Optimal inventory policy through dual sourcing


Profit maximization in the retail and manufacturing industry is currently focused on offshore production to utilize the resulting low production costs. However, in the face of uncertain customer demand, it is difficult to determine an optimal order quantity which maximizes profit. We consider a risk-averse firm that utilizes dual-sourcing for perishable or seasonal goods with uncertain customer demand. Using real options theories, we provide two models aimed at determining optimal order quantities to maximize the firm’s expected profit. Furthermore, we can consider the demand to be an observable process correlated to a traded asset, which can be hedged to reduce profit uncertainty. A single offshore single local order period model provides a pseudo-analytical solution which can be easily solved to determine optimal offshore and local order quantities based on the manufacturers’ lead times, and a more realistic single offshore multiple local order period model which uses numerical methods to determine optimal order quantities. Finally, a method for matching distributions of expected demands based on managerial estimates can be applied to the two models, providing managers a tool for practical application.

This is a preview of subscription content, log in to check access.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10
Fig. 11
Fig. 12
Fig. 13
Fig. 14
Fig. 15
Fig. 16
Fig. 17


  1. 1.

    We note that the process, \(X_t\), can be modelled in the actual or risk-neutral probability space.

  2. 2.

    For a risk-neutral valuation, the drift can be adjusted accordingly.


  1. Agrawal V, Seshadri S (2000) Impact of uncertainty and risk aversion on price and order quantity in the newsvendor problem. Manuf Serv Oper Manag 4:410–423

    Article  Google Scholar 

  2. Arrow K, Harris T, Marshak J (1951) Optimal inventory policy. Econometrica 19:250–272

    Article  Google Scholar 

  3. Barankin E (1961) A delivery-lag inventory model with an emergency provision (the single-period case). Naval Res Logist Q 8:285–311

    Article  Google Scholar 

  4. Borison A (2005) Real options analysis: where are the emperor’s clothes? J Appl Corp Finance 17(2):17–31

    Article  Google Scholar 

  5. Chen F, Federgruen A, Zheng Y-S (2001) Coordination mechanisms for a distribution system with one supplier and multiple retailers. Manag Sci 47–5:693–708

    Article  Google Scholar 

  6. Chen L, Li S, Wang L (2014) Capacity planning with financial and operational hedging in low-cost countries. Prod Oper Manag 23–9:1495–1510

    Article  Google Scholar 

  7. Cheng Y, Xu M, Zhang Z (2009) A risk-averse newsvendor model under the CVaR criterion. Oper Res 57–4:1040–1044

    Article  Google Scholar 

  8. Chou Y, Cheng C-T, Yang F-C, Liang Y-Y (2007) Evaluating alternative capacity strategies in semiconductor manufacturing under uncertain demand and price scenarios. Int J Prod Econ 105:591–606

    Article  Google Scholar 

  9. Dixit AK, Pindyck RS (1994) Investment under uncertainty. Princeton University Press, Princeton

    Google Scholar 

  10. Fu H, Dan B, Sun X (2014) Joint optimal pricing and ordering decisions for seasonal products with weather-sensitive demand. Discrete Dyn Nat Soc 2014:1–8

    Google Scholar 

  11. Fukuda Y (1964) Optimal policies for the inventory problem with negotiable leadtime. Manag Sci 10:690–708

    Article  Google Scholar 

  12. Gaur V, Seshadri S (2005) Hedging inventory risk through market instruments. Manuf Serv Oper Manag 7–2:103–120

    Article  Google Scholar 

  13. Gotoh J, Takano Y (2007) Newsvendor solutions via conditional value-at-risk minimization. Eur J Oper Res 179:80–96

    Article  Google Scholar 

  14. Ho W, Zheng T, Yildiz H, Talluric S (2015) Supply chain risk management: a literature review. Int J Prod Res 53(16):5031–5069

    Article  Google Scholar 

  15. Jaimungal S, Lawryshyn Y (2015) Incorporating managerial information into real option valuation. In: Aid R, Ludkovski M, Sircar R (eds) Fields volume on commodities and environmental finance. Springer, Berlin, pp 213–238

