strategic analysis of dual sourcing and dual … poms dual...strategic analysis of dual sourcing and...

Strategic Analysis of Dual Sourcing and Dual Channelwith an Unreliable Alternative Supplier

Baozhuang NiuSchool of Business Administration, South China University of Technology, No. 381, Wushan Road, Tianhe 510640, Guangzhou,

Guangdong, China, [email protected]

Jiawei LiRoss School of Business, University of Michigan, 1780 Broadway Street, Apt. 309, Ann Arbor, Michigan 48105, USA, [email protected]

Jie Zhang*School of Business Administration, Guangdong University of Finance and Economics, No. 21 Luntou Road, Haizhu District, 510320

Guangzhou, Guangdong, China, [email protected]

Hsing Kenneth ChengWarrington College of Business, University of Florida, P.O. Box 117169, Gainesville, Florida 32611-7169, USA,

[email protected]

Yinliang (Ricky) TanA.B. Freeman School of Business, Tulane University, 7 McAlister DR, New Orleans, Louisiana 70118, USA, [email protected]

I n today’s increasingly interconnected world, co-opetition has emerged as a new business practice among many high-tech firms. The boundaries between cooperation and competition becomes vague, and rivals engage in collaborative

activities. This study develops an analytical model to investigate the dual sourcing decision of the original equipmentmanufacturer (OEM) in the presence of a competitive supplier (i.e., frenemy) as well as a non-competitive supplier whonevertheless suffers from unreliable production yield. We study the competitive supplier’s dual channel decision if it pre-fers operating both component-selling business and self-branded business, and find that the OEM always prefers supplierdiversification even though the additional non-competitive supplier is unreliable. Interestingly, our results reveal that thenon-competitive supplier’s expected profit is unimodal in its production technology level, which suggests the non-compe-titive supplier may not have incentive to improve its production technology once it reaches a threshold. Furthermore, weanalyze the credibility of the competitive supplier’s threat to terminate the supply of the components to OEM as aresponse of OEM’s engagement of a new supplier. We show that this termination of component-selling business by com-petitive supplier is a non-credible threat to prevent OEM from seeking the alternative supplier.

Key words: co-opetitive supply chain; production technology; yield uncertainty; dual sourcing; dual channelHistory: Received: August 2017; Accepted: July 2018 by Subodha Kumar, after 2 revisions.

1. Introduction

Apple and Samsung, two of the largest smartphonevendors in the world, have combined shipping over533 million units in 2017, which is equivalent to 36.3%of worldwide smartphone market (IDC 2018).Although they compete fiercely in the end-user market,where consumers take them as the substitutable choices(Chowdhry 2014), they also in many ways cooperatewith each other. Samsung has supplied processors forApple’s smartphone for many years (Reisinger 2015).According to a recent report on Bloomberg (Lee andKing 2015), Samsung will continue manufacturing themain chips for Apple’s next generation iPhone.

This “frenemy” business relationship betweenApple and Samsung becomes particularly evident inmany high-tech related industries. For example,Sharp who is the leading supplier of the Apple’s iPadand iPhone’s LCD screens, also competes with Applein the smartphone and tablet market. Another classi-cal example is between IBM and Cisco, where IBMbuys Cisco’s network equipment and Cisco also com-petes with IBM directly in the server market (Swartz2009). The above illustrations epitomize the conceptof “co-opetition” (Brandenburger and Nalebuff1996), where two firms cooperate in some activities,and at the same time compete with each other inother activities.

570

Vol. 28, No. 3, March 2019, pp. 570–587 DOI 10.1111/poms.12938ISSN 1059-1478|EISSN 1937-5956|19|2803|0570 © 2018 Production and Operations Management Society

One of the most critical challenges in the co-opeti-tive supply chain is the concern about the sourcingstrategy. From Apple and many other original equip-ment manufacturers’ (OEMs, that is, purchase compo-nents and then produce a new product with its brandname) perspective, it is generally risky to rely solelyon their competitors to provide the key components.One noteworthy problem is the wholesale price deter-mination, where the sole supplier maintains a strongpricing power. According to Forbes (Worstall 2013),Samsung raised the price of a key component sup-plied to Apple by 20% where Apple first disapprovedit, but finding no replacement supplier. It seems natu-ral for the OEM to seek an alternative supplier whowill not compete directly with the OEM in the endmarket by engaging the dual sourcing strategy. How-ever, due to the sophisticated and unstable technol-ogy process, many new entering non-competitivesuppliers suffers the yield loss from manufacturingdefects. Bohn and Terwiesch (1999) have documentedthat high-tech manufacturers such as Seagate experi-enced production yields as low as 50%. Recent mediareports reveal that the yield rates for TSMC1 who pro-vide the fingerprint sensor for Apple is only around70%–80% (Bora 2014). Similar to the previous litera-ture, we capture the notion of unreliable supplier inthe technology industry by proportional randomyield (Tang and Kouvelis 2011, Yano and Lee 1995).Essentially, the OEM can only expect to receive a por-tion of what it orders from the alternative supplier.This yield uncertainty not only hurts the quality ofthe products but also upsets the availability of theproducts. OEM is facing a critical decision on whetheror not to engage an alternative supplier (i.e., non-com-petitive supplier for the remainder of the study) byadopting the dual sourcing strategy. On the one hand,sourcing from an additional supplier may reduce thecomponent price through upstream competition. Onthe other hand, the non-competitive supplier suffersfrom the yield uncertainty which may cause the qual-ity and supply problems for the OEM. The above dis-cussion leads to our first research question: Is itbeneficial for the OEM to seek a non-competitive supplierwho suffers from inferior production technology? And ifso, how does the OEM optimally allocate its componentorders between the competitive supplier and non-competi-tive one?Securing an additional supplier may give the OEM

more leverage when it comes to component procure-ment with its suppliers in the future. Previous litera-ture have also established several benefits of the dualsourcing, such as mitigating the supply risk throughsupplier diversification (Cachon et al. 2008), fosteringthe upstream competition (Chen and Guo 2014), andreducing the inefficiency caused by random yield(Tang and Kouvelis 2011). However, this new

engagement of a new supplier might also irritate thecompetitive supplier who has enjoyed its exclusivitywith the OEM, and thus it is possible that the compet-itive supplier terminates its component-selling busi-ness as a threat. In practice, we observe that Samsunghas considered terminating its LCD supply contractwith Apple when the Apple squeezes the componentwholesale price through supplier diversification fromLG and Sharp (Tibken 2012). Thus, the above exampleoccurring in reality raises some intriguing issues thathave not been well understood in the literature. Howwill the competitive supplier respond to the OEM’s seekingan alternative supplier? Will the competitive supplier ter-minate his component-selling business as a response?In this study, we consider a supply chain associated

with the high-tech industry consisting of an OEM, acompetitive supplier and a non-competitive supplier.We investigate the aforementioned research questionsby considering the following three scenarios: (i) Thebase scenario where the competitive supplier acts asthe sole supplier of the OEM, (ii) The dual sourcingscenario, in which the OEM sources from both thecompetitive supplier and the non-competitive sup-plier, although the non-competitive supplier’s compo-nents have uncertain yield, (iii) The terminationscenario, in which the OEM sources componentssolely from the non-competitive supplier, and thecompetitive supplier generates profits from self-branded business solely. We develop a stylized modelto investigate the incentives of the strategic decisionof each player. The main insights of our research aresummarized as below.First, OEM always prefers the supplier diversifica-

tion although the non-competitive supplier suffersfrom inferior technology and uncertain yield. Essen-tially, by shifting component orders to the non-com-petitive supplier, OEM’s dual sourcing strategybenefits itself by inducing a price war in the upstreamcomponent supply market. The price war drivesdown the component price, which helps the OEM toprocure more components at a lower cost. This givesthe OEM an advantage competing with the competi-tive supplier in the end product market. Although thebenefits of supplier diversification have been wellestablished in other settings (Chen and Guo 2014,Tang and Kouvelis 2011), we have extended thisresult to a co-opetitive supply chain where the bene-fits of supplier diversification come from bothupstream and downstream markets.Second, the non-competitive supplier’s expected

profit is unimodal in its production technology level inthe dual sourcing scenario. That is, the non-competi-tive supplier has a most preferred technology level,and it has no incentive to further improve the produc-tion technology (even it’s costless). The intuition driv-ing this result is as follows: As the non-competitive

Niu, Li, Zhang, Cheng, and Tan: Dual Sourcing with Unreliable SupplierProduction and Operations Management 28(3), pp. 570–587, © 2018 Production and Operations Management Society 571

supplier improves its production technology, initiallythe OEM tends to purchase more components from it.However this allocation will intensify the upstreamcompetition between the non-competitive supplierand competitive supplier, where the latter will cut hiswholesale price as a response. As a result, OEM willgradually reduce her component orders to the non-competitive supplier and shift orders back to thecompetitive supplier. Therefore, we observe that thenon-competitive supplier’s profit is unimodal in itstechnology level.Third, the termination of component-selling busi-

ness by competitive supplier is a non-credible threatto stop the OEM from seeking an alternative supplier.As we have discussed, the OEM always prefers thedual sourcing strategy; however, one of the concernsthat may hinder the OEM seeking the alternative sup-plier lies in the fact that the competitive supplier mayterminate its component-selling business. We showthat termination is never a credible threat by the com-petitive supplier. When the non-competitive sup-plier’s production technology is relatively low, thecompetitive supplier is willing to maintain its collabo-ration with the OEM, because the competition in thecomponent market is not severe. As the non-competi-tive supplier continues to improve its technologylevel, the OEM can ignore the termination threat bythe competitive supplier as the production quality ofthe non-competitive supplier is high enough such thatthe OEM can be better by relying solely on the compo-nents from the non-competitive supplier.The remainder of our study is organized as follows.

