the impact of outsourcing on the timing decision for entry
TRANSCRIPT
The Impac
Decision fo
Faculty & Research
t of Outsourcing on the Timing
r Entry into Uncertain Markets
by
S. Ülkü B. Toktay
and E. Yücesan
2003/82/TM
Working Paper Series
The Impact of Outsourcing on the Timing
Decision for Entry into Uncertain Markets
Sezer Ulku
105G Old North, McDonough School of Business
Georgetown University, 20057, Washington, DC
Telephone: (202) 687 0377
Fax: (202) 687 1366
e-mail: [email protected]
Beril Toktay
INSEAD, 77305 Fontainebleau, France
Telephone: +33 1 60 72 44 96
e-mail: [email protected]
Enver Yucesan
INSEAD, 77305 Fontainebleau, France
Telephone: +33 1 60 72 40 17
e-mail: [email protected]
The Impact of Outsourcing on the Timing
Decision for Entry into Uncertain Markets
Abstract
A new production technology becomes available, enabling the introduction of new
products in various end-markets. The opportunity windows in the end-markets are
short in the sense that delaying introduction leads to a loss in expected profits. How-
ever, early entry is risky: Initially, the precision of the demand forecasts for the new
products is low, and the firms learn about the market over time. We consider two
scenarios where the investments into the new technology are made by an original
equipment manufacturer (OEM) selling to the end-market and by a contract manu-
facturer (CM) serving M OEMs, respectively. The decision maker chooses the time
of entry and the capacity that maximize its expected profit. We compare the profits,
time-to-market and expected market size in the two scenarios. We identify time-to-
market as a reason for outsourcing production in certain cases. However, contrary to
the assertion in popular press, outsourcing does not always guarantee faster time to
market.
Key Words: Time-to-market, Outsourcing, Technology Adoption, Uncer-
tainty
1 Introduction
In this paper, we consider the impact of outsourcing on the time-to-market and profit
performance of a firm facing a short opportunity window to introduce a new product
with uncertain demand. The introduction of the new product requires the adoption
of a new process technology. We compare original equipment manufacturer (OEM)
performance under two scenarios where the investments into the new technology are
made by an original equipment manufacturer (OEM) selling to the market and by a
contract manufacturer (CM) serving several OEMs, respectively. The decision maker
chooses the time of entry and the capacity that maximize its own expected profits
over the life-cycle of the technology.
Let us start with two motivating examples. The first example is from the electron-
ics industry. This industry is characterized by “short windows of opportunity, rapid
obsolescence of products, and decreasing prices over time” (Solectron Website). In
the last few years, there has been a major move towards outsourcing production to the
so called Electronics Manufacturing Services (EMS) firms. Among other advantages,
EMS firms propose improved time-to-market and “access to the latest equipment,
process knowledge and manufacturing expertise without making substantial capital
investments.” (Solectron Website)
The second example is from the semiconductor industry. This industry is char-
acterized by product life cycles measured in months and rapidly decreasing prices
in the order of 25-30 percent per year (Smith and Reinertsen 1991, Macher 2001,
Leachman et al. 1999). Therefore, time-to-market is crucial for profitability. Intro-
ducing a new/superior product often requires an investment into new manufacturing
facilities and equipment, and the industry migrates to a new process technology every
2-3 years (Cataldo 2002). For instance, in the microprocessor market,“...the faster
a company can move to 0.1 micron [technology], the better it can scale up its pro-
cessor clock-speeds and introduce chips that consume less power than its previous
generation.”(Electronic News)
Due to short life-cycles of the production technologies and high variability in
1
demand, there is a significant risk associated with such investments. For example,
the 0.13 micron technology is considered as a lost generation: After investments by
many chip companies into the 0.13 micron technology, the demand for chips dipped.
By the time it picked up again, there was a new process technology to replace the
0.13 micron technology (Cataldo 2002).
Integrated manufacturers (OEM) and foundries (CM) co-exist in the semiconduc-
tor industry as well; the difference between the two is that the latter focus only on
wafer production and they are not involved in product design. Integrated manufac-
turers and foundries differ in the timing of the adoption of new process technologies
(Manners 2002, Cataldo 2002).
Time-to-market is defined by Mahajan and Muller (1996) as “the time elapsed
between making the decision to start product development and introduction of the
product into the marketplace.” In this paper, “time-to-market” is the time of adop-
tion of the new process technology. We focus on the delays in the introduction of
new products resulting from the strategic timing of this investment. In particular, we
focus on how the time-to-market and resulting OEM profits are impacted depending
on whether it is the OEM or the CM that invests in building capacity with the new
process technology.
We examine a product market that is characterized by a short window of oppor-
tunity at the end of which superior new products replace previous ones. The OEM
may benefit from early entry because of three reasons: First, a fraction of sales would
be lost if the entry into the market were delayed. Second, competitive intensity is
low during the early phases, and the market becomes more competitive later with an
increasing number of competitors to share the total market volume with. (Increasing
competition may also lead to falling prices over time, but we limit our scope to de-
creasing sales quantities in this paper.) Finally, first mover advantages can accrue to
early entrants (Lieberman and Montgomery 1988, Kerin et al. 1992).
By introducing its product early, the OEM avoids lost sales, benefits from low
competitive intensity and possibly pioneering advantages. However, early entry is
more risky: Initially, the level of uncertainty concerning the market demand for new
2
product is high, and the OEM has the opportunity to learn about the market by
delaying entry. Learning can occur by observing the sales of competitors (Chatterjee
and Sugita 1990, Hoppe 2000), by market research (McCardle 1985) or by following
trade journals (Jensen 1982). The decision maker decides on the entry time consid-
ering both the economic benefits of early introduction and the associated risks.
We consider two scenarios to study the impact of outsourcing on time-to-market
and OEM profit. In the first scenario, the OEM decides on the timing of adoption
and the investment level. At each point in time, it can invest or it can postpone the
investment and learn more about the market. In the second scenario, production is
outsourced to a CM who has invested in the process technology and who serves M
OEMs. In this case, the OEM’s ability to introduce the new products depends on the
CM’s timing and capacity decisions. We compare time-to-market and profits in these
two scenarios and answer the following questions: Does a CM always adopt a new
production technology earlier than an OEM? Under what conditions does outsourcing
lead to better time to market for the outsourcing OEM? How does outsourcing impact
OEM profits? What are drivers other then time-to-market that favor outsourcing?
We identify time-to-market as a reason for outsourcing production in certain cases.