    Google Scholar 

  16. Jia J, Hub Q (2011) Dynamic ordering and pricing for a perishable goods supply chain. Comput Ind Eng 60:302–309

    Article  Google Scholar 

  17. Keren B (2009) The single-period inventory problem: extension to random yield from the perspective of the supply chain. Omega 37:801–810

    Article  Google Scholar 

  18. Kouvelis P, Li R, Ding Q (2014) Managing storable commodity risks: the role of inventory and financial hedge. Manuf Serv Oper Manag 15–3:507–521

    Google Scholar 

  19. Li S, Wang L (2010) Outsourcing and capacity planning in an uncertain global environment. Eur J Oper Res 207:131–141

    Article  Google Scholar 

  20. Minner S (2003) Multiple-supplier inventory models in supply chain management: a review. Int J Prod Econ 81–82:265–279

    Article  Google Scholar 

  21. Okyay H, Karaesmen F, Ozekici S (2014) Hedging demand and supply risks in the newsvendor model. OR Spectr 37:475–501

    Article  Google Scholar 

  22. Rabbani M, Zia N, Rafiei H (2016) Joint optimal dynamic pricing and replenishment policies for items with simultaneous quality and physical quantity deterioration. Appl Math Comput 287–288:149–160

    Google Scholar 

  23. Shu L, Wu F, Ni J, Chu L (2015) On the risk-averse procurement strategy under unreliable supply. Comput Ind Eng 84:113–121

    Article  Google Scholar 

  24. Tang C (2006) Perspectives in supply chain risk management. Int J Prod Econ 103:451–488

    Article  Google Scholar 

  25. Trigeorgis L (1996) Real options: managerial flexibility and strategy in resource allocation. MIT Press, Cambridge

    Google Scholar 

  26. Uryasev S, Rockafellar R (2000) Optimization of conditional value-at-risk. J Risk 2–3:21

    Google Scholar 

  27. Whittemore A, Saunders S (1977) Optimal inventory under stochastic demand with two supply options. SIAM J Appl Math 32:293–305

    Article  Google Scholar 

  28. Xanthopoulos A, Vlachos D, Iakovou E (2011) Optimal newsvendor policies for dual-sourcing supply chains: a disruption risk management framework. Comput Oper Res 39:350–357

    Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Matthew Davison.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.



A Derivation of \(U^*\)

To derive \(U^*\), we rewrite \({\mathbb {E}}_{\tau }[G(X_T;M,U)]\) from Eq. 2 as

$$\begin{aligned} {\mathbb {E}}_{\tau }[G(X_T;M,U)]= & {} \int ^{\infty }_{0} [\underbrace{\min (x,M+U)P}_\text {A1} + \underbrace{P_{Salv}(M + U - x)^{+}}_\text {A2} \nonumber \\&- \underbrace{P_{Strat}(x - M - U)^{+}}_\text {A3}]f_{X_T|X_\tau }(x;y=X_\tau )dx - \underbrace{UC_{U}}_\text {A4}\nonumber \\ \end{aligned}$$

and maximize the expected profit by taking the derivative with respect to U and equating it to zero. Separating the equation into four parts \(A_{1}\), \(A_{2}\), \(A_{3}\) and \(A_{4}\) allows for easy expression management. Beginning with \(A_1\), the discontinuous nature of the equation makes it necessary to separate it into components where integration is possible as shown with,

$$\begin{aligned} \begin{aligned} A_{1} =&\int ^{\infty }_{0} (1_{x < M+U} \cdot x+1_{x > M+U} \cdot (M+U)) \cdot P \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx. \end{aligned} \end{aligned}$$

Next, taking the derivative of \(A_1\) with respect to U and integrating the function yields,

$$\begin{aligned} \begin{aligned} \dfrac{dA_{1}}{dU} =&\int ^{\infty }_{0} 1_{x > M+U} \cdot P \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx\\ =&\int ^{\infty }_{M+U} P \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx\\ =&P \cdot (1-F_{X_T|X_\tau }(M+U;y=X_\tau )). \end{aligned} \end{aligned}$$