In the next section, we first review the most relevantliterature and position our paper with respect to theliterature to highlight our contributions. This is fol-lowed by the introduction of our model setting andthe base scenario in section 3. Sections 4 and 5 investi-gate the dual sourcing scenario and terminationscenario respectively, and we also compare the com-panies’ performances to understand their strategicdecisions. Section 6 presents several extensions of ourmodel where we consider other factors that mayimpact the strategic decisions, which includes capac-ity constraint, alternative demand models andpositive production cost. This study ends withconclusions and avenues for future research. All theproofs are placed in the Appendix S1.

2. Literature Review

Our work is closely related to the studies on unreli-able supply and yield uncertainty problems. Earlyresearches are mostly concerned with optimal pro-duction, procurement and inventory replenishmentin the presence of yield uncertainty. Yano and Lee(1995) provide an excellent comprehensive review for

the earlier literature. Later, there are more studies thatfocus on the mitigation of supply disruption risk bysupplier diversification and dual sourcing. Tomlin(2006) studies the strategies to manage the supply dis-ruption risk when there are two suppliers–one is reli-able but expensive while the other is cheaper but withyield uncertainty. Tang and Kouvelis (2011) assumesuppliers’ products have proportional random yieldand show that two competing buyers’ dual sourcingstrategies may be beneficial by mitigating the channelinefficiency caused by yield uncertainty. They con-sider exogenously given component prices and focuson the value of supplier diversification. Li et al. (2013)investigate a buyer’s supply diversification decisionby assuming the suppliers’ random capacities are cor-related and the buyer adopts a responsive pricingstrategy. They find that the insight “cost is the orderqualifier and reliability is the order winner” holdswith two suppliers but fails to hold with more suppli-ers. Chen and Guo (2014) study an asymmetrictwo-retailer-one-supplier model where one retailer(referred to as the focal firm) can source from both theunreliable supplier and the spot market, while itsrival (referred to as the rival firm) can only sourcefrom the unreliable supplier. They show that the focalfirm’s dual sourcing creates a win–win situation forboth firms. Even if the spot market price is low, strate-gically sourcing from the unreliable supplier can bebeneficial for the focal firm. Tang et al. (2014) charac-terize the buyers’ trade off between sourcing frommultiple suppliers and encouraging their preferredsupplier to reduce the degree of yield uncertainty.They identify the conditions under which dual sour-cing or sole sourcing strategy can be preferred by thebuyers. More recently, Li et al. (2015) consider a set-ting where there exists information asymmetrybetween the two heterogeneous suppliers and a com-mon retailer. They find that the equilibrium contractmenus depend on how much information rent thesupplier may need to pay.Compared to the aforementioned works, especially

the two most related papers—Tang and Kouvelis(2011) and Chen and Guo (2014), we investigate a co-opetitive supply chain in which the competitive sup-plier serves as both the buyer’s upstream businesspartner and downstream competitor. The buyer cansource from a unreliable alternative supplier, whosewholesale price is endogenously determined. Wefocus on the channel members’ strategic decisionswith respect to the buyer’s adoption of dual sourcingstrategy, the competitive supplier’s incentives of mar-ket withdrawal, and the non-competitive supplier’spreference of the degree of yield uncertainty. We notethat Tang and Kouvelis (2011) study a chain-to-chaincompetition model without the consideration of co-opetition issues, and Chen and Guo (2014) study a

Niu, Li, Zhang, Cheng, and Tan: Dual Sourcing with Unreliable Supplier572 Production and Operations Management 28(3), pp. 570–587, © 2018 Production and Operations Management Society

Hotelling model where the product wholesale pricesare exogenously given. With endogenized wholesaleprices, we show that dual sourcing does not necessar-ily sustain as the channel members’ win–win strategy.Our work is also closely related to the studies on

co-opetition in supply chain. This stream is originatedfrom the literature of dual channel management ineconomics (Spiegel 1993) and the early OM/IS inter-face (Tsay and Agrawal 2004). While early literaturefocuses mostly on the impact of a direct channel tothe related firms and the supply chain, more recentstudies investigate the supplier’s incentives to estab-lish a direct channel and the strategic interactionbetween the supply chain members, and thenthe stream of literature on co-opetition graduallyemerges. Kumar and Ruan (2006) assume customersare either brand-loyal or retailer-loyal, and show thatthe supplier has incentives to open a direct channeland hence, operates a dual channel business model.Dumrongsiri et al. (2008) assume consumers are bothprice and service quality sensitive, and examine themanufacturer’s incentives regarding direct and retailchannels. They show that, the channel cost difference,the demand variability, and the channel centralizationdegree greatly influence the manufacturer’s decisionsof opening a direct channel. Cai (2010) identifies achannel-adding Pareto zone and a contract-imple-menting Pareto zone in two single-channel and twodual-channel supply chains. In the former zone, boththe retailer and supplier have profit improvementswhen the supplier opens a direct channel. In the latterzone, the value of contract coordination is derived.Wang et al. (2013) investigate the timing issue of twofrenemies’ quantity decisions by solving an endoge-nous timing game. They find that the OEM tends tosource solely from a competitive contract manufac-turer (CM) regardless the downstream competition,

and the competitive CM tends to generate profitsfrom both contract manufacturing and self-brandedbusinesses. Recently, Adner et al. (2015) develop agame-theoretic model to explain the incentives fortwo platforms to become frenemies when the differ-ence in their profit foci is sufficiently large.Similar to Dumrongsiri et al. (2008) and Cai (2010),

we derive the outcomes in each scenario and thencompare them to analyze the supply chain parties’incentives toward alternative channel structures.However, different from the existing works, we con-tribute by considering the OEM’s dual sourcing strat-egy when it faces a competitive supplier and has theoption of sourcing from an unreliable alternative sup-plier, and analyzing the strategic interactions betweenthe OEM and the competitive supplier.

3. Model Settings and Benchmark

3.1. Notations and AssumptionsWe consider a three-player game comprising a com-petitive supplier (CS), a non-competitive supplier(NS) and an OEM where the NS does not competedirectly with the OEM in the end market. Both com-petitive supplier and non-competitive supplier arecapable of producing a key component, which is thenused to produce the end products for the consumers.Similar to Amaral et al. (2006), Chen et al. (2012), andWang et al. (2013), we define OEM as a company thatpurchases components and then finalizes a new pro-duct with its brand name. For the remainder of thestudy, we use the pronoun “he” to represent the com-petitive supplier (cs), “she” to represent the OEM (o)and “it” to represent the non-competitive supplier(ns). Based on the motivations discussed in introduc-tion, we analyze the following three scenarios, whichare illustrated in Figure 1.

Figure 1 Illustration of Three Different Scenarios


(1) The base scenario, in which the competitivesupplier serves as the sole supplier of theOEM.

(2) The dual sourcing scenario, in which theOEM sources from both competitive supplierand non-competitive supplier, although thenon-competitive supplier suffers from yielduncertainty.

(3) The termination scenario, where the OEMsources components solely from the non-com-petitive supplier. This may either result fromthe OEM’s choice when it shifts all the ordersto the non-competitive supplier, or the com-petitive supplier’s choice when it refuses tosell components to the OEM.

Without loss of generality, we assume that both thecompetitive supplier and OEM employ one unit ofcomponent to assemble one unit of end product. Wefocus on those industries where the retail prices aremainly determined by the supply quantities. Typicalexamples include influenza industry (Deo and Cor-bett 2009), microchip industry (Tang and Kouvelis2011), and smartphone industry in which Samsungand Apple compete (Autrey et al. 2014, Karp andPerloff 2012). Thus, we assume that the competitivesupplier and the OEM engage in a Cournot-typedcompetition in the end-user market.In particular, similar to Tang and Kouvelis (2011),

the consumer demand for the competitive supplier’sproduct is represented by a linear, downward slop-ing, (inverse) demand function Pcs = a � bQ, wherePcs is the retail price, a is the market potential, b repre-sents the quantity sensitivity and Q denotes the totalquantities available on the market. Without loss ofgenerality, we assume that the market potential a islarge enough such that the market demand is alwayspositive (Wu and Zhang 2014). To capture the branddistinction between the products from the OEM andthe competitive supplier, we assume that the demandfor OEM’s product is Po = a � bQ + m (m ≥ 0 anda ≥ m). Essentially, the OEM’s products enjoy a pre-mium perception from the consumers (Arru~nada andV�azquez 2006). For instance, consumers are willing topay a higher price for ThinkPad laptop (i.e., OEM)than Asus laptop (i.e., CS) although the configurationof the two computers are almost identical. This modelis originated from the Cournot competition model fordifferentiated goods (Singh and Vives 1984). Withoutloss of generality, we also normalize the productioncosts of the competitive supplier, the non-competitivesupplier, as well as the OEM’s assembly cost to bezero.2 In section 6.4, we analyze the situation in whichthere is a positive cost for the production in both thecompetitive and the non-competitive supplier, andwe find that our main findings are robust.