The paper is structured as follows. After an overview of the relevant literature
in §2, we describe our model in §3. §4 provides our results for general functional
forms for the evolution of the expected market potential and the market uncertainty.
A specific functional form is subsequently analyzed for additional insight. Next, in
§5, the implications of relaxing some of the assumptions are discussed. We conclude
in §6 with a summary of results, managerial implications, and a discussion of future
research directions.
2 Literature Review
There is a large body of literature on time-to-market, considering such issues as timing
of innovation, first-mover advantages and impact of uncertainty. Our contribution is
to study the impact of outsourcing on the time-to-market and profit for a firm that
3
faces an uncertain and short-lived market opportunity to adopt a process innovation.
To our knowledge, this is the first paper to consider time-to-market in a two-level
supply chain. In developing our model, we draw on some elements from models in
the literature, as described below.
The literature on R&D races examines “how the expected benefits, R&D costs, and
interaction among firms determine the pattern of expenditure, date of introduction,
and identity of the innovating firm” (Reinganum 1989). Kamien and Schwartz (1982)
and Reinganum (1989) provide excellent surveys for decision theoretical and game
models in the timing of innovation. Once the innovation is produced, there may be
delays in its adoption due to insufficient demand, decreasing cost of adoption over
time, high uncertainty (Tirole 1988) and network externalities (Reinganum 1989).
We assume that innovations are given (a process innovation provided by an external
supplier, and product innovations to be launched at the OEM). The question of
interest to us is when to adopt a process innovation in the presence of economic
benefits to moving early and high initial uncertainty.
The literature on first mover (dis)advantages examines the impact of the order of
entry on firm performance (Lieberman and Montgomery 1988, 1998, Robinson and
Min 2002, Bowman and Gatignon 1996, Lilien and Yoon 1990, Moore et al. 1991,
Kerin et al. 1992, Tellis and Golder 1996). Apart from initial monopoly status,
first movers can benefit from leadership in product and process technology, preempt
resources, and develop buyer switching costs (Lieberman and Montgomery 1988).
However, first movers are subject to some disadvantages. Among these, the resolution
of market and technology uncertainty is “... the major factor affecting the timing of
entry in practice” (Lieberman and Montgomery 1988). Hence, the pioneer’s business
is a high risk and high return one. Typically, the pioneer’s temporary monopoly
and its first mover advantages amount to more than the risks associated with market
and technological uncertainties (Robinson and Min 2002). The pioneer survival rates
increase with the lead time between pioneer entry and the followers (Robinson and
Min 2002, Lieberman and Montgomery 1998). Similarly, fast followers have a higher
chance of survival (Robinson and Min 2002). In our model, we capture first mover
4
advantages with a decreasing sales potential based on the time of entry, without
reference to specific sources of advantage. Similarly, the associated risk decreases
over time.
The uncertainty that surrounds the market potential of a new product is important
in the determination the timing of its introduction (Lieberman and Montgomery 1998,
Lilien and Yoon 1990). Jensen (1982) studies the adoption decision of a single firm
regarding a new technology of uncertain profitability. The success probability of the
firm is updated at zero cost in a Bayesian manner as external information becomes
available. Sources of information can be the innovator itself or trade press. It is shown
that the firm stops collecting information when the posterior probability of success is
above a certain threshold. Firms with different priors adopt at different times, leading
to diffusion, and all firms eventually adopt. In McCardle (1985) information gathering
is costly. The sources of information can be testing of the process, the product or the
market. In the optimal solution, the firm takes one of three possible actions at any
point of time: If estimated profitability is high, the firm adopts innovation. If it is low,
it is rejected. More information is collected if the probability takes an intermediate
value. The firm can end up adopting a bad innovation or rejecting a good one. Hoppe
(2000) and Chatterjee and Sugita (1990) consider adoption and market entry when
learning from the leader is possible. Hoppe (2000) shows that, in a duopoly, firms can
play a preemption game or a waiting game depending on the probability of success
for the adoption. When the probability of success is low, it is preferable to wait and
learn the true value of the innovation by observing the leader. In these papers, and
in the economics literature in general, the firm adopting an innovation incurs a fixed
cost; however, there is no consideration of a constraint on the firm’s ability to fulfill
demand. Only the last two papers take early mover advantages into account. In
our model, we assume that uncertainty concerning the market demand is resolved
costlessly over time and is observed by both the OEM and the CM.
The pertinent papers in the operations management literature focus on quality,
time and cost trade-offs, and the design of processes for product development. Cohen
et al. (1996), Bayus et al. (1997) and Morgan et al. (2001) examine the time-
5
to-market and product performance trade-off. Rushing to the market may reduce
the chances of success, or the product performance may be very different from the
customer preferences. Therefore, the entry time should take into account the firm
characteristics, competitors in the market, profit margins, and the length of the time
window. Furthermore, when the introduction time is fixed, there is a trade-off be-
tween fine-tuning the product to customer preferences and reducing unit costs (Bhat-
tacharya et al. 1998). MacCormack et al. (2001), Eisendhardt and Tabrizi (1995)
and Ward et al. (1995) focus on the design of development processes and suggest the
use of flexible processes to shorten time-to-market in dynamic and highly uncertain
markets.
Finally, an important element of our model is the possibility of outsourcing. The
decision to vertically integrate or outsource has been studied extensively in the eco-
nomics and strategy literatures. Three recent papers in the operations management
literature examine reasons for outsourcing. Van Mieghem (1999) considers a two-
firm setting where an OEM faces the choices of manufacturing in house, or fully or
partly outsourcing. He evaluates the option value of subcontracting and examines the
ability of price-only contracts, state-dependent contracts and incomplete contracts to
coordinate the channel. Plambeck and Taylor (2001) consider the trade-off between
pooling effects and the incentive to innovate when manufacturing and innovation take
place in different organizations, at the CM and OEM, respectively. They conclude
that outsourcing production to CMs does not improve profits, unless the OEM has
higher bargaining power. Otherwise, the gains from production efficiency are more
than offset by the decrease in the investment into innovation, and it may be better
to share capacity with other OEMs. Finally, Cachon and Harker (2002) examine
competition between firms in the presence of economies of scale. In such a context,
competition is fiercer as firms can lower their costs with a higher throughput. The
authors identify economies of scale as a motivation for outsourcing, since it mitigates
competition by eliminating the dependence of costs on volume. In the same spirit,
our paper explores time-to-market as a reason for outsourcing.
6
3 Model Description
This paper examines the impact of outsourcing on the time-to-market and profit per-
formance of a firm facing an uncertain market opportunity that decreases over time.
