Vol. 10, No. 3, May 2010, pp.

105–112
issn 1532-0545
10
1003
0105 informs ®

doi 10.1287/ited.1100.0045
I N F O R M S © 2010 INFORMS

Transactions on Education

The Mexico-China Sourcing Game:

Teaching Global Dual Sourcing
Gad Allon, Jan A. Van Mieghem
Kellogg School of Management, Northwestern University, Evanston, Illinois 60208
{[email protected], [email protected]}

W e describe a three-hour class on global dual sourcing built around a game that demonstrates the challenges
in making operational decisions, and transfers recent academic insights to the classroom. Student teams
manage a firm with access to a responsive but expensive supply source (Mexico) and a cheap but remote source
(China). Each team must determine a sourcing strategy to satisfy random demand that is revealed throughout
the game. In each period, teams place orders to both sources and manage two assets: inventory and their bank
account. The goal is to maximize each team’s value (final bank balance). During the debriefings, we analyze the
policies used by different teams along both financial and operational metrics, present the optimal strategy, and
summarize the experiential learning points.
Key words: dual sourcing; strategic sourcing; experiential learning; inventory management; total landed cost;
simulation game
History: Received: July 24, 2009; accepted: March 1, 2010.

1. Introduction (2) to highlight the role of working capital estima-

In many retail settings, and business in general, firms tion in the concept of total landed cost;
can source goods from multiple suppliers. We inves- (3) to appreciate the added complexity of manag-
tigate a typical choice between two suppliers: One is ing a supply portfolio over single sourcing;
low cost but has long lead times whereas the other (4) to understand the futility of guessing and over-
provides quick response but at higher cost. This clas- reacting to demand.
sical problem is faced by many companies and most The game has been used with MBA and execu-
students understand its relevance. However, effective tive education students and although some knowl-
management of dual sourcing is surprisingly chal- edge of basic inventory models is useful, it is not nec-
lenging and conveying its complexity through tra- essary. In the remainder, we will describe sequentially
ditional means such as a case study or a lecture is the three-hour class we have taught in three parts:
difficult: Solving a dual-sourcing case requires ana- Part A introduces global sourcing and related theory,
lytic tools that are not readily available to students or Part B describes the game, and Part C summarizes
instructors. A lecture on dual sourcing runs the risk the debriefing phase as well as a discussion of recent
that students may not appreciate the added level of academic research. We conclude this paper by sum-
complexity in day-to-day, as well as strategic, dual marizing student reactions and possible extensions.
sourcing relative to single sourcing.
We have developed a sourcing game as an expe-
riential learning tool that addresses the above short- 2. Class Part A: Introduction to
comings. In this game, students play the role of Global Sourcing
sourcing managers who must make strategic alloca- This section describes the first part of the class
tion decisions as well as place day-to-day orders to that introduces global sourcing and related theory
two suppliers. During this process, they experience (about 30 to 45 minutes of class time).
the operational, financial and service related conse-
quences of their decisions. 2.1. The Motivating Case
The sourcing game is designed with four pedagog- As preparation for the class, the students are asked
ical objectives in mind: to read the Mexico-China minicase in Van Mieghem
(1) to develop intuition on possible sourcing strate- (2008). The case describes a $10 billion high-tech U.S.
gies and to appreciate simple, robust policies to guide manufacturer of wireless transmission components.
sourcing decisions; The company was at a crossroads regarding its global
105
 Allon and Van Mieghem: The Mexico-China Sourcing Game: Teaching Global Dual Sourcing
106 INFORMS Transactions on Education 10(3), pp. 105–112, © 2010 INFORMS