When the OEM sources from a non-competitivesupplier, she has to be concerned about its yielduncertainty problem. That is, for an order of size q,the realized delivered quantity is eq, where e is a ran-dom variable with mean l and variance r2 (Yano andLee 1995). We restrict the support of e to be [0, 1] andr < 1/2.3 It’s worth noting that the increase of theexpected yield l and/or the decrease of the yield vari-ance r2 can be interpreted as “quality improvement.”In practice, these improvements can be achievedthrough better process management and investmentin new technology. In the presence of the yield uncer-tainty problem, firms can only form an expectation oftheir own and the rival’s profits. We assume that allsupply chain members are risk-neutral profit maxi-mizers. We also assume that the OEM pays for whatshe actually receives rather than what she orders. Thisis consistent with the industrial practice and previousliterature.4 Nevertheless, we show that these twoassumptions lead to almost identical results.For the remainder of this study, we incorporate

superscripts on the optimums: B, D, T to denote thebase scenario, the dual sourcing scenario, and thetermination scenario, respectively. For example, PB

cs

stands for the competitive supplier’s optimal profitin the base scenario. We denote qo (qb) as the OEM’s(the competitive supplier’s) production quantity inthe end market, qcs as the component order quantityallocated to the competitive supplier, and qns as theorder quantity to the non-competitive supplier. Inaddition, Wcs represents the wholesale pricebetween the competitive supplier and OEM, andWns represents that between the non-competitivesupplier and the OEM.

3.2. The Benchmark: Base ScenarioIn this scenario, the OEM sources solely from thecompetitive supplier. The event sequence is describedas follows: First, the competitive supplier determinesthe unit wholesale price Wcs. Second, the OEM placesan order to the competitive supplier, who also deter-mines the production quantity for his self-brandedproducts. Third, the OEM receives the componentsand both firms assemble the components to the endproducts. Finally, products are sold to consumers atmarket clearing price. Note that, the OEM’s produc-tion quantity in the end market, qo, exactly equals herorder quantity placed to the competitive supplier, thatis, qo = qcs, because the competitive supplier does nothave the yield uncertainty problem.Thus, the competitive supplier and the OEM’s

profit functions are given as:

Pcs ¼ ða� bqb � bqcsÞqb þWcsqcs;

Po ¼ ða� bqb � bqcs þmÞqcs �Wcsqcs:


The competitive supplier’s profit comes from thefollowing two sources: (i) component-selling busi-ness; (ii) self-branded business. If he competes withthe OEM intensively in the downstream market, itwill not only squeeze the OEM’s selling quantity butalso reduces the competitive supplier’s own revenuefrom component-selling business. Hence, the OEMand the competitive supplier have a co-opetitive rela-tionship, under which the competitive supplier has tobalance his revenue from these two sources. Theresults are summarized in Lemma 1.

LEMMA 1. In the base scenario, where the OEM sourcessolely from a competitive supplier, the equilibrium whole-

sale price is WBcs ¼ 5aþ 4m

10 , the production quantities are

qBb ¼ 5a� 2m10b , qBo ¼ qBcs ¼ 2m

5b , and the supply chain par-

ties’ profits are PBcs ¼ 5a2 þ 4m2

20b , PBo ¼ 4m2

25b .

Clearly, the competitive supplier’s productionquantity is decreasing in m, while the OEM’s isincreasing in m. This result is in line with intuition,because the OEM’s brand image helps her productsdifferentiate from the competitive supplier’s. We alsofind that the equilibrium wholesale price is increasingin m, but the selling quantity qBo still increases in mbecause the OEM’s profit margin increases. That is,the positive effect (i.e., larger selling quantity) due tothe increase of m outweighs the negative effect (i.e.,higher wholesale price) in this scenario. Meanwhile,the competitive supplier’s overall profit is alsoincreasing in m, indicating that his profit loss fromself-branded business is compensated by its profitgains from component-selling business. When mincreases, the competitive supplier gains from ahigherWB

cs and sells more components qBcs. On the con-trary, if m approaches to 0, that is, the OEM’s producthas little brand advantage toward its competitor, thenthe OEM only gains very limited profits and canhardly survive in the market.

4. The Dual Sourcing Scenario

In this scenario, the OEM adopts dual sourcing strat-egy and allocates her component orders to both thecompetitive supplier and the non-competitive sup-plier. The event sequence is as follows: First, both thecompetitive supplier and the non-competitive sup-plier determine their component wholesale prices.Second, after observing both wholesale prices, theOEM and the competitive supplier determine theirproduction quantities simultaneously. Note that theOEM’s orders are allocated between the competitiveand the non-competitive suppliers, which are labeledas qcs and qns respectively, and her total received com-ponent quantity is qo = qcs + eqns, where the non-

competitive supplier suffers the yield uncertaintyproblem. Third, the OEM and the competitive sup-plier finalize the components into end products anddeliver them at the market-clearing price po and ps.The profit functions of the competitive supplier, theOEM and the non-competitive supplier are then:

Pcs ¼ ½a� bqb � bðqcs þ eqnsÞ�qb þWcsqcs;

Po ¼ ½aþm� bqb � bðqcs þ eqnsÞ�ðqcs þ eqnsÞ�Wcsqcs �Wnseqns;

Pns ¼ Wnseqns:

In the quantity-decision stage, both the competitivesupplier and OEM evaluate their expected profits inanticipation of the non-competitive supplier’s yielduncertainty. We summarize the optimums in the fol-lowing Lemma 2.

LEMMA 2. In the dual sourcing scenario, where theOEM allocates her component orders to competitive andnon-competitive suppliers, the equilibrium wholesale

prices are WDcs ¼ 4ð5aþ 4mÞr2

40r2 þ 27l2 , WDns ¼ 2ð5aþ 4mÞr2

40r2 þ 27l2 ; the

order quantities are qDb ¼ 4ð5a� 2mÞr2 þ 9ða�mÞl2bð40r2þ 2 7l2Þ , qDcs ¼

16mr2 þ 4al2 þ 14ml2

bð40r2 þ 27l2Þ , and qDns ¼ ð5aþ 4mÞlbð40r2 þ 27l2Þ. Correspond-

ingly, the equilibrium expected profits of the supply chainmembers are,

EPDns ¼

2ð5aþ4mÞ2r2l2bð40r2þ27l2Þ2 ;

EPDcs ¼

m2ð320r4þ368r2l2þ81l4Þ�2aml2

ð80r2þ81l2Þþ a2ð400r4þ440r2l2þ81l4Þbð40r2þ27l2Þ2 and

EPDo ¼

a2l2ð25r2þ81l2Þþ4aml2ð82r2þ81l2Þþ4m2ð64r4þ148r2l2þ81l4Þ

bð40r2þ27l2Þ2 ;

respectively.

For the remainder of this study, we denotex = (l/r)2 as the technology level of the non-competitive supplier, which is an indicator of thequality in its production process, and almost all thedecisions/outcomes can be represented as functionsof x. It’s straightforward to see that the increase of land/or the decrease of r leads to a higher level of x.Essentially, the higher the technology level x is, themore reliable the non-competitive supplier’s produc-tion process are. Suppliers can improve their produc-tion process not only through costly R&D innovationsbut also by inexpensive administrative efforts. For


example, Snow et al. (2006) have documented that thebio-technology firm Genentech worked very hard toimprove its yield through “monitoring the raw mate-rials, limiting human involvement in production, test-ing frequently and ensuring that all connectionsbetween pieces of equipment were tightly sealed.”Next we use x to rearrange the supply chain parties’profits shown in Lemma 2, which leads to the follow-ing two interesting results through sensitivity analy-sis. The results are also illustrated in Figure 2.

PROPOSITION 1. When the OEM adopts the dual sourcingstrategy, the competitive supplier’s expected profit decreasesin x while the OEM’s expected profit increases in x.