The precondition for entry is investment into a process innovation which becomes
available at time t = 0 and enables the introduction of new products in a variety of
markets. Time-to-market is defined as the adoption time of the process technology.
We do not consider other time-related advantages such as shorter production cycle
times or zero/shorter lead-times for capacity building and process perfection.
Two scenarios are considered. In the first scenario, an OEM performs production
in-house, and considers adopting the process innovation in order to introduce a series
of new products. In the second scenario, a CM has M ′ OEM customers, and M of
them can use the innovation. The OEMs can introduce their products to the market
only after the CM adopts the innovation. We assume that all the OEMs are identical,
that is, they all have the same cost parameters and their demands are drawn from
independent and identical normal distributions. The OEMs served by the CM are
not in direct competition with each other, even though they operate in competitive
product-markets. This assumption is valid when the CM has clients operating in
different industries or geographic markets.
An OEM entering the market at time t faces a market potential that is normally
distributed with mean S(t) and standard deviation σ(t). S(t) is the total expected
demand for the product from the time of entry, t, until the end of the time window
TD, whereas σ(t) is the standard deviation of the total demand during [t, TD]. The
coefficient of variation at time t is ν(t) = σ(t)S(t)
. Both the CM and the OEM have
access to the same information about the market.
We assume that the expected demand for the firm’s product strictly decreases with
the time of entry (S ′(t) < 0) until it drops to zero at time TD due to three reasons.
First, there is a finite opportunity window for commercializing the product, at the
end of which new products replace it, irrespective of the time of introduction. As
the product introduction is delayed, the sales opportunity for some customers is lost.
7
Second, competition is milder in the beginning, hence early entrants command higher
market shares, whereas competition becomes increasingly fiercer with new entries into
the market. Finally, there may be advantages that accrue to early-movers.
We allow S(t) to take different forms. A concave decreasing S(t) implies that the
demand rate is growing over time, and the initial periods are not very important,
since the sales are slow. A convex decreasing S(t) would be observed in a market
where the loss due to delays in introduction decreases over time. This is the case in
a market with pioneering advantages, where delays at early stages are costly. S(t)
can also take the shape of an inverted S-curve, if cumulative sales follow a diffusion
curve.
Even though there may be a high market potential at the early stages of the
time window, the firm knows very little about the new market opportunity; entry
at this stage is more risky. Therefore, the firm may delay entry and expand its
knowledge set prior to making a decision (Jensen 1982, McCardle 1985, Mamer and
McCardle 1987). However, there are diminishing returns to each additional piece
of information (McCardle 1985, Griffin and Hauser 1993), and the firm may stop
collecting information at some point and enter the market. We do not explicitly
model the acquisition of information. Rather, we assume that the standard deviation
for the sales potential is convex decreasing (σ′(t) < 0 and σ′′(t) > 0) to reflect the
fact that the market uncertainty is resolved over time, but at a diminishing rate.
The decision maker knows the past values of σ and S and that both are mono-
tonically decreasing in t (the market uncertainty will decrease and the expected total
sales will decrease as time progresses). However, the trajectories of S(t) and σ(t) are
not known to the decision maker. Therefore, it can not find the optimal t using the
first order conditions over t. We assume that the entry decision is taken based on
local information: At each point in time, the decision maker evaluates entry, given
the market uncertainty at that point of time. To this end, it solves a Newsboy prob-
lem, and finds the capacity K that balances the costs of over- and under-investment,
namely, excess capacity and lost sales. The resulting expected profit is compared to
the previous expected profit value. The entry decision is taken when the change in
8
profits is no longer positive. The investment level is the one that maximizes profits,
given the market information at that point of time. As we will show, this myopic deci-
sion rule is optimal when the expected profits are concave, convex or concave-convex
in t. In other cases, the entry time can be suboptimal.
Upon entry, the firm makes an irreversible commitment with a one-shot investment
into the process technology. Hence, there is an up-front fixed cost at entry that
depends on the capacity level. Any other fixed cost independent of the capacity level
is normalized to zero. Furthermore, the lead-time for building production capacity is
assumed to be equal for all OEMs and the CM, and is normalized to 0.
We assume that OEMs and the CM are risk-neutral and maximize expected prof-
its. In Scenarios I and II, the OEM and the CM, respectively, maximize their expected
profit as a function of capacity and time.
The economic parameters in the model are unit revenue (r), up-front investment
and subsequent production costs for the OEM and the CM ((cI , cp) and (c′I , c′p) re-
spectively), and the unit wholesale price charged to the OEM by the CM w. All the
economic parameters are assumed to be given. We subsequently perform a sensitivity
analysis to study the impact of these parameters.
The OEM earns a fixed unit revenue r for each item sold. Investment (cI , c′I)
and production (cp, c′p) costs are incurred by the party performing production. The
investment cost is incurred before demand realization and is linear in the capacity.
Production cost includes labor and material and is also linear in the realized output.
The OEM and the CM are assumed to have different unit costs. The source of the
difference in investment costs may be economies of scale at the CM due to higher
total capacity (cI > c′I). Similarly, the CM may have lower production costs (c′p < cp)
because of economies of scale in purchasing and economies of learning resulting from
a higher total volume of production.
When production is outsourced, a wholesale price w is charged by the CM for
each unit of production. Finally, the salvage value is normalized to zero.
As the capacity at the CM is allocated among multiple OEMs, an allocation rule
needs to be defined. Since allocation mechanisms are not of primary interest here, we
9
assume that a fair allocation rule is used, described in Netessine and Rudi (2001) as
“any inventory rule (deterministic or probabilistic) that does not give a preference to
any particular OEM”. Homogeneous OEMs achieve identical expected profits under
a fair allocation rule.
In §5, we discuss how our results would be impacted by relaxing some of these
assumptions.
4 Analysis
Several issues are analyzed in this section. We start with the timing decision of a firm
facing a short opportunity window of uncertain potential (§4.1). The OEM has the
option of outsourcing production, and thereby avoiding the investment in the face of
uncertainty. However in this case, it is dependent on the CM to invest into capacity,
and product introduction does not take place until the CM makes its investment.
We compare time-to-market in these two scenarios, and study their dependence on
costs, revenues, learning rate, decay rate for the market potential and the scale of the
contract manufacturer. Next we examine the value of outsourcing to the OEM, and
underline different motivations for outsourcing (§4.2). Finally, for additional insight,
we consider an example where S(t) and σ(t) have exponential functional forms (§4.3).