network; the case focuses on its two assembly plants, 2.3. The Concept of Total Landed Cost (TLC)
one in China and the other in Mexico. Although the To address the motivating case, the natural first step is
Chinese facility enjoyed lower costs, ocean transporta- to compare the two extreme solutions: single sourcing
tion made its order lead times 5 to 10 times longer from Mexico versus single sourcing from China. This
than those from Mexico. With highly uncertain prod- requires accounting for the difference in cost of goods
uct demand—coefficients of variations of monthly sold (COGS), shipping costs, duties, and working-
demand for some products were as high as 1.25—sole
capital requirements.1 For that purpose we review the
sourcing was unattractive: Mexico was too expensive
concept of total landed cost (TLC), which represents
and China too unresponsive.
The firm had to decide how it could best utilize the end-to-end cost to transform inputs at the source
these two sources. At the strategic level, this amounts to outputs at destination. The TLC captures not only
to properly allocating product demand to each source. the traditional cost of goods sold but also accounts
Strategic allocation refers to the expected cumula- for supply chain costs such as transportation, cus-
tive demand allocated to each source over the plan- toms, duties, and taxes, as well as required working
ning horizon. At the tactical level, the firm had to capital carrying costs to achieve desired service lev-
choose a dynamic ordering policy that implements els and protect against supply and demand risks (see
that strategic allocation at lowest cost. In practice, Figure 1).
specifying strategic allocations and ordering policies We will refer to all but working capital cost com-
are key tasks of any sourcing strategy—be it global ponents as the sourcing cost. Computing the sourcing
or domestic—because it affects costs and supplier
cost is tedious yet straightforward. In contrast, work-
management.
ing capital is driven by pipeline and safety inventory,
2.2. Discussion on Practice which depend on lead times, volatility, and service
To connect the game to real sourcing practices, and levels.
if students have prior experience with similar set-
tings, we start with an open discussion covering two
questions: 2.4. Computing TLC for Single Sourcing
(1) What policies have you used in practice? This Working capital and inventory requirements are eas-
discussion typically reveals some of the following ily estimated for single sourcing using readily avail-
practices: able standard inventory formulae. Before moving to
(a) The allocation of products to plants is based dual sourcing and to the game, we remind the stu-
solely upon historic allocations (e.g., new products dents of the concept of safety stock and compute the
within a family are allocated to plants in the exact total landed cost for each of the locations separately.
same fashion as existing products in that family This computation reveals that single sourcing from
regardless of supply and demand characteristics). China appears to be the favorite solution.
(b) The allocation of products to plants is based Asking the students’ reaction to this analysis gen-
upon internal politics (e.g., “pet” production facilities
erates a discussion of the merits of sourcing from
get higher volume, lower complexity products to keep
Mexico and of the expected benefits of dual sourcing
utilization high and cost structure low).
(c) The allocation of products to plants is based in a quest to combine the strengths of both sources.
upon basic understanding of costs (e.g., China’s This also leads to a discussion on the risk of using a
cost base is lower—thus source full requirements single source. We usually discuss two types of sup-
offshore). ply risks: stochastic lead times and supply disruptions
(d) Product allocations follow a simple primary- (such as natural disasters strikes.) This naturally leads
secondary allocation (e.g., if primary location to the question of how to design an effective dual-
cannot fulfill demand—requirements cascade to the sourcing policy, which is the transition point to the
secondary location). game.
(e) Product-to-plant allocation decisions are rare-
ly revisited or adjusted (e.g., allocation decision is a
“life sentence”). 1
It should be noted that working capital includes the pipeline
(2) What data or metrics are used to make sourcing deci- inventory as well as the safety stock. The pipeline inventory
sions? This discussion follows the practices above and depends on the transformation process, which takes place between
highlights the typical cost, lead times, and service risk the moment of ordering and the moment of receipt of the order. If
part of the lead time is caused by inflexibility in production because
metrics. The instructor then can ask the natural ques-
of set-up times, then there is no per se investment in material dur-
tion: How can we combine these various components ing the lead-time period. Thus, in order to accurately assess the
into a single metric? This brings us to the concept of working capital requirements one needs to know the precise timing
total landed cost. of material inflow to build a sequence of cash flows over time.
Allon and Van Mieghem: The Mexico-China Sourcing Game: Teaching Global Dual Sourcing
INFORMS Transactions on Education 10(3), pp. 105–112, © 2010 INFORMS 107

Figure 1 Total Landed Cost (TLC) Is the Total Cost Incurred From Source to Destination