The same process is applied for parts \(A_2 - A_4\);

$$\begin{aligned} A_{2}= & {} \int ^{\infty }_{0} P_{Salv} \cdot 1_{x < M+U} \cdot (M+U-x) \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx, \end{aligned}$$
$$\begin{aligned} \dfrac{dA_{2}}{dU}= & {} \int ^{\infty }_{0} P_{Salv} \cdot 1_{x < M+U} \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx\nonumber \\= & {} \int _{0}^{M+U} P_{Salv} \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx\nonumber \\= & {} P_{Salv} \cdot F_{X_T|X_\tau }(M+U;y=X_\tau ), \end{aligned}$$
$$\begin{aligned} A_{3}= & {} \int ^{\infty }_{0} P_{Strat} \cdot 1_{x > M+U} \cdot (x-M-U)) \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx, \end{aligned}$$
$$\begin{aligned} \dfrac{dA_{3}}{dU}= & {} -\int ^{\infty }_{0} P_{Strat} \cdot 1_{x > M+U} \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx\nonumber \\= & {} -\int ^{\infty }_{M+U} P_{Strat} \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx\nonumber \\= & {} -P_{Strat} \cdot (1-F_{X_T|X_\tau }(M+U;y=X_\tau )), \end{aligned}$$
$$\begin{aligned} A_{4}= & {} UC_{U}, \end{aligned}$$
$$\begin{aligned} \dfrac{dA_{4}}{dU}= & {} C_{U}. \end{aligned}$$

Substituting Eqs. 20, 22, 24, 26 into Eq. 18 and isolating for U, allows us to determine the optimal onshore order quantity \(U^*\), which maximizes the overall profit at time \(t=\tau \),

$$\begin{aligned} U^* = \underbrace{F^{-1}_{X_T|X_{\tau }}\bigg (\dfrac{P + P_{Strat} - C_{U}}{P+P_{Strat}-P_{Salv}};y\bigg )}_{\gamma (y)} - M \equiv \gamma (y)-M. \end{aligned}$$

B Derivation of the Profit

Following a procedure similar to the derivation of \(U^*\), we begin with Eq. 5

$$\begin{aligned} \begin{aligned}&{\mathbb {E}}[Profit(M,U^*)] \\&\quad = \int ^{\infty }_{0}[\underbrace{\min (X,M+(\gamma (y) - M)^+)P}_\text {B1} + \underbrace{P_{Salv} \cdot (M + (\gamma (y) - M)^+ - X)^{+}}_\text {B2}\\&\qquad - \underbrace{P_{Strat} \cdot (X - M - (\gamma (y) - M)^+)^{+}}_\text {B3}]f_{X_T|X_\tau }(x;y=X_\tau )dx \\&\qquad \underbrace{-MC_{M} - (\gamma (y) - M)^+C_{U}}_\text {B4}, \end{aligned} \end{aligned}$$

and separate it into four parts \(B_{1}\) to \(B_{4}\) and solve each individually,

$$\begin{aligned} \begin{aligned} B_{1} =&\int ^{\infty }_{0} (1_{x< M+U} \cdot x+1_{x>M+U} \cdot (M+U)) \cdot P \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx,\\ B_{2} =&\int ^{\infty }_{0} (1_{x< M+U} \cdot (M+U-x)) \cdot P_{Salv} \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx,\\ B_{3} =&\int ^{\infty }_{0} (1_{x > M+U} \cdot (x-(M+U))) \cdot P_{Strat} \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx,\\ B_{4} =&{\left\{ \begin{array}{ll} -MC_{M} - (\gamma (y) - M)C_{U} &{} \text { if } \gamma (y) \ge M\\ -MC_{M} &{} \text { if } \gamma (y)<M. \end{array}\right. } \end{aligned} \end{aligned}$$

Since the proposed functions are discontinuous, we must determine appropriate integration boundaries to the functions in Eq. 29, as can be seen for \(B_1\) where

$$\begin{aligned} B_{1} =&\int ^{\infty }_{0} (1_{x < M+U} \cdot x+1_{x>M+U} \cdot (M+U)) \cdot P \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx. \end{aligned}$$