All proofs are provided in the appendix.
4.1 The Optimal Time of Investment
The optimization problem of the OEM and the CM in Scenarios I and II, respectively,
are similar. In Scenario I (II), the OEM (CM) maximizes its expected profit as a
function of capacity and time. This corresponds to solving a Newsboy problem at
each point in time and choosing the time that results in the highest Newsboy profit.
At time t∗I (t∗II), each OEM (CM) invests into the new process technology and starts
producing and selling the product.
Two trade-offs are faced by the decision makers, which are resolved by choosing
the time of entry and the capacity respectively. The market potential is higher with
10
earlier entry, however the variability is also higher, whereas both are lower if the firm
postpones investment. Time of entry is chosen to balance the profit improvements re-
sulting from early entry with the associated risks. Furthermore, demand is uncertain:
demand realization can exceed or fall short of the level of investment. This trade-off
is solved by choosing capacity to maximize the expected profit function, which is
concave over capacity for each t. The optimal capacity at time t can be found by
solving the first order condition for K, using the market information (S(t), σ(t)).
Integrated OEM
In Scenario I, the OEM is integrated into production. Hence, the OEM itself
decides on the time of investment and the capacity to be built. The expected profit
for an OEM that invests at time t into capacity K is as follows:
πI(KI , t) = −cIKI + S(t)(r − cp) − (r − cp)
∫ ∞
KI
(D(t) − KI)dF (D(t)).
The first term in this expression is the cost of investment into the new technology;
the second term is the total revenues minus the production costs in the absence of
uncertainty; the final term is the expected loss due to uncertainty. The overage cost
for the OEM is cI , the cost of a unit of excess capacity investment, and the underage
cost is r − cp − cI , the opportunity cost of a unit of lost sale.
Lemma 1 At time t, the profit maximizing capacity is K∗I (t) = S(t) + zoσ(t).
Furthermore, the expected profit on the optimal trajectory is given by
πI(K∗I (t), t) = mo (S(t) − ξ(zo)σ(t)) where mo = r − cp − cI , zo = Φ−1(1 − cI
r−cp) and
ξ(zo) = φ(zo)Φ(zo)
.
Using Lemma 1, the maximization problem can be rewritten as
t∗I = arg maxt≥0
(S(t) − ξ(zo)σ(t)). (1)
While the profit function may have many local maxima in the general case, the
optimal entry time can be easily characterized in certain cases.
11
Proposition 1 If S(t) and σ(t) are three-times differentiable; S ′(t) < 0, S(TD) = 0,
σ′(t) < 0 and σ′′(t) > 0, the optimal time of entry t∗I is given by one of the following
mutually exclusive cases:
(i) t∗I = ∞ if S(t) < ξ(zo)σ(t)∀t ∈ (0, TD).
(ii) t∗I = 0 if S ′(t) < ξ(zo)σ′(t)∀t ∈ (0, TD); or if πI(0) > 0 and S ′′(t) > ξ(zo)σ
′′(t)∀t ∈(0, TD); or S ′′′|π′′
I=0 > ξ(zo)σ
′′′|π′′
I=0 and πI(0) > πI(t0).
(iii) t∗I = t0 ∈ (0, TD) if S ′′ < ξ(zo)σ′′ ∀t ∈ (0, TD) and ∃to ∈ (0, TD) such that
π′I(to) = 0 ; or if S ′′′|π′′
I=0 < ξ(zo)σ
′′′|π′′
I=0; or if S ′′′|π′′
I=0 > ξ(zo)σ
′′′|π′′
I=0 and
πI(0) < πI(t0).
(iv) For more general S(t) and σ(t), the optimal t∗I can be found through line-search.
Proposition 1 characterizes the optimal time of entry for various conditions on
S(t) and σ(t). The first case states that if the expected profit is never positive, the
firm should never enter. The second case says that entry should be immediate if
the profit function is strictly decreasing, or convex, or convex-concave with the profit
at time 0 greater than the profit at the interior maximum. Stated in terms of the
primitives S(t) and σ(t), when πI(0) > 0, the firm should enter immediately if the
ratio of time derivatives of market potential and uncertainty is always less than ξ(zo),
or if the market potential is more convex than the uncertainty scaled by ξ(zo). In this
case, information that can be obtained about the market does not justify the delay
in entry. The third case indicates that it is optimal to enter after a delay if the profit
function is concave, or concave-convex, or convex-concave with the expected profit at
time 0 less than the profit at the interior maximum.
The condition on the third derivative allows the profit function to switch between
concavity and convexity at most once. This condition is useful, for instance, when
the difference of the two convex curves inherits convexity for some ranges of t and
concavity for some others. The resulting profit function has at most one interior
maximum. Some diffusion type S(t) specifications satisfy this condition. For instance,
when S(t) is specified with the Bass diffusion model S(t) = S0(a + 1) e−γt
1+ae−γt (Bass
12
1969) and σ(t) = σ0e−βt with β > γ, the profit function is concave-convex (case iii)
and the optimal time of investment is the smaller root of the derivative of the profit
function.
For all but one of the specific cases described above, the myopic decision rule yields
the optimal entry time. Entry is premature in the case where the profit function is
convex-concave and the profit at time 0 is lower than at the interior maximum. In
this case, the decision maker is misled by the decrease in the profits at t > 0 and
decides to enter at t = 0 even though the optimal time is yet to come.
Virtual OEM
In the second scenario, the CM invests into production capacity while no invest-
ment is made by the OEM. The timing and capacity are chosen to maximize the CM
profits. In this scenario, the OEM is dependent on the CM, since it can commercialize
its products only after an investment is made by the CM.
The CM may have two advantages over OEMs: A CM may have a lower cost of
delivery for deterministic demand; that is, it may benefit from deterministic efficiency
(cI > c′I , cp > c′p). Furthermore, a CM serving multiple customers has stochastic
efficiency, arising from risk pooling over a broader customer base (M > 1).
The CM profit function is similar to that of the OEM in Scenario I, only with
different parameters:
πcmII (K, t) = −c′IK + MS(t)(w − c′p) − (w − c′p)
∫ ∞
K
(D(t) − K)dFcm(D(t)).
Lemma 2 At time t, the profit maximizing capacity is K∗(t) = MS(t) + zc
√Mσ(t).
Furthermore, the expected profit on the optimal trajectory is given by
πcmII (K∗(t), t) = Mmc(S(t) − ξ(zc)√
Mσ(t)) where mc = w − c′I − c′p, zc = Φ−1(1 − c′
I
w−c′p)
and ξ(zc) = φ(zc)Φ(zc)
.