Our findings toward the profits of the competitivesupplier and the OEM are in line with expectation:The higher the technology level of the non-competitivesupplier is, the more intense the competition betweenthe suppliers will be. This reduces the competitive sup-plier’s profit from component-selling business. For theOEM, it always benefits from the non-competitive sup-plier’s production quality improvement, and the mainreason is the component price war induced betweenthe suppliers. Taking a closer look at the wholesaleprices and quantities of the competitive supplier, wehave the following comparative statics:

@WDcs

@x\0;

@qDcs@x

[ 0;@qDb@x

\0:

When the technology level x increases, the non-competitive supplier becomes a threat to the

competitive supplier as the non-competitive supplierhas a relative price advantage (WD

ns \WDcs ). Its

improvement on production quality increases thecompetitive supplier’s pressure significantly andhence, Wcs and WD

ns both decrease. That is, the com-petitive supplier’s component price has to be low-ered along with the non-competitive supplier’s. As aresult, from Proposition 1, we conclude that a pricewar between the suppliers can be successfullyinduced by the OEM’s dual sourcing strategy. Whenthe competitive supplier lowers WD

cs , to respond, weobserve the increase of qDcs. That is, the OEM shiftssome component orders back due to the competitivesupplier’s price undercutting behavior.Regarding the OEM, when x increases, the lowered

average component wholesale price provides her a lar-ger cost advantage in the downstream market, becauseshe can procure more components from the competi-tive supplier at a lower cost. Eventually, the competi-tive supplier’s self-branded business is hurt, andtherefore qDb decreases in x. That is, the OEM benefitsfrom both the suppliers’ price war and the resolvedyield uncertainty at the non-competitive supplier, andthus she prefers a higher value of x. However, the com-petitive supplier’s loss from downstream self-brandedbusiness prevails his gains due to component orderincrease in the component-selling market, and thus hesuffers from the non-competitive supplier’s improve-ment of it technology level, x. Then, we reach the fol-lowing result: the OEM is strictly better off while thecompetitive supplier is strictly worse off in the dualsourcing scenario, because the base scenario can be

0 1 2 3 4 50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Prof

it

0 1 2 3 4 50

0.01

0.02

0.03

0.04

0.05

0.06

Prof

it

OEM

Competitive Supplier

Non−competitive Supplier

Non−competitive Supplier

Figure 2 Illustration of Impact of x on Profits of Each Supply Chain Party [Color figure can be viewed at wileyonlinelibrary.com]


www.wileyonlinelibrary.com

regarded as a special case when x = 0 and the non-competitive supplier is completely incapable. Thisfinding is consistent with the existing literature of dualsourcing and dual channel.

PROPOSITION 2.

1. The non-competitive supplier’s expected profit EPDns

is unimodal in x;2. WD

ns is decreasing in x; qDns is unimodal in x for allgiven feasible r.

We then study the non-competitive supplier’s pref-erence of x. Conventional wisdom suggests that thenon-competitive supplier should improve its technol-ogy level as much as possible in order to attract moreorders from the OEM. However, our findings towardthe non-competitive supplier’s preference of x israther surprising: Its profit is unimodal in its technol-ogy level x. In other words, the non-competitive sup-plier has a most preferred technology level, and it hasno incentives to further improve the production tech-nology, if it still can.To understand this finding, we further investigate

how the quantities and the wholesale prices change inthe technology level x. Interestingly, we find that qDnsis non-monotone in x in the support of a given r. Thisis the key reason to explain why the profit of the non-competitive supplier is unimodal in x. When the tech-nology level x improves, the OEM tends to shift herorder to the non-competitive supplier, that is, qDnsincreases. Then, the competitive supplier faces a

fiercer competition in the component selling market,and its optimal response is to reduce WD

cs . As a result,the non-competitive supplier will reduce WD

ns corre-spondingly, but the price war becomes less intensebecause the difference between the prices becomessmaller. In anticipation of the competitive supplier’sresponse, the OEM may be willing to shift some ofher order back to the competitive supplier when x isat a high level. That is, we observe an increasing qDcsand a decreasing qDns. The intuition is as follows: Whenthe non-competitive supplier’s quality is approachingto the competitive supplier’s, although WD

ns keepsdecreasing, the price advantage of the non-competi-tive supplier is not very significant. Therefore, theOEM’s gains from component price war become lim-ited, and is more concerned about her gains from thedownstream market, where the competitive supplieris her major competitor. It has been illustrated by pre-vious literature that, placing more orders to a rivalcan limit the rival’s incentives to develop the self-branded business (Spiegel 1993, Wang et al. 2013),and hence, reduce downstream competition. There-fore, to generate more profits from the downstreammarket, we observe an increasing qDcs and a decreasingqDns, even if WD

ns is further decreasing and x is increas-ing. Being aware of this, to snatch more componentorders, the non-competitive supplier has no otherchoices but to further lower its wholesale price WD

ns,though its production quality is already high. Weillustrate the above results in Figure 3.If we think the competitive supplier’s products as

“luxury goods” with high price elasticity for its high

0 1 2 3 4 50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Who

lesa

le P

rice

0 1 2 3 4 5

0.5

1

1.5

2

2.5

Qua

ntity

Figure 3 Illustration of Impact of x on Wholesale Prices and Quantities [Color figure can be viewed at wileyonlinelibrary.com]



quality, and the non-competitive supplier’s productsas “economy substitutes” with low price elasticity forits inferior quality, then it’s easy to understand thatthe marginal order increase due to price reduction of“luxury goods” will be more obvious. This also helpsexplain the finding that qDns decreases while qDcsincreases when the suppliers’ wholesale prices areboth decreasing. In summary, we find that when x issmall, the OEM shifts orders to the non-competitivesupplier as x increases, and then it has incentive ofimproving technology. When x becomes large, theOEM tends to shift orders back to the competitivesupplier. As a result, the non-competitive supplierwill lose some orders even if it has further reducedthe wholesale price due to the intensive upstreamcompetition. Combining the foregoing forces, thenon-competitive supplier has no incentive to furtherimprove its production technology when its most pre-ferred quality level has been already achieved.We also conduct the sensitivity analysis with respect

to the price premium m. Most insights are similar tothose in the base scenario: The wholesale prices isincreasing in m. When m is increasing, the competitivesupplier’s production quantity of his own product willdecrease, and the OEM’s ordering quantities from bothsuppliers as well as her profit will increase. With analternative supplier, the OEM can always survive inthe market, even if m = 0, because now its supply isnot fully controlled by its competitor. A different resultis that, the competitive supplier’s profit is decreasingin mwhen it is smaller than a threshold, but increasingin m when it is sufficiently large. The driving forcecomes from the price war induced by the OEM’s dualsourcing strategy, which weakens the positive impactof an increasing WD

cs and strengthens the negativeeffect of a decreasing qDb , when m is increasing. By con-trast, the competitive supplier’s pricing power ensureshim to gain more from an increasing m in the basemodel, even if qBb is decreasing in m.

5. The Termination Scenario

In this scenario, the OEM sources solely from the non-competitive supplier. The event sequence is as fol-lows. First, the non-competitive supplier determinesthe component wholesale price Wns. Second, the com-petitive supplier and the OEM determine their pro-duction quantities simultaneously. Third, both firmsassemble the components to end products. Finally,products are sold at market clearing prices. Note thatthe OEM’s production quantity of the end productsis qo = eqns, where the non-competitive supplierbecomes the sole supplier for the OEM.The profit functions of the competitive supplier,

the OEM and the non-competitive supplier are,respectively

Pcs ¼ ða� bqb � beqnsÞqb;Po ¼ ðaþm� bqb � beqnsÞeqns �Wnseqns;

Pns ¼ Wnseqns:

The equilibrium decisions and outcomes are sum-marized in Lemma 3.

LEMMA 3. In the termination scenario, where the OEMcan only source from the non-competitive supplier, the

equilibrium component wholesale price is WTns ¼ aþ 2m

4 ;

the supply chain parties’ production quantities are qTb ¼8ar2 þð5a� 2mÞl2

16br2 þ 12bl2 , and qTns ¼ ðaþ 2mÞl8br2 þ 6bl2; and the supply chain

parties’ profits are EPTcs ¼ ð8ar2 þð5a� 2mÞl2Þ2

16bð4r2 þ 3l2Þ2 , EPTo ¼

ðaþ 2mÞ2l2ðr2 þl2Þ4bð4r2 þ 3l2Þ2 , EPT

ns ¼ ðaþ 2mÞ2l28bð4r2 þ 3l2Þ respectively.

Again, we find that the OEM’s production quantityis increasing while the competitive supplier’s produc-tion is decreasing in the brand difference m. TheOEM’s profit as well as the non-competitive sup-plier’s profit both increase in m. Without the compo-nent-selling business, there is no tradeoff and thecompetitive supplier’s profit will decrease in m.Regarding the impact of the non-competitive sup-plier’s production technology level, our findings areas follows: (i) the wholesale price WT

ns is a constant.(ii) qTb is deceasing in x, which indicates that the com-petitive supplier, acting as a pure competitor of theOEM, will be hurt when OEM finds a high qualityalternative supplier. (iii) qTns is unimodal in x for allgiven r, which implies that the OEM’s order quantitymight be decreasing in x when x is sufficiently high.(iv) EPT

o is increasing in x, because when the non-com-petitive supplier has a high quality level, although theordering quantity at the non-competitive supplier qTnsis lower but the actual received component quantitylqTns is higher.Performance comparison between the termination and

base scenarios. Shifting from the base scenario to thetermination scenario may benefit or hurt the OEM,because the non-competitive supplier can provide thecomponents to the OEM at a lower price but at thesame time it has yield uncertainty. As for the competi-tive supplier, he has an incentive to terminate hiscomponent supply business because he competeswith the OEM directly in the end-user market. We areinterested in the changes of performances of the OEMand the competitive supplier from the base scenarioto the termination scenario. The results are summa-rized in the following proposition.