Substituting \(U^*\) for U transforms the first part of the discontinuous function \((1_{x < M+U} \cdot x)\) into

$$\begin{aligned}&1_{x<M+(\gamma (y)-M)^+} \rightarrow {\left\{ \begin{array}{ll} 1_{x<\gamma (y)} &{} \text { if } \gamma (y) \ge M\\ 1_{x<M} &{} \text { if } \gamma (y)<M\\ \end{array}\right. } = 1_{\gamma (y)>M} \cdot 1_{x<\gamma (y)} + 1_{\gamma (y)<M} \cdot 1_{x<M} \end{aligned}$$

which exists under the conditions that demand is between M and \(\gamma (y)\) when \(\gamma (y)\) is greater than M, or when demand is less than M when \(\gamma (y)\) is less than M, as shown in the following plot line


The second part of the discontinuous function \((1_{x>M+(\gamma (y)-M)^+})\) is handled in a similar manner, initially by substituting \(U^*\) for U, and then determining the integration boundaries:

$$\begin{aligned} 1_{x>M+(\gamma (y)-M)^+} = 1_{\gamma (y)>M} \cdot 1_{x>\gamma (y)} + 1_{\gamma (y)<M} \cdot 1_{x>M} \end{aligned}$$

As a result, the function \(B_1\) simplifies to:

$$\begin{aligned} B_{1} =&P \left\{ \begin{array}{ll} \int ^{\gamma (y)}_{0} x \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx + \int ^{\infty }_{\gamma (y)} \gamma (y)&{} \text { if } \gamma (y) \ge M\\ \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx &{}\\ \int ^{M}_{0} x \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx + \int ^{\infty }_{M} M &{} \text { if } \gamma (y)<M. \\ \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx &{}\\ \end{array} \right. \end{aligned}$$

Applying a similar process, functions \(B_2\) and \(B_3\) simplify to

$$\begin{aligned} B_{2} =&P_{Salv} \left\{ \begin{array}{ll} \int ^{\gamma (y)}_{0} (\gamma (y)-x) \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx &{} \text { if } \gamma (y) \ge M\\ \int ^{M}_{0} (M-x) \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx &{} \text { if } \gamma (y)<M \end{array}\right. \end{aligned}$$
$$\begin{aligned} B_{3} =&P_{Strat} \left\{ \begin{array}{ll} \int _{\gamma (y)}^{\infty } (x-\gamma (y)) \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx &{} \text { if } \gamma (y) \ge M\\ \int _{M}^{\infty } (x-M) \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx &{} \text { if } \gamma (y)<M. \end{array}\right. \end{aligned}$$

By defining \(G(y;a,b) \equiv \int ^{b}_{a}xf_{X_T|X_\tau }(x;y=X_\tau )dx\) we are able simplify the function \(\int ^{\gamma (y)}_{0} (\gamma (y)-x) \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx\) such that

$$\begin{aligned} \begin{aligned} \int ^{\gamma (y)}_{0} (\gamma (y)-x) \cdot f_{X_T|X_\tau }(x;y=&X_\tau )dx \equiv \gamma (y) \cdot F_{X_T|X_\tau }(\gamma (y);y) \\&- G(y;0,\gamma (y)). \end{aligned} \end{aligned}$$

Applying this process to Eqs. 30, 31, and 32, we are able to reduce \(B_1\) to \(B_3\) to