Using Lemma 2, the maximization problem can be written as
t∗II = arg maxt≥0
(S(t) − ξ(zc)√M
σ(t)).
13
This is the same expression as Equation (1), only with a different multiplier for
σ(t). Therefore, when ξ(zo) is replaced by ξ(zc)√M
, Proposition 1 characterizes the opti-
mal investment time of the CM for different conditions on S(t) and σ(t).
The expected OEM profit in this scenario is
πoemII = (r − w)(S(t) − 1√
Mσ(t)L(zc)), (2)
where L(z) is the standard loss function. The first term in this expression is the OEM
profit in the absence of uncertainty. The second term reflects the expected profits
lost due to smaller CM capacity than the total demand. Since capacity is rationed
fairly among OEMs, expected loss is shared equally. On the other hand, OEMs do
not bear any portion of the over-investment risk in this scenario.
The following property will be useful in understanding the impact of the economic
environment on the time of entry and comparing the entry times of different parties.
Lemma 3 ξ(z) ∈ (0,∞) and it decreases in z. Equivalently, ξ(z) decreases in the
unit revenue (dξ(z)∂r
< 0) and increases in costs (dξ(z)∂cp
> 0 and dξ(z)∂cI
> 0).
Lemma 3 states that lower values of ξ(z) correspond to a more desirable economic
climate with higher unit revenues, lower unit investment or production costs.
Let us assume that the optimal time of entry is 0 < t < TD. The following
proposition summarizes the results regarding the time of entry.
Proposition 2 If OEM (CM) profit πoemI (t) (πcm
II (t)) is maximized at t∗ ∈ (0, TD),
the following is true for the optimal investment time t∗:
(i) t∗ increases with ξ(z0) and ξ(zc), and decreases with M .
(ii) t∗ increases with S ′(t), the rate of decrease in market potential.
(iii) t∗ decreases with σ′(t), the rate of resolution of market uncertainty.
In addition, there exists a ξ0 such that t∗ = 0 if π(0) > 0 and ξ < ξ0.
14
To summarize Proposition 2, the time of entry depends on margins (ξ(z)), the
rate of decrease in the sales potential over time (S ′(t)) and the rate of resolution of
the market uncertainty (σ′(t)). In addition to these factors, the number of OEMs
(M) also affects the decision when it is taken at the CM. Finally, if the margin is
high enough, entry is immediate (t = 0).
By Lemma 3, increasing ξ(z) implies decreasing margins for the decision maker.
In a market with low margins, the willingness of the decision maker to take risks
is limited. The gains from early entry are not high enough to offset the associated
risks. Therefore, investment into the new technology is delayed until there is more
information available about the demand. On the other hand, the decision maker has
a lot to lose with late entry in a market with high margins (low ξ(z)); thus, the
investment is done earlier.
The adoption time of the CM is accelerated with a larger number of OEMs served.
This is due to pooling at the CM. With increasing M, the variance of the market
potential for each t decreases. Therefore, t∗ decreases with increasing M .
Everything else being equal, the optimal investment time is earlier in a market
with a rapidly decreasing potential. Such a case might correspond to having a shorter
life-cycle with the same initial potential. In such a market, products become obsolete
quickly and firms have a short time interval within which to commercialize products.
Similarly, a firm is expected to enter earlier if competitive intensity is expected to
increase quickly, eroding the market potential for the firm. On the other hand, if
the firm is a monopoly and can control the obsolescence time of the product to be
introduced, the entry can be delayed until more information is gathered about the
market.
The rate at which uncertainty is resolved also affects the time of entry. Entry is
earlier in a market where the rate of improvement for demand information is slow
(σ′(t) is close to zero). In this case, delaying entry leads to losses in sales. However, the
increase in the accuracy of demand information obtained in exchange is not significant.
Therefore there is little reason for delaying entry. Finally, beyond a threshold ξ0, all
ξ lead to immediate adoption.
15
Even though the market characteristics are the same for both, the OEM and
the CM differ in two ways. First, the OEM and the CM have different economic
parameters, summarized by ξ(zo) and ξ(zc). Second, the CM has the ability to pool
over different customers, hence faces lower overall demand variability. Because of
these reasons, the OEM and the CM make different decisions regarding entry time
and capacity. The following proposition compares the two in terms of their chosen
time of investment.
Proposition 3 OEM and CM entry times can be compared as follows:
(i) If ξ(zo) < ξI0 and ξ(zc) < ξII
0 , t∗I = t∗II = 0.
(ii) Otherwise, if ξ(zo) < ξ(zc)√M
, then t∗I < t∗II (and if ξ(zo) ≥ ξ(zc)√M
, then t∗I ≥ t∗II).
Since time-to-market cannot be reduced below zero, any ξ(zo) (ξ(zc)) below the
threshold ξI0 (ξII
0 ) leads to immediate entry. Beyond these threshold values, there is
no difference between the OEM and the CM in terms of time-to-market, but there is
still a profit differential as the capacity decision is not the same for fixed t.
Given that the OEM and the CM have access to the same market information, the
difference in the times of investment arises from differences in economic parameters
and pooling. When the CM serves a single customer (M = 1), the pooling effect dis-
appears and the one with a better newsvendor ratio invests earlier. If the investment
and production costs at the OEM and the CM are equal (c′I = cI , c′p = cp), this party
is the OEM unless w = r. This changes as M increases. As the pooling advantage
becomes available to the CM, t∗I > t∗II becomes possible for w < r, even with identical
unit cost parameters.
A CM may have cost advantages due to larger volumes and experience in produc-
tion (c′I < cI , c′p < cp). First of all, the capacity cost can be lower at the CM. Due to
higher total capacity (possibly including facilities other than the one under study),
the CM may benefit from economies of scale in building capacity. Second, the CM
may have lower production costs (c′p < cp) due to economies of scale in purchasing,
and economies of learning resulting from a higher aggregate volume of production.
16
Any of these unit cost advantages leads to an improvement in the CM margins, and
therefore to earlier investment.
Our first conclusion is that outsourcing does not always lead to a better time
to market for an OEM as compared to in-house production. Faster time to market
depends on the market characteristics (S(t), σ(t)), unit margins at the CM and the
OEM and finally the scale of the CM.
