8): 1200–1214.

Heath, D., P. Jackson. 1994. Modeling the evolution of demand forecasts with application to safety stock analysis in produc- tion/distribution systems. IIE Trans. 26(3): 17–30.

IBISWorld. 2014a. IBISWorld industry report 44811: Men’s cloth- ing stores in the US. IBISWorld.

IBISWorld. 2014b. IBISWorld industry report 44812: Women’s clothing stores in the US. IBISWorld.

IBISWorld. 2016. IBISWorld industry report C1311-GL: Global apparel manufacturing. IBISWorld.

International Apparel Federation. 2017. About IAF. Available at http://iafnet.eu/about/ (accessed date February 25, 2017).

Klotz, D., K. Chatterjee. 1995. Dual sourcing in repeated procure- ment competitions. Management Sci. 41(8): 1317–1327.

Larsen, M., S. Manning, T. Pedersen. 2013. Uncovering the hidden costs of offshoring: The interplay of complexity, organiza- tional design, and experience. Strateg. Manag. J. 34(5): 533–552.

Li, C. 2013. Sourcing for supplier effort and competition: Design of the supply base and pricing mechanism. Management Sci. 59(6): 1389–1406.

Li, D., C. O’Brien. 2001. A quantitative analysis of relationships between product types and supply chain strategies. Int. J. Prod. Econ. 73(1): 29–39.

Li, Z., S. Gilbert, G. Lai. 2014. Supplier encroachment under asymmetric information. Management Sci. 60(2): 449–462.

Loudin, A. 2007. Bean there, returned that. Available at http:// www.inboundlogistics.com/cms/article/bean-there-returned- that/ (accessed date October 15, 2015).

eMarketer, 2017. Commerce snapshot. Available at https:// www.emarketer.com/public_media/docs/ (accessed date February 25, 2017).

Manyika, J., J. Bughin, S. Lund, O. Nottebohm, D. Poulter, S. Jauch, S. Ramaswamy. 2014. Global flows in a digital age: How trade, finance, people and data connect the world economy. McKinsey Global Institute. Available at http:// www.mckinsey.com/business-functions/strategy-and-corporate- finance/our-insights/global-flows-in-a-digital-age (accessed date December 1, 2014).

New York Times. 2017. Amazon to open retail store in Manhattan in Time Warner Center.

Ofek, E., Z. Katona, M. Sarvary. 2011. “Bricks and clicks”: The impact of product returns on the strategies of multichannel retailers. Market. Sci. 30(1): 42–60.

Petersen, J., V. Kumar. 2009. Are product returns a necessary evil? Antecedents and consequences. J. Market. 73(3): 35–51.

Pisano, G., P. Adams. 2009. VF Brands: Global Supply Chain Strat- egy. Harvard Business School, Boston, MA.

Reitzig, M., S. Wagner. 2010. The hidden costs of outsourcing: Evidence from patent data. Strateg. Manag. J. 31(11): 1183–1201.

Schweitzer, M., G. Cachon. 2000. Decision bias in the newsvendor problem with a known demand distribution: Experimental evidence. Management Sci. 46(3): 404–420.

Selldin, E., J. Olhager. 2007. Linking products with supply chains: Testing Fisher’s model. Supply Chain Manag. 12(1): 42–51.

Taylor, T. 2002. Supply chain coordination under channel rebates with sales effort effects. Management Sci. 48(8): 992–1007.

Treleven, M., S. Schweikhart. 1988. A risk/benefit analysis of sourcing strategies: Single vs. multiple sourcing. J. Oper. Manag. 7(4): 93–114.

de Treville, S., L. Trigeorgis. 2010. It may be cheaper to manufac- ture at home. Harv. Bus. Rev. 88(10): 84–87.

U.S. Census Bureau. 2017. Quarterly retail e-commerce sales: 4th quarter 2016. Report, U.S. Census Bureau, Washington, DC.

Weinswig, D. 2017. Five ways that Amazon is shaking up retail.Available at https://www.forbes.com/sites/deborahwe inswig/2016/05/24/5-ways-that-amazon-is-shaking-up-retail/ (accessed date April 30, 2017).

Wood, S. 2001. Remote purchase environments: The influence of return policy leniency on two-stage decision processes. J. Mark. Res. 38(2): 157–169.

WSJ. 2013. Rampant returns plague e-retailers. Wall Street Journal.

Wu, X., F. Zhang. 2014. Home or overseas? An analysis of sour- cing strategies under competition. Management Sci. 60(5): 1223–1240.

Yoon, D. 2016. Supplier encroachment and investment spillovers. Prod. Oper. Manag. 25(11): 1839–1854.

Zhang, W., Z. Hua, S. Benjaafar. 2012. Optimal inventory control with dual-Sourcing, heterogeneous ordering costs and order size constraints. Prod. Oper. Manag. 21(3): 564–575.

Supporting Information Additional supporting information may be found online in the supporting information tab for this article:

Appendix S1: Online Companion.

Copyright of Production & Operations Management is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.