PROPOSITION 3. The competitive supplier is strictlyworse off in termination scenario than that in the basescenario; The OEM is better off under the termination


scenario when x exceeds a threshold value xO, otherwiseit prefers the base scenario.

Note that the competitive supplier becomes thepure competitor of the OEM in the termination sce-nario. Compared with the base scenario, we find thatthe competitive supplier’s strategy of terminating hiscomponent-selling business will indeed hurt the OEMwhen x < xO. This indicates that the competitive sup-plier’s termination decision of the component-sellingbusiness can induce a lose–lose situation in the down-stream competition. This situation no longer holdswhen x ≥ xO, where we find that the OEM can be bet-ter off in the termination scenario even when the non-competitive supplier’s production technology is notvery stable. Essentially, this is because that in the ter-mination scenario, the wholesale price offered by thenon-competitive supplier is always lower than theprice offered by the competitive supplier in the basescenario, that is, WT

ns \WBcs. When the technology

level x is low, the cost advantage of the OEM in thetermination scenario is not strong enough to offset thedrawback brought by the low production technology.However, as x increases, the drawback in productionquality becomes less significant, and eventually whenx surpasses xO, the OEM becomes better off in the ter-mination scenario.Proposition 3 establishes the fact that the competi-

tive supplier always benefits from engaging thecomponent-selling business with OEM and termi-nating the component supply will shrink OEM’sprofit when x < xO. We have also shown that theOEM has strong incentives to diversify her supplysources. Thus, an intriguing and important questionarise naturally, if the OEM wants to adopt the dual-sourcing strategy by purchasing from an additionalunreliable supplier, can the competitive supplierprevent that by the threat of terminating his supplyof components? Next, we will investigate this ques-tion by comparing the supply chain parties’ perfor-mances under the termination and dual sourcingscenarios.Performance comparison between the termination and

dual sourcing scenarios. This comparison can be themost interesting one—as we have mentioned in theintroduction, emerging debates have recently takenplace in the OM/marketing field (Epstein 2014, Kaiser2013, Kim 2012, Worstall 2013, Humphries 2012).Essentially, the OEM prefers the situation where thecompetitive supplier does not terminate the compo-nent-selling business even if she shifts some compo-nent orders to a new alternative supplier. Especiallywhen x < xO, the supplier’s termination threat seemscredible and indeed hurts the OEM’s profitability.The results of our analysis are summarized in the fol-lowing proposition.

PROPOSITION 4. Comparing the profits in the dual sour-cing scenario and termination scenario,

1. the OEM is always better off in the dual sourcingscenario;

2. the competitive supplier is better off in the dualsourcing scenario when x is smaller than athreshold value xC, otherwise the competitivesupplier prefers the termination scenario.

We find that the incentives of the OEM and thecompetitive supplier can be aligned when the newlyentered non-competitive supplier’s technology levelis lower than the threshold value, xC, such that bothOEM and competitive supplier prefer the dual sour-cing strategy to the termination scenario. For the OEM,we can conclude that the dual sourcing is its most pre-ferred strategy as it generates the largest profit amongthe three scenarios. An immediate insight is that theOEM will always benefit from the newly entered alternativesupplier even it suffers from technology inefficiency.That is, the OEM has no incentive to place all compo-nent orders to a sole supplier.For the competitive supplier, intuitively speaking,

it’s more profitable to engage both component-sellingbusiness and self-branded business (Wang et al. 2013).We find that this intuition only holds when the compe-tition in the upstream component market is not veryintense. That is, when the non-competitive supplier’sproduction technology is relatively low, the reductionin component wholesale price is not very significant.Furthermore, engaging in self-branded business willcontribute to the competitive supplier’s profit throughrevenue diversification. In practice, many technologi-cal firms have realized this and producing and sellingself-branded products, such as Samsung and Intel.However, we find that this intuition breaks when thealternative non-competitive supplier’s productiontechnology is sufficiently high. The resulting fierceprice war will hurt the competitive supplier while pas-sing the gains to OEM. On the one hand, the competi-tive supplier’s benefit from component-sellingbusiness diminishes due to the intense competition inthe component market. On the other hand, the OEMbenefits from the component price war which strength-ens her business of the end product and hurts the com-petitive supplier’s self-branded business. Being awareof this, it’s worthwhile for the competitive supplier toterminate his component-selling business so as toincrease the component costs for OEM and squeezeOEM’s margin. This also benefits the competitive sup-plier’s self-branded business. Our finding here helpsexplain Samsung’s recent termination of LCD contractwith Apple (Humphries 2012).We now turn our attention to the critical but open

question: Is the termination of component-selling


business by the competitive supplier a credible threat forthe OEM? In other words, when the OEM tries toadopt the dual sourcing strategy, is it possible thatthe competitive supplier terminates his supply ofcomponents as a response such that the OEM earnseven less than in the base scenario? If this is true, theOEM has to consider more carefully when solicitingthe alternative supplier. In Proposition 3, we showthat it is possible that the OEM becomes worse off inthe termination scenario than in the base scenario,and there exists a threshold value xO. In order todetermine whether this termination is a crediblethreat, we compare the two critical threshold value:xC and xO. The result is provided in the followingproposition.

PROPOSITION 5. For any given parameters of a, m, l andr, xO < xC.

We illustrate this proposition in Figure 4. A directimplication of Proposition 5 is that when the technologylevel x is strong enough to convince the competitive sup-plier to terminate his component-selling business, it is alsostrong enough to enable the OEM to be better off in the ter-mination scenario than in the base scenario. We summar-ize the strategic interactions between the OEM andthe competitive supplier as follows. From the compe-titive supplier’s perspective, when the OEM conductsdual sourcing strategy and when x is relatively low,the competitive supplier is still willing to cooperatewith the OEM since the competition in the componentmarket is not very intense. As x gradually increases,the upstream competition in the component marketbecomes fiercer, and eventually when x surpasses xC,it’s actually better for the competitive supplier to ter-minate his component-selling business. However,Proposition 5 indicates that it’s already “too late” forthe competitive supplier to do so. The productionquality of the non-competitive supplier is so high thatthe OEM can be better off by relying solely on thecomponents from the non-competitive supplier,

although the yield uncertainty is not yet completelysolved.From the OEM’s perspective, dual sourcing is

always the preferred strategy. When x is relativelylow, the competitive supplier is willing to continuethe component-selling business with OEM, hencetheir incentives are coordinated and the dual sourcingstrategy can be successfully adopted. As the technol-ogy level x gradually increases and finally surpassesxC, the component business will be terminated by thecompetitive supplier due to the intense competitionin the component market. However, the OEM stillbecomes better off compared to the base scenario.With the foregoing analysis, we can now concludethat: Termination is never a credible threat by the competi-tive supplier. It’s always optimal for the OEM to seekan alternative supplier (even with lower productiontechnology) and adopt the dual sourcing strategy. Nomatter the competitive supplier accepts it or not, theOEM will always be better off compared with the ori-ginal (base) sourcing strategy. In summary, we havethe general equilibrium of the supply chain structuresin the following Table 1.From the above discussions, dual sourcing is

always OEM’s dominating strategy and the competi-tive supplier’s best response depends on x: if x < xC,he chooses to maintain the co-opetition relationshipwith the OEM; otherwise when x ≥ xC, he will chooseto terminate its component-selling business andbecome a pure competitor of the OEM. Therefore,“Dual Sourcing” and “Termination” can be the equili-brium structure in general, depending on the alterna-tive supplier’s technology level.

Figure 4 Illustration of Performance Comparison [Color figure can be viewed at wileyonlinelibrary.com]

Table 1 General Equilibrium of Supply Chain Structures

Competitive supplier

Co-opetition Competition

OEM Single Base N/ADual Dual sourcing (x < xC) Termination (x ≥ xC)



6. Extensions and Discussions

6.1. Capacity Constraint of theCompetitive SupplierIn practice, the capacity of technological firms istypically limited and difficult to adjust oncedetermined in the short term. In this subsection,we explore how the capacity constraint of thecompetitive supplier influences our results in thedual sourcing scenario. Let K denote the competi-tive supplier’s capacity. As long as K � qDb þ qDcs,it is obvious that the competitive supplier willfully utilize its capacity and satisfy all the compo-nent orders by adjusting the wholesale priceaccordingly. That is, qDK

b þ qDKcs ¼ K, where the

superscript “K” is used to denote the equilibriumwith limited capacity. Then, we obtain the equi-librium and outcomes similar to Lemma 2, butwe omit the lengthy formulations here for exposi-tion. We find that the competitive supplier has amost preferred level of K because his profit isconcave in K. Furthermore, this “most preferred”capacity level is smaller than qDb þ qDcs, the equilib-rium output of the competitive supplier withoutcapacity constraint. That is, the competitive sup-plier actually benefits from insufficient capacity.This finding is presented in the followingproposition.

PROPOSITION 6. Let KDKcs denote the most preferred

capacity level that maximize EPDKcs , then KDK

cs \qDb þ qDcs.