Let us now compare two CMs in the light of Propositions 2 and 3. Higher degrees
of deterministic and stochastic efficiency, that is, lower (c′I , c′p) and higher M lead to
faster adoption of new technologies at the CM. Thus, a faster time to market can
be expected from an efficient CM, in addition to the more obvious cost advantage.
Therefore, power considerations aside, a CM of larger scale is a better choice for the
OEM seeking faster time to market.
4.2 Other Motives for Outsourcing
In the previous section, it was shown that outsourcing does not necessarily lead to
faster time-to market. In this section, we identify other motives for outsourcing by
dissecting OEM profits.
OEM profits in the two scenarios are given by
πoemI = (r − cp − cI)S(t∗I)[1 − ν(t∗I)ξ(zo)] and
πoemII = (r − w)S(t∗II)[1 − ν(t∗II)
L(zc)√M
],
where ν(t) = σ(t)/S(t). The OEM chooses to outsource production if the expected
profits are improved as compared to in-house levels, that is, if there is a gain from
outsourcing, where gain is defined as Gπ =πoem
II
πoemI
. The two scenarios differ in four
dimensions: the unit costs faced by the OEM for each unit delivered to the market,
the time of entry and consequently the market potential, the variability faced at the
time of entry, and finally the opportunity cost of unit uncertainty to the OEM for unit
value. Outsourcing gain can be written as a combination of these four components:
Gπ =πoem
II
πoemI
= GmGS
(
1 − GRGνν(t∗I)ξ(zo)
1 − ν(t∗I)ξ(zo)
)
,
17
where Gm = r−wr−cI−cp
, GS =S(t∗
II)
S(t∗I), Gν =
ν(t∗II
)
ν(t∗I)√
Mand GR = L(zc)
ξ(zo).
The first component, Gm, is the ratio of the OEM margins in the two scenarios for
each unit of product delivered to the market. The wholesale price w can be written as
the sum of CM margin and costs: w = c′I + c′p + mc. Everything else being equal, Gm
is higher for lower costs and lower margin at the CM. That is, higher deterministic
efficiency and lower margins at the CM improve the OEM profits.
The second component, GS, is the relative change in the expected market poten-
tial due to differences in the timing of entry. An early investment by the contract
manufacturer allows the OEM to introduce its products earlier, and achieve a higher
expected market potential. GS depends on the relative newsvendor ratios, the num-
ber of CM clients, and the functional forms of S(t) and σ(t). As opposed to Gm, GS
increases in mc.
The third component, Gν , is the ratio of the demand variability per OEM in the
two scenarios. Like GS, Gν depends on the relative newsvendor ratios, the number
of CM clients, and the functional forms of S(t) and σ(t). At the time of entry, the
variability for each individual market can be higher. Gν < 1 is possible due to pooling
over multiple OEMs.
Finally, GR captures the relative cost of uncertainty in the two scenarios, for unit
margin and unit variability. When the production is in-house, OEM incurs both
underage and overage costs due to uncertainty. When production is outsourced, the
excess capacity risk is fully transferred to the CM. However, the OEM still faces the
risk of not being able to fulfill demand. If total demand is higher than capacity set
by the CM, it is rationed to OEMs; as a result, OEMs may lose sales. GR increases
in zo and decreases in zc. In addition, for zo = zc, GR < 1 since L(z) < ξ(z).
Outsourcing results in improved profits when Gπ > 1. The value of outsourcing
increases in Gm and GS. Higher margins are achieved for each unit delivered, and
the market served is larger due to earlier entry. On the other hand, the value of
outsourcing decreases in GR. In this case, the transfer of risk is not significant due to
insufficient investment at the CM. Even though the cost of excess capacity is borne
by the CM, the OEM faces a higher cost of lost sales. Finally, Gπ > 1 decreases in
18
Gν : due to a lower number of customers, the ability to achieve statistical efficiency is
lower.
4.3 Example: Exponential S(t) and σ(t)
For further insight, let us assume that S(t) and σ(t) are exponentially decreasing
functions of t: S(t) = S0e−γt; σ(t) = σ0e
−βt and the corresponding coefficient of
variation ν(t) = σ(t)S(t)
= σ0
S0
e(γ−β)t. The parameter of the market potential γ can be
interpreted as the rate of obsolescence: the market potential decreases over time with
parameter γ. On the other hand, β is the learning parameter: learning is faster at a
higher β compared to a lower one.
Consider a CM that does not have any cost superiority over the OEM (cI = c′I ,
cp = c′p). Nevertheless, it earns a non-zero profit margin (w − cI − cp > 0). Figure
1 illustrates the gain in profits, Gπ, and the difference in times of entry, ∆t for two
parameter sets. ∆t < 0 and Gπ > 1 correspond to improvements in entry time and
OEM profits due to outsourcing.
Our first observation is that ∆t is non-increasing in w: At low wholesale prices,
the OEM entry is earlier than the CM entry (∆t > 0), and the opposite is true at
high wholesale prices. Above a threshold dependent on M , a further increase in w
does not provide any improvement in ∆t, as t∗II = 0 already. This threshold is lower
for a higher M .
Our second observation is that outsourcing does not always improve OEM profits.
For instance, when M = 1, profits for any w under outsourcing are lower than those
with in-house production, and for M = 5, profits are marginally improved using
outsourcing in a narrow range of w.
Third, there is a wholesale price w∗ that maximizes the gain from outsourc-
ing, and its value is higher than the unit cost of delivery with in-house production
(w∗ > cI + cp). Therefore the OEM benefits from paying a premium to the contract
manufacturer. This premium decreases in the number of OEMs served at the CM
(M).
Fourth, the OEM benefits from a larger scale CM (high M) both in terms time-
19
(a) (b)
(c) (d)
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
60 70 80 90 100
w
M=1
M=5
M=10
M=50
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
70 75 80 85 90 95 100
w
M=1
M=5
M=10
M=50
II Itt t
II Itt t
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
60 65 70 75 80 85 90 95 100
w
M=1
M=5
M=10
M=50
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
70 75 80 85 90 95 100
w
M=1
M=5
M=10
M=50
II
I
GII
I
G
Figure 1: The impact of outsourcing on time-to-market and profits. (Parameters: σ0 = 40,
S0 = 100, β = 1.3, γ = 0.4, r = 100, c′p = cp = 30; c′I = cI = 40 for (a) and (c), c′I = cI = 30
for (b) and (d).)
to-market and total profit gains. The CM does not provide any value to the OEMs
unless it has a critical number of customers.