3% 2% 100%
3%
7%
8%
9%
68%

Typical COGS Supply chain

+ service
Illustrative

Material Labor/service Overhead Working Outbound Customs, Inbound TLC

capital freight duties, and taxes freight

3. Class Part B: The Mexico-China product is launched in the fourth period, allowing
Sourcing Game the groups to build pipeline inventory from both
This section covers the main part of the class by sources before sales start. The product sells for $10,000
describing the different stages of the game (about 1 to per unit. The starting version of the game assumes
1 12 hours of class time). that any unit demanded but not available on hand
is backlogged for a cost of $20 thousand per unit
3.1. Game Setup (Mostly Done in Advance of and each team plays in a separate yet identical
Class) market.
The idea behind the game is for students to act as All teams start with zero money in the bank and
sourcing managers of a firm making periodic order- can borrow to finance inventory. Any bank balance
ing decisions for a new product. Demand is highly (both negative and positive) incurs a 1% interest rate
uncertain and a probabilistic forecast is provided. per period. All team bank balances are projected
Setting: A class with as few as 5 or as many as on the classroom screen continuously (which allows
60 students is divided into 5 to 10 groups. teams to compare their position to other teams and
Props: Each student group needs (at least) one lap- creates a sense of competition and excitement). Any
top computer. Each group receives a log-in code for leftover inventory at the end of the game is liqui-
the website hosting the game.2 The instructor needs dated at zero value (although the instructor can easily
a laptop that is connected to the Internet and to a change the salvage value).
classroom screen projector. The objective for the teams is to maximize their
Setup: Similar to the Mexico-China minicase, each firm value, which is their bank balance, at the end of
team represents an identical company that is intro- the game. The termination time of the game is deter-
ducing a new product that can be sourced from China mined by the instructor but is not told to students in
or Mexico. Each team manages two assets: cash and advance to prevent end-of-horizon effects.
inventory.
The demand distribution for the product is known 3.2. Kickoff and Brainstorming (Between 15 and
and shown to all groups in advance. (The specific dis- 30 Minutes of Class Time)
tribution that we have used is gamma with mean 10 The goal of this stage is to allow each group to
and standard deviation 15, reflecting a highly volatile develop a sourcing strategy. Each group is given
demand.) The actual demand realization is identical an Excel spreadsheet that allows students to simu-
for all groups and projected onto the classroom screen late and test their strategies. Specifically, the spread-
dynamically over time (see Figure 2). sheet shows how the pipeline inventory, sales, and
Each team can place an order to Mexico and an bank balance evolve given user-entered orders and
order to China in each period; sources differ in cost demand.
and in lead times. Specifically, orders placed to Mex- Groups are asked to think through how they would
ico have a sourcing cost of $8,000 per unit and arrive react (in terms of placing orders), given a new
at the beginning of the subsequent period. Orders demand realization and a certain inventory status.
placed to China have a sourcing cost of $7,000 per To steer their thinking, each group is also asked
unit but arrive only after four time periods. The to estimate the fraction of orders they will place
to China and to Mexico and hand this in to the
2
For access to the game, please contact the authors. instructor.
 Allon and Van Mieghem: The Mexico-China Sourcing Game: Teaching Global Dual Sourcing
108 INFORMS Transactions on Education 10(3), pp. 105–112, © 2010 INFORMS

Figure 2 During the Game, the Time as Well as Corresponding Demand and Bank Balances Are Projected on the Classroom Screen

Results for Period 24

Instructor options: Periods Results Edit demand/periods Edit teams Data download MASTER RESET
Page reloads automatically
Total team value
600 K
Period:

24 480 K

360 K

($)
240 K
Demand:

0 120 K

Athens

Brussels

Chicago

Daker

Edmonton

Fuxin
Team order status: Athens , Brussels , Chicago , Daker , Edmonton , Fuxin

3.3. Playing the Dual-Sourcing Game (Between Figure 3 for a screenshot of the team’s Web browser
45 to 60 Minutes of Class Time) interface).
The game starts with a four-period prelaunch stage (3) Previous orders are received automatically.
during which students can fill their pipeline before (4) Teams place orders to the two sources.
sales start. Demand and sales start following the (5) Each team’s pipeline inventory status and its
prelaunch stage. During each period, all groups are bank balance are updated automatically for the next
informed simultaneously of the same period demand period.
as shown in Figure 2. (Showing all students the same The instructor terminates the game at a time of his
demand stream creates a positive class atmosphere, or her choosing. (We have typically played 30 to 35
manifested by students shouting “YES!” or “Oh Jeez” periods, for about 50–60 minutes.) It is critical not to
inform students of the termination period in advance
etc., where all students together experience the game.)
to prevent end-of-horizon effects.
In each period, the following typical inventory
actions take place:
(1) Teams observe the period demand (which is 4. Class Part C: Debrief and
zero during prelaunch). Connection to Recent Research
(2) Sales are determined automatically as the min- This section describes the last part of the class during
imum of period demand and on-hand inventory (see which we debrief the game insights, discuss recent

Figure 3 Screenshot of a Team’s Web Browser Interface Where Students Can Place Orders and Track Their Performance
Allon and Van Mieghem: The Mexico-China Sourcing Game: Teaching Global Dual Sourcing
INFORMS Transactions on Education 10(3), pp. 105–112, © 2010 INFORMS 109

Figure 4 A Dashboard That Summarizes Financial (Top Row) and Operational (Bottom Row) Metrics of All Teams Is Projected on the Classroom
Screen During the Debriefing of the Game

Game Over. Final Results.

Instructor options: Periods Results Edit demand/periods Edit teams Data download MASTER RESET Log out

Quantity sourced from China

Total sales and Mexico Team value
300 400 600 K

240 320 480 K

180 240 100 146 360 K

150

($)
261 75
120 160 240 K
116
80 180 140 163
60 133 120 K
55 48
0

Athens

Brussels

Chicago

Daker

Edmonton

Fuxin
0 0
Athens

Brussels

Chicago

Daker

Edmonton

Fuxin

Athens

Brussels

Chicago

Daker

Edmonton

Fuxin
academic research, and summarize takeaways (about 4.2. Recent Academic Research: TBS and Dual
30 minutes of class time). Index Policies
The objectives of the game include answering the
4.1. Debriefing the Game following two questions: (1) What features do effec-
After ending the game, the instructor asks differ- tive dual-sourcing policies share? and (2) how can we
ent groups (including the winning team, the runner- determine near-optimal policies and average sourcing
up, and a low performer) for their sourcing strategy. allocation?
Teams are encouraged to share their thoughts and Effective dual-sourcing policies take into account
experiences on the benefits of their strategy and what not only the on-hand inventory and the current
they would do differently next time. demand but also the entire pipeline status, as well as
During that discussion, every argument is vali- the entire demand forecast. Research has shown that
dated against metrics compiled during the game. The the structure of the optimal policy is very complex.
instructor shows a dashboard (Figure 4) that summa- Therefore, the dual-sourcing literature has tradition-
rizes the performance of each team along financial ally focused on determining sophisticated dynamic
policies that approach optimal performance. Typi-
and operational metrics:
cally, these policies are characterized by one or two
(1) Financial metrics include the value of each
target inventory levels (base-stock levels) and keep
team’s firm (bank values at the end of the game),
track of one or two inventory positions (indices).
costs (total number of units sourced), and revenues
For example, Veeraraghavan and Scheller-Wolf (2006)
(the cumulative number of units sold).
show that a dual-index policy with two target levels
(2) Operational metrics include each team’s actual
performs close to optimally, which represents state-
strategic allocation (fraction of total units sourced of-the-art dual-sourcing research.
from each location), average on-hand inventory, and Unfortunately, determining the target levels of
service level measured by the fill rate (total number sophisticated policies such as the dual-index policy
of units sold divided by the total demand). requires sophisticated computational work (optimiza-
We are also able to show the actual orders placed tion via simulation). In addition, under these poli-
over time from each source for each group. cies, the associated strategic sourcing allocation can
The discussion leads to the observation that differ- be determined only through simulation. For a strate-
ent teams may perform equally well financially and gic sourcing manager, it would be desirable to have a
employ different strategies. The groups usually differ simple policy with simple guidelines that determine
with respect to average on-hand inventory as well as the strategic allocation during the planning phase
the fraction of units sourced to China versus Mexico. without significantly compromising performance.
The common thread for all successful groups is the Recently, Allon and Van Mieghem (2010) have pre-
ability to execute the strategy and not change it in the sented such a policy that is used in practice: the tai-
face of lower (or higher) than expected demand or lored base-surge (TBS) policy. Under this policy, a
inventories. Determining the features of a good dual- constant (standing) order is placed to the low-cost
sourcing policy and the guidelines provided by aca- supplier in each period. The responsive source is used
demic theory are the focus of the next topic. only to bring the total inventory (on-hand + pipeline)
 Allon and Van Mieghem: The Mexico-China Sourcing Game: Teaching Global Dual Sourcing
110 INFORMS Transactions on Education 10(3), pp. 105–112, © 2010 INFORMS