From Proposition 6, we find that an insufficientcapacity may benefit the competitive supplier. Thekey to understand this finding is the yield uncer-tainty problem of the non-competitive supplier. Ifthe non-competitive supplier has 100% yield rate, allunmet demand will be immediately shifted to thenon-competitive supplier, and the component priceswill remain unchanged. However, due to yielduncertainty, the OEM can only shift a partial order.This enhances the competitive supplier’s negotiationadvantage, and enables him to raise the wholesaleprice. In other words, the capacity constraint actuallyalleviates the price war in the component market.Thus, a small K benefits the competitive supplier andharms his rival, the OEM. This explains whyKDKcs \ qDb þ qDcs. As a result, with a tight capacity

constraint the total component ordering quantity isreduced. One recent example supporting our findingis that Samsung outsmarted Apple by raising chipprices using limited production capacity. Accordingto a report from Financial Times, Samsung follows asimilar strategy when it sold application processorsto Apple (Song 2012).

We also investigate the impact of the non-competi-tive supplier’s production technology level x on itsprofit when its competitor does not have sufficientcapacity. It is not difficult to show that EPDK

ns is stillunimodal in x. However, we find that the thresholdbecomes even smaller than the previous one, that is,the non-competitive supplier is less willing toimprove its technology if its competitor has limitedcapacity. This is because when the competitive sup-plier has limited capacity, the OEM has to rely on thenon-competitive supplier to a higher degree, whichreduces the latter’s incentive of improvement.Furthermore, in section 4, we have shown that the

OEM is always better off while the competitive sup-plier is always worse off in the dual sourcing scenariocomparing with the base scenario. We now investi-gate if this result changes when there is a capacityconstraint for the competitive supplier.

PROPOSITION 7. When the competitive supplier’s produc-tion quantity is constrained by a limited capacity K,

1. There exists a threshold Ko ¼ 14a�mþ 4ða�mÞxbð29þ 12xÞ such

that, the OEM is better off in the dual sourcingscenario when K > Ko, and vice versa;

2. There exists a threshold Kc ¼ 12aþ 4mþð5aþmÞxbð28þ 15xÞ such

that, the competitive supplier is worse off in thedual sourcing scenario when K > Kc, and viceversa;

3. Kc > Ko if and only if x [ 11a� 36m24m .

When the competitive supplier has limited capac-ity, the performance comparison becomes more com-plicated and depends on the relationship among a, m,and x. We illustrate this proposition in Figure 5 andhave the following observations: (i) when the compet-itive supplier has very limited capacity, the perfor-mance comparison is just the reverse of traditionalwisdom: the OEM is worse off while the competitivesupplier is better off in the dual sourcing scenario.This is because, the component price war will bringlimited gains to the OEM, and protect the competitivesupplier’s component-selling business. The OEM hasto procure more components from the low-qualityalternative supplier at the average high purchasingprice, and thus its profitability is hurt. (ii) Interest-ingly, when the competitive supplier has moderatebut limited capacity, dual sourcing may result ineither a win–win situation or a lose–lose situation,depending on the OEM’s price premium m and thenon-competitive supplier’s quality x. If the non-com-petitive supplier has a high quality, and/or theOEM’s premium is small, both the competitive sup-plier and the OEM obtain more profits in the dualsourcing scenario; otherwise, both companies preferthe base scenario.


6.2. Capacity Constraint of the Non-competitiveSupplierWe now explore how the capacity constraint of thenon-competitive supplier influences our results. Theexistence of this constraint means that the OEM can-not mitigate the yield uncertainty problem by simplyincreasing its order quantity qns. Let s denote the non-competitive supplier’s capacity. Similarly, it is obvi-ous that this constraint is binding when s � qDns. Withthis constraint, l and r are not homogeneous in theformulations and thus they cannot be rearranged asfunctions of the technology level x. As a result, weexamine the comparative statics for l and r respec-tively. Let the superscript “t” denote the correspond-ing equilibrium and outcomes. We have the followingproposition:

PROPOSITION 8. When the non-competitive supplier’stotal output is constrained by s,

1. WDtcs and qDt

b both decrease in l and increase in r;2. qDt

cs decreases in r, and increases in l if x < 40/9;

3. EPDtcs decreases in l and increases in r; EPDt

o

increases in l and decreases in r;4. EPDt

ns is unimodal in l and/or r.

Although the comparative statics become more com-plicated, they share the similar qualitative insights intoour previous results. The two decisions made by thecompetitive supplier, WDt

cs and qDtb are negatively

related to its competitor’s technology levels. When thecompetitive supplier’s wholesale price decreases, it isintuitive that the OEM should place a larger order tohim, but now the effects of l and r become diverse.The intuition holds when r decreases, but is not neces-sarily true when l increases, unless the technologylevel is not very high. When the non-competitive sup-plier is constrained by the capacity, the effectiveexpected quantity that the OEM can obtain is exactlyls. Therefore, the qualified component quantity fromthe non-competitive supplier that the OEM can obtainis increasing in l, but she does not necessarily increaseher order size to the competitive supplier.

(a)

(b)

Figure 5 Illustration of Performance Comparison with Capacity [Color figure can be viewed at wileyonlinelibrary.com]



Regarding the non-competitive supplier’s decisionsand outcomes, we find that similar insight holds: Itmay not have incentives to improve the technologylevels, that is, increase l or decrease r. We show that,its profit is quasi-concave in both l and r, that is, (i)the non-competitive supplier is not willing to increasel when it is large enough; (ii) the non-competitivesupplier will decrease r only when it is large enough.These insights are highly consistent with Proposition2. Furthermore, we show that the OEM still benefitsfrom diversifying its component orders to both sup-pliers, which hurts the competitive supplier.We also consider a different scenario in which the

non-competitive supplier has sufficient capacity. Weshow that it will produce at its full capacity and sellsthe excess components to an outside spot market at aconstant market price W. Then, the non-competitivesupplier’s profit function becomes

Pns ¼ Wnseqns þWeðs� qnsÞ:We further investigate the non-competitive sup-plier’s incentives to cooperate with the OEM, anduse “� ” on the top of the equilibrium and out-comes in this scenario.

PROPOSITION 9. If W\ 10aþ 8m20þ 9x , then

~WDtcs [ ~WDt

ns [W

and ~qDtns [ 0; otherwise, ~qDt

ns ¼ 0.

This proposition shows that, when the market priceof the spot market is sufficiently high, the non-compe-titive supplier is more likely to sell its components tothe spot market than to cooperate with the OEM,because the equilibrium wholesale price is lower thanW. We also note that, this threshold is a decreasingfunction of the technology level x. That is, a firm of alower technology level is more willing to cooperatewith the OEM. Furthermore, we re-examine the non-competitive supplier’s profit function with respect toits technology level,

~EPDtns ¼ sWlþ xð10aþ 8m� 20W � 9WxÞ2

2bð40þ 27xÞ2 :

It can be shown that the second part is unimodal inx. However, the first part is a linear increasing func-tion of l, representing the potential profit that thenon-competitive supplier can obtain from the spotmarket. Therefore, it has a strong incentive toimprove its technology level, especially the averageyield rate l, when it has such an outside option.This incentive is even stronger when the non-com-petitive supplier has more excess capacity.

6.3. Alternative Demand ModelsIn this subsection, we examine whether our findingsare robust under two alternative demand models.

The first one is an extension to the original modelwith a variable price premium of the OEM, m = bQand b > 0, that is, the OEM’s price premium is a linearincreasing function of the total output quantity. In thiscase, the OEM has a greater advantage toward itscompetitor as the total market becomes larger due tothe OEM’s branding effect. The OEM’s inversedemand function can also be rewritten asPo = a � bQ + m = a � (b � b)m, that is, the OEM’sproduct has a smaller price elasticity than its competi-tor’s. It is worth noting that b cannot be too large

(b\ffiffiffiffi

19p � 2

3 b � 0:786b). Otherwise the competitive

supplier will charge a sufficiently high wholesaleprice leading to a boundary solution.The second one is the Cournot competition model

with substitutable products Pi = a � qi � cqj, i,j 2 {cs, o}, where c (0 < c < 1) measures the degree ofproduct substitutability. The larger the c is, the lessdistinction between the end products of the competi-tive supplier and the OEM. Correspondingly, thedownstream competition will become more intense.The price premium m is thus absent in this model.To verify the robustness of our findings, we con-

duct the same analysis in the previous sections forthese alternative demand models. We first derive theequilibrium of the three scenarios: Base, Dual Sour-cing and Termination, under the two alternative mod-els, and compare the profits among the threescenarios. We find that, under either alternativemodel, the OEM still benefits from dual sourcing, butthe competitive supplier suffers from this strategy.We also find that, similar to that in section 4, we canrearrange the non-competitive supplier’s profit func-tion with the notation x = (l/r)2, and the non-compe-titive supplier’s expected profit is unimodal in x. Itaffirms that our findings and explanations are robustunder these alternative demand models.We next consider the strategic interaction between

the competitive supplier and the OEM. Similarly, westudy how the competitive supplier responds to theOEM’s order-shifting behavior, and whether termina-tion is a credible threat to the competitive supplier.Unfortunately, it is too complicated to derive analyti-cal results. We thus conduct extensive numericalstudies to see how b (in the first model), c (in the sec-ond model) and x affect the outcomes. Parametersthat we used are provided in Table 2. We have variedthousands of different feasible combinations and findthat the results are robust to the changes of