Finally, even without time (t∗o = t∗c) and unit price (w > cI + cp) advantages
(Gm < 1 and GS < 1), it is possible to have a gain from outsourcing (Gπ > 1). In
this case, the motivation for outsourcing is the risk transfer to the CM. This happens
for higher values of M , where the CM can better pool risks.
5 Discussion
In this section, we discuss the implications of relaxing some of the earlier assumptions.
Fixed Cost of Entry. Any fixed cost of entry independent of the capacity level is
20
normalized to zero in this paper. The effect of such a cost is to lower the profit curves
by a constant amount, possibly making outsourced production the only alternative
for entry. This is the case, for instance, for the fabless start-ups that are supplied
by semiconductor foundries. In 2001, there were about 1300 IC design companies
that did not have production capabilities (Business Week 2001). If the profit after
the fixed cost is still positive, the presence of a fixed cost does not impact the timing
of investment. Therefore the high fixed cost of building a fab does not explain why
Intel, an integrated manufacturer, and TSMC, a foundry, differ in their timing of new
process technology adoption.
Nonzero Salvage Value. In this paper salvage values are normalized to 0. The
CM may have a higher salvage value for capacity that is originally built to serve a
specific market, since it can be put into other uses afterwards, for instance to produce
lower margin products. Since the risks faced are lower, one can expect such a party
to invest earlier and at a higher level.
Substitution. We did not consider substitution effects. On the product side, poten-
tial cannibalization of the markets for old generation products may result in delays
in the introduction of the new product. On the capacity side, it may be possible to
use old process technology for introducing the new products with lower efficiency. In
both cases, investment into new process technology would be delayed further.
Margins Decreasing over Time. Throughout the paper, we have assumed that the
OEM margins are constant. In reality, product price may decrease through the life-
cycle (r′(t) < 0). There may also be an accompanying decrease in the production costs
over time. As long as the difference r(t)−cp(t) is constant, the current results remain
unchanged. If the profit margins decrease over time, entry by an OEM is earlier, as
there is a higher penalty for postponing investment. Unless there is a change in the
wholesale prices, w, the entry time for the CM does not change. Therefore, the time
advantage provided by the CM decreases. In this case, outsourcing is motivated more
by lower risks and increased efficiency, and less by an improvement in time-to-market.
Nonidentical OEMs. The OEMs need not be identical in terms of their unit costs
and revenues. In this case, by Proposition 2, the OEMs with better economics (lower
21
ξ(zo)) invest earlier. If the CM has identical costs across all OEMs and charges
uniform prices, the party that benefits the most from outsourcing, both in terms of
time-to-market and profits, is the one with the worst in-house economics (highest
ξ(zo)).
Asymmetric Allocation. With non-identical OEMs, allocation need not be ‘fair’:
The CM may favor high-margin, high-volume or low-variability customers. An OEM
facing an asymmetric allocation mechanism is dependent on the demands of the higher
priority customers of the CM. Therefore, even if outsourcing production to a CM
can improve time-to-market for such a firm, its profits may decrease due to limited
allocated capacity.
Correlated Demands at the OEMs. If the OEM demands are positively corre-
lated, variability reduction at the CM due to pooling is less significant. The optimal
investment time for the CM increases compared to the no correlation case, resulting
in a smaller market potential. The advantages provided by the CM decrease with
positive correlation. The opposite is true if demand correlation is negative.
Information Asymmetry. Being closer to end markets, OEMs may be able to
reduce uncertainty faster than the CM through more frequent and higher quality
market input. In this case, the time advantage provided by a CM would be lower.
Sequential Investment. In some cases, the firm may have the opportunity to phase
the investments and build capacity sequentially. When this is possible, the firms can
probe the market through smaller-scale initial investments and expand capacity once
the market uncertainty has diminished. This would be possible when the fixed costs
of each additional investment is not prohibitive. The decision maker still makes a
risk-return trade-off, only at a smaller scale.
6 Conclusion
As noted in the introduction, time-to-market is one of the many benefits promised by
contract manufacturers. We find that when a new investment is required, as in the
case of adopting a new process technology, outsourcing does not necessarily result in
22
a time advantage compared to in-house production. In particular, we show that the
time (dis)advantage through outsourcing depends on relative economics at the CM
and the OEM, the scale of the CM, and the evolution of the market potential and
the associated risks.
Apart from time-to-market, several other potential benefits can be achieved through
outsourcing production. First, outsourcing may imply a transfer of risks: In our
model, the risk of excess capacity is transferred to the CM. Second, the CM can
reduce the demand losses due to uncertainty. This is because the CM benefits from
pooling over multiple OEMs, and is therefore in a better position to own capacity:
While the utilization of specialized OEM investment is dependent on the success of
a few products, a CM may allocate the capacity left idle by unsuccessful OEMs to
successful ones. Despite these potential advantages, the risk of lost demand still re-
mains at the OEM. It is possible that the overall risk exceeds that under in-house
production if the right incentives are not provided to the CM and the CM builds
insufficient capacity. A third advantage is that the cost to the OEM of each unit can
be reduced if the wholesale prices are sufficiently low. On the other hand, due to the
other advantages listed above, the OEM may benefit from outsourcing even when the
wholesale price quoted by the CM is higher than its total unit in-house capacity and
production cost.
In the semiconductor industry, anecdotal evidence suggests that there are indeed
cases where outsourcing leads to better time-to-market as defined by faster adoption
of new process technologies. For example, TSMC, the largest foundry (CM), leads the
integrated manufacturers and foundries alike. For the 90 nanometer technology, Intel
is expected lag behind TSMC by one year (Manners 2002, Cataldo 2002). TSMC’s
lead is made possible by an extensive customer base and high margins: TSMC serves
600 fabless semiconductor firms, and its profits in year 2000 were $1.9B over revenues
of $5.3B (Business Week 2001). This is consistent with our results concerning the
impact of CM margin and scale on time-to-market.
We also show that waiting for the CM to make an independent investment into
the process technology may not be advantageous for OEMs. In our model, where we
23
allow for either CM or OEM investment, it is the OEM that makes an investment
in such a setting. In reality, intermediate options are available and various forms of
risk sharing arrangements are observed in industry. While some OEMs, such as Intel
and Conexant, make equity investments into CMs (Cameron 2002), some others form
joint ventures with CMs (for instance AMD and UMC) (Dickie 2002) for investments
into new process technologies. One motivation for such arrangements is to accelerate
the adoption of new technologies by reducing the investment costs, and consequently
the risks for the CMs. In addition, these agreements reduce the delivery risks faced
by the OEMs as they ensure capacity and priority at the CM. Joint ventures are
also formed by OEM alliances, allowing OEMs to pool resources without relying on
contract manufacturers. For example, NEC, Toshiba, Fujitsu, Hitachi and Mitsubishi
Electric have recently announced a joint venture (Dickie 2002). These agreements
allow OEMs to share risks, as well as the fixed costs of new process technologies.