Figure 5 A Tailored Base-Surge Policy Orders a Constant Fraction (3) the supply-demand volatility
is small. This
From the Low-Cost Source and Orders From the Responsive arises with stable or high-volume products (in the
Source to Respond to Surges maturity phase of the product life cycle) and requires
Volume
stable (level) production in China. This factor also
Supplier 1 with key competency
shows how the sourcing allocation should change as
Surge demand
focused on responsiveness a product moves through its product life cycle.
Students are asked to compute the strategic alloca-
Supplier 2, cost efficient but
tion using the square-root formula for the parameters
Base demand less flexible of the game, which is easy:
• h = about $7,000∗ 1%/period = $70/period,
Time • c = $1000
• COVD = 15
so that the√near-optimal base allocation is approxi-
to a single target level. The TBS policy thus echoes a
mately 1 − 007/2∗ 15 = 1 − 028 = 072.
fundamental tenet in strategy: it aligns the ordering
(As a homework assignment, the instructor can ask
patterns with the core competencies of the suppliers for a similar computation for each of the different
(Figure 5). The constant base allocation allows China SKUs in the minicase.)
to focus on cost efficiency whereas Mexico’s quick In the class we also compare the performance of
response is utilized only dynamically to guarantee different classical policies for a finite horizon prob-
high service. lem, similar to the game. Table 1 compares the profits
Besides its simplicity, the TBS policy directly deter- of four strategies: single sourcing from Mexico, sin-
mines the strategic allocation, which equals the stand- gle sourcing from China, the tailored base-surge pol-
ing order divided by the average demand. Janssen icy, and a dual base-stock policy. For each policy, we
and de Kok (1999) study a similar policy to the TBS have optimized via simulation over the policy param-
and show using numerical study that a firm should eters. For single-sourcing policies we computed the
allocate close to 90% of its demand to the cheaper optimal base-stock policy. For the tailored base-surge
source. Allon and Van Mieghem (2010) optimized the policy we used the formula given in class, and for
TBS policy analytically and present a simple square- the dual base-stock policy we used, again, simulation-
root formula to specify the near-optimal strategic based optimization to compute the optimal threshold
allocation: levels. For each policy we present both the value a

firm would have obtained using that specific policy
h
Base allocation  1 −
 and the specific demand sample path played in the
2c game, as well as the expected value corresponding to
where h is the unit holding cost, c is the unit sourc- the demand distribution.
ing cost differential,  is the demand rate, and
is We also compare these policies with the case in
the supply-demand volatility.3 Simple formulae also which the firm has perfect demand information. With
exist for the corresponding target inventory level and perfect demand foresight, the optimal policy is sim-
cost. When focusing only on demand volatility, the ple: order the exact demand quantities four periods
formula simplifies further: early from China. The comparison with perfect infor-
mation highlights the dramatic cost of uncertainty.