Table 2 Summary of Parameters

Market potential a = 2, b = 1

Brand effect b 2 [0, 0.75], step length = 0.025Substitutability c 2 [0, 1], step length = 0.025Technology level x 2 [0, 5], step length = 0.025


parameters. We illustrate one of the typical curves inFigure 6.The bottom-right corner in Figure 6a and top-right

corner in Figure 6b represents the cases when the com-petitive supplier decides to terminate his component-selling business, but the OEM also performs betterthan the base scenario (i.e., Competitive supplier’s ter-mination is not threatening). The top-left corner in Fig-ure 6a and bottom-left corner in Figure 6b representsthe cases when both the competitive supplier and theOEM are happy with continuing their component-sell-ing business. The area in the middle represents thecases where termination could make the OEM performworse than the base scenario, but the competitive sup-plier tends to continue the component-selling business(i.e. Competitive supplier has no incentive of termina-tion). xC and xO are two curves to divide the threeareas. For either model, we observe xC > xO, so theredoes not exist an area where the competitive supplier’stermination threatens the OEM effectively to stop theorder-shifting behavior as well as benefits himself.Regarding the sensitivity analysis, we have the fol-

lowing observations. For the first alternative model,we observe that the large the b is, the less likely thatthe competitive supplier will terminate his compo-nent-selling business. For the second alternativemodel, we observe that the smaller the c is, the lesslikely that the competitive supplier will terminate.These observations consistently imply that the com-petitive supplier will be more tolerant toward thecompetition in the up-stream component market if hisend product in the downstream market is highly dif-ferentiated from the OEM’s.In short, under the alternative demand models, we

find that our main qualitative insights hold. The

non-competitive supplier’s profit is unimodal in x.The competitive supplier will terminate his compo-nent-selling business when the non-competitive sup-plier’s production technology level is higher than athreshold value and the competitive supplier’s termi-nation of component-selling business is not a crediblethreat to prevent the OEM engaging an additionalalternative supplier.

6.4. Suppliers’ Positive Production CostsIn this subsection, we relax the zero production costassumption by assuming a positive unit productioncost c for both suppliers. Although all equilibriumscan be obtained in closed-form, additional analysisbecomes intractable. Therefore, we resort to thenumerical analysis to derive the findings here.We first re-investigate the impact of x on the order-

ing quantity qcs and qns. When there is a positive pro-duction cost, the non-competitive supplier may not beable to gain profits, if its quality x is too low. Weobserve qns ≤ 0, in either the dual sourcing scenario orthe termination scenario. Then, the base scenario isthe OEM’s sole choice. The reason is that, when thenon-competitive supplier has a positive productioncost, if x is too low, the received payment from theOEM may not cover the production cost. In otherwords, when the OEM considers the dual sourcingscenario, she should seek a relatively reliable alterna-tive supplier, so that the dual sourcing strategy is fea-sible. It is also obvious that the threshold of xincreases in c. That is, the higher the production costis, the larger x the non-competitive supplier shouldhave for surviving.Our second observation is that, Proposition 2 may

not hold when c is sufficiently high. When c becomes

(a) (b)

Figure 6 Illustration of Strategic Interaction with Alternative Demand Models [Color figure can be viewed at wileyonlinelibrary.com]



very high, qns is no longer unimodal but increasing inx. An explanation is, when there is a large productioncost related to both suppliers, their component priceswill be raised correspondingly, because they still needa reasonable profit margin. This induces the OEM tokeep increasing the order size to the non-competitivesupplier with the hope of further lowing the averagecomponent purchasing cost. This also affects the non-competitive supplier’s profit and changes the result inProposition 2. Since the increasing order quantitymay have larger impact than the decreasing whole-sale price on its profit, the non-competitive supplier’sprofit is increasing in x.We find the remaining results are consistent with

the results derived from the base model when there isa positive production c > 0 both suppliers:

1. The blue lines in Figure 7 represent the OEM’sprofits in the three scenarios. We observe thatEPD

o is the highest one and dominates theother two, and the OEM is better off in the ter-mination scenario than in the base scenariowhen x > xO.

2. The red lines represent the competitive sup-plier’s profits in the three scenarios. Weobserve that EPB

cs is the largest one and domi-nates the other two, and he is better off in thedual sourcing scenario than in the terminationscenario when x < xC.

3. We note that xC > xO, that is, the competitivesupplier’s threat on termination is non-credible.

4. In the dual sourcing scenario, the competitivesupplier’s expected profit EPD

cs decreases whilethe OEM’s expected profit EPD

o increases in x.

7. Conclusion

The co-opetitive relationship between Apple andSamsung motivates their strategic decisions such asthe component sourcing quantity allocation and self-branded business development. In this study, wecharacterize the main properties of such a co-opetitivesupply chain and find that the dual sourcing strategyis always in the OEM’s best interest. It induces a pricewar between component suppliers, which reduces theOEM’s total procurement cost. Such a cost reductionfurther strengthens the OEM’s competition advantagein the downstream market of end products. However,the OEM’s component order-shifting seriously harmsthe competitive supplier, resulting in two possiblestrategic choices: stop its self-branded business, or ter-minate the component-selling business and become apure competitor of the OEM. We show that the latteris never a credit threat for the competitive supplier toprevent the OEM from engaging an additional alter-native supplier.Contrary to conventional wisdom, we also find

that the non-competitive supplier may not haveincentive to improve its technology and raise its yieldrate. When its quality indicator has surpassed a mostpreferred level, the upstream price competitionbecomes so intense that the price advantage of thenon-competitive supplier almost vanishes. Thus, theOEM tends to shift back most component orders tothe competitive supplier who has 100% yield rate.This indicates that that the OEM may fail to find anon-competitive supplier that has sufficiently goodquality of components. Interestingly, we find that thenon-competitive supplier’s wholesale prices aredecreasing and production quantities are unimodalin its quality level, which helps explain the non-com-petitive supplier’s incentive to manage its yielduncertainty problem. We also consider the impacts ofthe two suppliers’ capacity constraints respectivelyand find that the upstream component price war willbe alleviated and the downstream market supply willbe reduced when the competitive supplier has lim-ited capacity.We note a few limitations of this study and provide

promising avenues for future research. First, weadopt a Stackelberg game framework to facilitate ouranalysis. For the future research, it will be interestingto study the wholesale price negotiation problemsamong the three players, rather than the take-it-or-leave-it pricing scheme that we assumed here. Sec-ond, in this study, we consider the scenario that OEMdoes not have the capability to influence the non-com-petitive supplier’s production technology. It is possi-ble that the OEM can co-invest in the non-competitivesupplier’s quality improvement so as to givemore pressure to the competitive supplier. The0.88 5 10 15 20 25

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Prof

it

Figure 7 Illustration of Impact of x on the Competitive Supplier’s andthe OEM’s Profits with a Positive Cost [Color figure can beviewed at wileyonlinelibrary.com]



corresponding cost-sharing mechanisms and contracttypes may influence the supply chain parties’ strate-gic decisions.We find that termination is always better for the

non-competitive supplier because in this scenario itbecomes the sole supplier of the OEM and theupstream price war no longer exists. Then, the non-competitive supplier may have more incentives inimproving its technology level x to be higher than xCand induce the competitive supplier to terminate itscomponent-selling business. However, in practice, theimprovement of x can be prohibitively costly. If thenon-competitive supplier can afford this technologyimprovement cost, the global equilibrium must beTermination. Otherwise, the global equilibrium willbe Dual-sourcing. Noting that the technologyimprovement cost issue can be overly complicatedthat distract readers from the main objectives of ourresearch, we chose to keep x exogenously given andderived the global equilibriums conditional on thelevel of x. Notwithstanding these limitations, thisstudy presents a first step in understanding the OEMand competitive supplier’s attitude toward the dualsourcing and dual channel strategies. We believe thegrowing popularity of frenemy relationship in thehigh-tech industry presents an exciting area ofresearch.

Acknowledgments

The authors are grateful to the Department Editor, theSenior Editor and the two anonymous reviewers for theirhelpful comments and suggestions. Authors contributeequally to this study and are ordered alphabetically accord-ing to their preferred first name. Jie Zhang is the corre-sponding author. The authors’ work was supported in partby NSFC Excellent Young Scientists Fund (No. 71822202),NSFC (No. 71871067, 71571194, 71301032), Chang JiangScholars Program (Niu Baozhuang 2017), GDUPS (NiuBaozhuang 2017).