$$\begin{aligned} B_{1}= & {} P \left\{ \begin{array}{ll} G(y;0,\gamma (y)) + \gamma (y) \cdot (1-F_{X_T|X_\tau }(\gamma (y);y)) &{} \text { if } \gamma (y) \ge M\\ G(y;0,M)+M \cdot (1-F_{X_T|X_\tau }(M;y)) &{} \text { if } \gamma (y)<M \end{array}\right. \end{aligned}$$
$$\begin{aligned} B_{2}= & {} P_{Salv} \left\{ \begin{array}{ll} \int ^{\gamma (y)}_{0} (\gamma (y)-x) \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx &{} \text { if } \gamma (y) \ge M\\ \int ^{M}_{0} (M-x) \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx &{} \text { if } \gamma (y)<M \end{array}\right. \nonumber \\= & {} P_{Salv} \left\{ \begin{array}{ll} \gamma (y) \cdot F_{X_T|X_\tau }(\gamma (y);y)-G(y;0,\gamma (y)) &{} \text { if } \gamma (y) \ge M\\ M \cdot F_{X_T|X_\tau }(M;y)-G(y;0,M) &{} \text { if } \gamma (y)<M \end{array}\right. \end{aligned}$$
$$\begin{aligned} B_{3}= & {} P_{Strat} \left\{ \begin{array}{ll} \int _{\gamma (y)}^{\infty } (x-\gamma (y)) \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx &{} \text { if } \gamma (y) \ge M\\ \int _{M}^{\infty } (x-M) \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx &{} \text { if } \gamma (y)<M\\ \end{array}\right. \nonumber \\= & {} -P_{Strat} \left\{ \begin{array}{ll} \gamma (y)[1-F_{X_T|X_\tau }(\gamma (y);y)]-\int ^{\infty }_{\gamma (y)}(x) &{} \text { if }\gamma (y) \ge M\\ \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx&{}\\ M[1-F_{X_T|X_\tau }(M);y)]-\int ^{\infty }_{M}(x) &{} \text { if } \gamma (y) \ge M\\ \cdot f_{X_T|X_\tau }(x;y=X_\tau )dx&{}\\ \end{array}\right. \nonumber \\= & {} -P_{Strat} \left\{ \begin{array}{ll} \gamma (y)-\gamma (y) \cdot F_{X_T|X_\tau }(\gamma (y);y)-G(y;\gamma (y),\infty ) &{} \text { if } \gamma (y) \ge M\\ M-M \cdot F_{X_T|X_\tau }(M;y)-G(y;M,\infty ) &{} \text { if } \gamma (y)<M. \end{array}\right. \nonumber \\ \end{aligned}$$

Rearranging and cancelling terms in the above expression reduces the profit function to

$$\begin{aligned} Profit = \left\{ \begin{array}{ll} \gamma (y) \cdot (P+P_{Strat}-C_{U}) + \gamma (y) \cdot F(\gamma (y);y) &{}\text {if } \gamma (y) \ge M\\ \cdot (P_{Salv}-P-P_{Strat}) -P_{Strat} \cdot G(y;\gamma (y),\infty ) &{}\\ + (P-P_{Salv})\cdot G(y;0,\gamma (y))+ M \cdot (C_{U}-C_{M})&{}\\ &{}\\ M \cdot (P+P_{Strat}-C_{M})+M \cdot F_{X_T|X_\tau }(M;y) &{} \text { if } \gamma (y)<M.\\ \cdot (P_{Salv}-P-P_{Strat})+G(y;0,M) \cdot (P-P_{Salv})&{}\\ -P_{Strat} \cdot G(y;M,\infty )&{} \end{array}\right. \end{aligned}$$

After some algebraic manipulation, the expected profit at \(t=0\) can be determined as

$$\begin{aligned} \begin{aligned} {\mathbb {E}}[Profit(y)]&= \int ^{\infty }_{y=\gamma ^{-1}(M)}\big [ \gamma (y) \cdot (P+P_{Strat}-C_{U}) + \gamma (y) \cdot F_{X_T|X_\tau }(\gamma (y);y) \\&\quad \cdot (P_{Salv}-P-P_{Strat}) -P_{Strat} \cdot G(y;\gamma (y),\infty ) + (P-P_{Salv}) \\&\quad \cdot G(y;0,\gamma (y)) + M \cdot (C_{U}-C_{M}) \big ]f_X(y)dy \\&\quad +\int ^{\gamma ^{-1}(M)}_{y=0}\big [ M \cdot (P+P_{Strat}-C_{M})+M \cdot F_{X_T|X_\tau }(M;y)\\&\cdot (P_{Salv}-P-P_{Strat}) +G(y;0,M) \cdot (P-P_{Salv})\\&\quad -P_{Strat} \cdot G(y;M,\infty ) \big ]f_X(y)dy, \end{aligned} \end{aligned}$$

where \(f_X(\cdot )\) is the density of X.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Davison, M., Lawryshyn, Y. & Miklyukh, V. Optimal inventory policy through dual sourcing. Comput Manag Sci (2020). https://doi.org/10.1007/s10287-020-00371-8

Download citation