h As one may expect, the dual base-stock policy out-
Base allocation  1 − COVD  performs the other policies, in expectation. However,
2c
on the specific sample path played in the game, the
where COVD is the coefficient of variation of the TBS generates a higher profit. It is important to note
single-period demand distribution.
The formula also provides insight by identifying Table 1 Comparison of Four Sourcing Strategies
the key drivers of dual sourcing and quantifying their
interaction. Specifically, the allocation to the low-cost Optimal target Actual value
Expected
Strategy inventory levels game demand ($) value ($)
supplier is high when
(1) the key monetary trade-off c/h is large, which Single source Mexico 23 557 422
implies a large cost advantage or a small cost of cap- Single source China 48 423 569
Tailored base-surge 18, standing order 616 547
ital (small commercial risk);
to China = 5
(2) the expected demand rate  is high; Dual base-stock 10, 45 514 586
Perfect information Order exact demand 2
321 1
878
3 + single four periods earlier
Specifically,
2 is the sum of the squared coefficients of variation
source China from China
of the interdemand times and the China supply times.
Allon and Van Mieghem: The Mexico-China Sourcing Game: Teaching Global Dual Sourcing
INFORMS Transactions on Education 10(3), pp. 105–112, © 2010 INFORMS 111

that, in expectation, single sourcing from China in this The game also provides a background to illustrate
game would result in higher profits than the TBS pol- the benefits that companies have experienced after
icy. However, our analytic model of the TBS does not transitioning from sole to dual sourcing:
capture dynamic control from China, because it essen- • Significant air freight reduction—by moving to
tially assumes the lead times are too long to allow for dual sourcing some companies have seen 70% +
closed-loop control. Therefore, a single-sourcing pol- reductions in air freight.
icy that does allow for dynamic control may do better • Faster response to changes in demand, as
than the TBS policy. (Also note that the significant dif- opposed to missing demand. Under single sourc-
ferences between the expected value of a policy and ing from a low cost source, by the time a demand
the profit under the specific realization of demand are change was recognized, it was too late in the season
attributable to the high volatility of demand.) to respond.
• Improved ability to manage new product intro-
5. Summary and Extensions ductions. When products were sole sourced from
This section describes how we conclude the class and an offshore low-cost supplier, six to nine months of
summarizes the contributions of the game and possi- products were in the pipeline before the company rec-
ble extensions, as well as student reactions. ognized poor sell-through, which left them with sig-
nificant obsolete inventory.
5.1. Summarizing Key Takeaways of the Game
Dual sourcing is a surprisingly complex manage- 5.2. Possible Customizations and Extensions of
ment problem with potentially high rewards. The the Game
first step in determining an effective dual-sourcing The game can easily be extended along several dimen-
policy is to consider its entire source-to-destination sions, in order of suggested importance and ease of
cost. Although it is straightforward (yet tedious) to implementation:
compute most components of this total landed cost, Customization: The game software currently allows
the impact on inventory and thus working capital is the instructor to change the names and locations
important yet difficult to assess under dual sourcing. of the sources. For example, the fast source can be
Effective dual sourcing uses suppliers with very domestic and two locations could be France and
different strengths. In this game, we combined a Poland. The instructor can also modify the demand
low-cost (but slow) supplier with a fast (but expen- realization and any other parameters.
sive) source. Effective dual-sourcing policies are also Operational extensions: Changing the backlogging
tailored to the strengths of each source. The tai- assumption in the game and allowing for lost sales
lored base-surge policy has a standing order with
is a straightforward modification to the game. So is
the low cost supplier (allowing that supplier to level
adding a physical holding cost to the current finan-
the workload and reduce costs further) and orders
cial opportunity cost or a fixed ordering cost. In addi-
from the fast source only to react to demand surges.
tion, it is easy to modify the software and allow for
Such a policy allows the firm to reduce total cost by
stochastic lead times (but we found that this makes
placing the majority of orders to the low-cost sup-
the game much harder to understand for students
plier while guaranteeing high service levels by plac-
without offering any new insights).
ing occasional, small orders from the responsive, yet
expensive supplier. Financial extensions: It is straightforward to have a
Although determining optimal dual-sourcing poli- higher interest rate for borrowing than saving. One
cies is computationally complex (and requires simula- can also introduce a default threat by adding a con-
tion), it is valuable—and often sufficient in practice— straint on maximal borrowing.
to have some guiding formulae to support dual- Nonstationary parameters: For example, instead of
sourcing decisions. The square-root formula provides assuming i.i.d. demand, the demand distribution
a simple starting point when determining strategic could reflect the different stages of the product life
sourcing allocations. cycle of introduction, growth, maturity, and decline.
Successful dual sourcing often requires having a The same can be done with prices, costs, exchange
strategy and sticking to it. Having a suboptimal pol- rates, and interest rates.
icy is typically better than continuously changing the Competitive market: For example, any demand
decision process. For example, a strategy should allow unmet by one team spills over to the other teams.
the manager to react to high demands by increasing This adds another layer of complexity, even for a sim-
orders; however, reducing the target inventory level ple single-period, single-source model where one can
(base-stock) when observing a few low-demand peri- choose different excess demand splitting rules; see
ods usually leads to low performance. Lippman and McCardle (1997).
 Allon and Van Mieghem: The Mexico-China Sourcing Game: Teaching Global Dual Sourcing
112 INFORMS Transactions on Education 10(3), pp. 105–112, © 2010 INFORMS