Notes

1TSMC is a non-competitive supplier of Apple and itdeclares that it will be “staying away from designing,manufacturing or marketing semiconductor productsunder its own brand name so it can avoid competingagainst its customers” (Forbes 2017).2A positive assembly cost is equivalent to the reduction ofm, which does not affect our main results.3Note that Var(e) = E(e2) � (E(e))2. e ranges from 0 to 1,thus E(e2) < E(e). We have Var(e) < E(e) � (E(e))2. BecauseE(e) also ranges from 0 to 1, E(e) � (E(e))2 reaches its max-imum 1/4 when E(e) = 1/2. Hence, Var(e) < 1/4. Thisindicates that r is smaller than 1/2.4If the OEM pays for what she orders instead of what shereceived, as Tang and Kouvelis (2011), we find that all ourresults and findings remain unchanged, except that the

non-competitive supplier’s wholesale price WT0ns ¼ lWT

ns

which becomes lower, because it does not bear the yieldcost in that case. We realize that these two paymentschemes (i.e., pay for what is ordered, and, pay for whatis received) are mostly equivalent. Similar results andexplanations can be found in Tang and Kouvelis (2011).

ReferencesAdner, R., J. Chen, F. Zhu. 2015. Frenemies in platform markets:

The case of Apple’ iPad vs. Amazon’ Kindle. Harvard Work-ing Paper Series No. 15–087.

Amaral, J., C. Billington, A. Tsay. 2006. Safeguarding the promiseof production outsourcing. Interfaces 36: 220–233.

Arru~nada, B., X. V�azquez. 2006. When your contract manufacturerbecomes your competitor. Harv. Bus. Rev. 84(9): 135–144.

Autrey, R., F. Bova, D. Soberman. 2014. When gray is good: Graymarkets and market–creating investments. Prod. Oper. Manag.24(4): 547–559.

Bohn, R., C. Terwiesch. 1999. The economics of yield–driven pro-cesses. J. Oper. Manag. 18: 41–59.

Bora, K. 2014. Apple iPhone 6 rumors: TSMC to use 8–inch pro-cessing method to produce fingerprint sensors for next–GeniPhone. Available at http://www.ibtimes.com/apple-iphone-6-rumors-tsmc-use-8-inch-processing-method-produce-finger-print-sensors-1554550. February 11 (accessed date August 10,2017).

Brandenburger, A., B. Nalebuff. 1996. Co-opetition: A RevolutionMindset that Combines Competition and Cooperation. BantamDoubleday Dell Press. New York.

Cachon, G., C. Terwiesch, Y. Xu. 2008. On the effects of consumersearch and firm entry in a multiproduct competitive market.Market. Sci. 27: 461–473.

Cai, G. 2010. Channel selection and coordination in dual-channelsupply chains. J. Retail. 86: 22–36.

Chen, J., Z. Guo. 2014. Strategic sourcing in the presence of uncer-tain supply and retail competition. Prod. Oper. Manag. 23:1748–1760.

Chen, Y., S. Shum, W. Xiao. 2012. Should an OEM retain compo-nent procurement when the CM produces competing prod-ucts? Prod. Oper. Manag. 21: 907–922.

Chowdhry, A. 2014. Apple and Samsung drop patent disputesagainst each other outside of the U.S. Available at https://www.forbes.com/sites/amitchowdhry/2014/08/06/apple-and-samsung-drop-patent-disputes-against-each-other-outside-of-the-u-s/#656d7c8418ca. August 6 (accessed date August 10,2017).

Deo S., C. Corbett. 2009. Cournot competition under yield uncer-tainty: The case of the US influenza vaccine market. Manuf.Serv. Oper. Manag. 11: 563–576.

Dumrongsiri, A., M. Fan, A. Jain, K. Moinzadeh. 2008. A supplychain model with direct and retail channels. Eur. J. Oper. Res.187(3): 691–718.

Epstein, Z. 2014. Apple and Samsung: Doomed to be frenemiesforever? Available at http://bgr.com/2014/07/01/apple-a9-processor-production-samsung/. July 1 (accessed date August10, 2017).

Forbes, 2017. Taiwan’s richest: Morris Chang, pioneering chip-maker and list newcomer. Available at https://www.forbes.com/sites/luisakroll/2017/06/28/taiwans-richest-morris-chang-pioneering-chipmaker-and-list-newcomer/#4d2a1e056d40. June28 (accessed date May 24, 2018).

Humphries, M. 2012. Samsung terminates LCD contract withApple. Available at http://www.geek.com/apple/samsung-


terminates-lcd-contract-with-apple-1523689/. October 22 (ac-cessed date August 10, 2017).

IDC. 2018. Smartphone shipments down 6.3% in Q4 2017, Appleovertakes Samsung for top spot. Available at https://www.techspot.com/news/73082-idc-apple-overtakes-samsung-smartphone-market-global-shipments.html. February 2 (ac-cessed date March 2, 2018).

Kaiser, T. 2013. Samsung takes hit as Apple partners with TSMCfor iPhone, iPad chips. Available at http://www.dailytech.com/Samsung+Takes+Hit+as+Apple+Partners+with+TSMC+for+iPhone+iPad+Chips/article31871.htm. July 1 (accesseddate August 10, 2017).

Karp, L., J. Perloff. 2012. Vertical contracts and mandatory univer-sal distribution. BE J. Econ. Anal. Policy 13: 595–626.

Kim, M. 2012. Analysis: Friend and foe; Samsung, Apple won’twant to damage parts deal. http://www.reuters.com/article/us-samsung-apple-supply-idUSBRE87Q06N20120827. August27 (accessed date August 10, 2017).

Kumar, N., R. Ruan. 2006. On manufacturers complementing thetraditional retail channel with a direct online channel. Quant.Market. Econ. 4: 289–323.

Lee, J., I. King. 2015. Samsung said to win Apple A9 chip ordersfor next IPhone. Available at https://www.bloomberg.com/amp/news/articles/2015-04-03/samsung-said-to-win-apple-a9-chip-orders-for-next-iphone. April 2 (accessed date August10, 2017).

Li, T., S. Sethi, J. Zhang. 2013. Supply diversification with respon-sive pricing. Prod. Oper. Manag. 22: 447–458.

Li, Z., K. J. Ryan, L. Shao, D. Sun. 2015. Supply contract designfor competing heterogeneous suppliers under asymmetricinformation. Prod. Oper. Manag. 24: 791–807.

Reisinger, D. 2015. Apple, Samsung to remain bedfellows fornext iPhone–report. Available at https://www.cnet.com/news/apple-samsung-still-bedfellows-as-they-eye-next-iphone-report-says/. January 26 (accessed date August10, 2017).

Singh, N., X. Vives. 1984. Price and quantity competition in a dif-ferentiated duopoly. RAND J. Econ. 15(4): 546–554.

Snow, D. C., S. C. Wheelwright, A. B. Wagonfeld. 2006. Genen-tech–Capacity planning. Harv. Bus. School Case 606–052.

Song, J. 2012. Samsung: Playing tough with Apple. Available athttp://blogs.ft.com/beyond-brics/2012/11/13/samsung-play-ing-tough-with-apple/. November 13 (accessed date August10, 2017).

Spiegel, Y. 1993. Horizontal subcontracting. RAND J. Econ. 24:570–590.

Swartz, J. 2009. Cisco goes up against IBM, HP in computer ser-ver market. Available at http://abcnews.go.com/Business/story?id=7098334. May 27 (accessed date August 10, 2017).

Tang, S., P. Kouvelis. 2011. Supplier diversification strategies inthe presence of yield uncertainty and buyer competition.Manuf. Serv. Oper. Manag. 13: 439–451.

Tang, S., H. Gurnani, D. Gupta. 2014. Managing disruptions indecentralized supply chains with endogenous supply processreliability. Prod. Oper. Manag. 23: 1198–1211.

Tibken, S. 2012. Samsung kicking Apple to curb for LCD panels,report says. Available at https://www.cnet.com/news/samsung-kicking-apple-to-curb-for-lcd-panels-report-says/. May23 (accessed date August 10, 2017).

Tomlin, B. 2006. On the value of mitigation and contingencystrategies for managing supply chain disruption risks. Man-agement Sci. 52: 639–657.

Tsay, A., N. Agrawal. 2004. Modeling conflict and coordination inmulti-channel distribution systems: A review. D. Simchi-Levi,D. Wu, and M. Shen, eds. Handbook of Quantitative SupplyChain Analysis: Supply Chain Analysis in the Ebusiness Era.Springer, Boston, MA, 557–606.

Wang, Y., B. Niu, P. Guo. 2013. On the advantage of quantityleadership when outsourcing production to a competitivecontract manufacturer. Prod. Oper. Manag. 22: 104–119.

Worstall, T. 2013. No surprise in Global Foundries being used tomake Apple’s processors by Samsung. Available at https://www.forbes.com/sites/timworstall/2013/11/12/no-surprise-in-globalfoundries-being-used-to-make-apples-processors-by-samsung/. November 12 (accessed date August 10, 2017).

Wu, X., F. Zhang. 2014. Home or overseas? An analysis of sourcingstrategies under competition. Management Sci. 60: 1223–1240.

Yano, C., H. Lee. 1995. Lot sizing with random yields: A review.Oper. Res. 43: 311–334.

Supporting InformationAdditional supporting information may be found onlinein the Supporting Information section at the end of thearticle.

Appendix S1: Proofs.


strategic analysis of dual sourcing and dual … poms dual...strategic analysis of dual sourcing and...

Documents