Design-marketing-sales decisions: For example, teams complexity inherent in dual sourcing. After playing
can boost demand by spending money on marketing the game using only intuition, the class enhances their
campaigns and sales incentives. This would necessar- perceived value of the academic research and sim-
ily imply that teams no longer observe a common ple guideline formulae. Students often suggest that
demand. Another extension could be to include prod- the game should be played with a cross-functional
uct and supply network design, similar to the Global team, representing their marketing, sales, financial,
Supply Chain Management Simulation of Enspire Learn- and operational groups. Such experience would con-
ing.4 This simulation focuses on the design of the vey the importance of interfunctional coordination
products and the supply network and allows the play- and collaborative forecasting.
ers to modify the sourcing allocation in a limited man- From an instructor’s perspective, the game is easy
ner. In contrast, our game is simpler and focuses on to explain, has minimal requirements (it needs only a
managing an existing supply network in a dynamic few laptops connected to the Internet), and is simple
setting. to set up. A single instructor can easily run the game
Multiple items or multiple stages: Although we do not with as few as five students or as many as 60 students
suggest this, one could extend to multiple products without needing additional assistants or diminishing
or to a multistage supply chain along the lines of the the effectiveness of the game. The game not only suc-
famous Beer Game. In essence, this would lead to a cessfully achieves the pedagogical objectives, it also
Beer Game with dual sourcing. highlights the value of academic research for a realis-
The game can also be used to conduct experi- tic and important business problem.
ments to test research hypotheses. For example, teams
may stick closer to their original plan and policy if Supplementary Material
they experience the game asynchronously or are not Files that accompany this paper can be found and
informed of other teams’ bank balances. downloaded from http://ite.pubs.informs.org.
Although this shows how the game can be
extended, we strongly advise to play the game in its Acknowledgments
current simple format because it allows students to We are grateful to Cort Jacobi and Ruchir Nanda of Deloitte
map their decisions to subsequent results and thus Consulting for ongoing collaboration and insightful discus-
sions on dual sourcing.
enhances the learning from the game.

5.3. Instructor Experience So Far References

So far, the authors have played the game in about Allon, G., J. A. Van Mieghem. 2010. Global dual sourcing: Tai-
10 business school class settings: with full-time MBA lored base-surge allocation to near and offshore production.
students in an operations strategy elective, with fac- Management Sci. 56(1) 110–124.
ulty members at a research conference, and in several Janssen, F., T. de Kok. 1999. A two-supplier inventory model. Inter-
nat. J. Production Econom. 59(1–3) 395–403.
executive education courses. Next we will play it with Lippman, S. A., K. F. McCardle. 1997. The competitive newsboy.
our doctoral students. We believe the game would Oper. Res. 45(1) 54–65.
also work well with graduate, as well as advanced Van Mieghem, J. A. 2008. Operations Strategy: Principles and Practice.
Dynamic Ideas, Belmont, MA, 230–232.
undergraduate, students in engineering.
Veeraraghavan, S., A. A. Scheller-Wolf. 2006. Now or later: A simple
The response has been uniformly positive. Exec- policy for effective dual sourcing in capacitated systems. Oper.
utives appreciate the realism of the game and the Res. 56(4) 850–864.

4
http://www.enspire.com/simulations/gscms. Also distributed by
Harvard Business Publishing under Prod. #: 6107-HTM-ENG.