Finally, consolidation in the industry not only mitigates competition, but also allows
risk pooling (Dickie 2002).
Our results have several implications for managers. One reason for outsourcing
in practice is rapid access to leading edge production technology. It is seen that a
contract manufacturer may in fact prefer to adopt a new technology later than an
OEM. This results from inadequate compensation, inefficiency, or insufficient client
base at the CM. These factors should be carefully evaluated before an outsourcing
decision.
Another reason for outsourcing in practice is to achieve cost reduction. Our results
show that cost focus in outsourcing relations may in fact reduce the profit and time-to-
market performance of the OEM. It is very striking that when the time dimension is
taken into account, the OEM may even benefit from paying a wholesale price above in-
house costs. Therefore, when comparing unit costs of delivery between in-house and
outsourced production, the yardstick should not be unit production and investment
costs. These measures do not capture the effects of risk transfer and earlier market
entry.
Finally, our results highlight that the choice of CM is important. One would
24
expect an OEM to obtain better prices from a lower-cost CM, thus achieving cost
efficiency, one of the major drivers of outsourcing. It is a pleasant surprise that this
CM can also provide a better time-to-market: With higher margins and more to gain,
the CM invests early, taking a higher risk than a low-margin CM. The same is true for
a CM with a broad customer base. Benefiting from pooling effects, such a CM faces
a lower risk compared to a smaller CM, and can afford to enter earlier. On the other
hand, a large CM may also be very powerful and appropriate a large fraction of the
profits. Power issues aside, a large CM emerges as the best source of manufacturing
capacity, not only due to efficiency, but also because of better time to market.
An important limitation of our model is that it is based on a single CM serving
multiple OEMs, and does not consider competition between CMs. CMs competing for
capacity may be forced to invest earlier into new technologies to acquire customers or
keep existing ones. Alternatively, some CMs may choose to be “followers” and invest
later. We expect that an extended model capturing competition between CMs would
provide further insights into the time-to-market (dis)advantage of outsourcing; this
is left for future research.
Acknowledgements
We gratefully acknowledge Luk Van Wassenhove for his valuable input. We would
also like to thank Harold Clark, Joe Bellefeuille, Norbert Schmidt, Enrique Salas and
Carlos Nieva from Lucent Technologies for sharing their industry knowledge. This
research was partially funded by the Center for Integrated Manufacturing and Service
Operations and the INSEAD-PwC Initiative on High Performance Organizations.
Appendix
Proof of Lemma 1. The OEM profit is given by
πI(KI , t) = −cIKI+(r−cp) min(D(t), KI) = −cIKI+(r−cp)(D(t)−max(D(t)−KI , 0)),
where Dt ∼ N(S(t), σ(t)2). The OEM maximizes its expected profit over entry time
25
and capacity:
maxKI ,t
πI(KI , t) = maxKI ,t
{−cIKI + S(t)(r − cp) + (r − cp)
∫ ∞
KI
(D(t) − KI)dF (D(t))}.
For a fixed t, πI(KI , t) is strictly concave in KI . The unique optimal capacity at time
t is therefore found by solving the first order conditions and is K∗I (t) = S(t) + zoσ(t).
Substituting the optimal capacity for each time t, the OEM expected profit as a
function of the entry time t can be written as
πI(K∗I (t), t) = (r − cp − cI)[S(t) − ξ(zo)σ(t)], (3)
where ξ(zo) = φ(zo)Φ(zo)
.
Proof of Proposition 1. Lemma 1 reduces the problem of finding the optimal
t to the following:
maxt≥0
πI(K∗I (t), t) = (r − cp − cI) max
t≥0[S(t) − ξ(zo)σ(t)].
The proof proceeds by examining the two cases where there is no point satisfying
the first order condition π′ = 0 and where there is at least one such point. In the
second case, we examine in detail the subcases where π′ has one or two roots. The
results are collected under the cases in Proposition 1.
Case 1. There exists no t1 ∈ [0, TD] s.t. π′I(t1) = 0: In this case, the profit is
either monotonically increasing (π′I < 0) or decreasing in t (π′
I > 0).
(a) If π′I(t) < 0 ∀t and πI(0) > 0 then t∗I = 0. If πI(0) < 0, then t∗I = ∞.
(b) S ′ < 0 and S(TD) = 0. Therefore, if π′I > 0, then πI > 0 is not possible for any
t. Therefore, πI < 0 for all t ∈ [0, TD) and t∗I = ∞.
Case 2. There exists a t1 ∈ [0, TD] s.t. π′I(t1) = 0:
(a) S ′′′|π′′
I=0 > ξσ′′′|π′′
I=0. π′
I is convex and has at most two roots. Since the case
with no roots is proven above, only one and two root cases are examined below:
26
(i) π′I = 0 has one root t1. If π′
I > 0 for t ∈ [0, t1] and π′I < 0 t ∈ [t1, TD] then
there exists a unique maximum at t∗I = t1. If π′I < 0 for t ∈ [0, t1] and π′
I > 0
t ∈ [t1, TD] then there exists a unique minimum at t1. In the latter case, the
entry is at t∗I = 0 if πI(0) > 0, and t∗I = ∞ if πI(0) < 0 since πI(t) first decreases
and then increases up to πI(TD) = 0.
(ii) π′I = 0 has two roots t1 < t2. π′
I > 0 for t < t1, π′I < 0 for t1 < t < t2 and
π′I > 0 for t > t2. Therefore, πI(t1) is a maximum and πI(t2) is a minimum.
t∗I = t1.
(b) S ′′′|π′′
I=0 < ξσ′′′|π′′
I=0. π′
I is concave and has at most two roots. Again, only one
and two root cases are considered below:
(i) π′I = 0 has one root t1. If π′
I > 0 for t ∈ [0, t1] and π′I < 0 t ∈ [t1, TD] then
there exists a unique maximum at t∗I = t1. If π′I < 0 for t ∈ [0, t1] and π′
I > 0
t ∈ [t1, TD] then there exists a unique minimum at t1. πI(TD) = 0 and the profit
function has a single inflection point at t1. Therefore, if the profit function has
a minimum at t1, then the non-negative maximum can only be at t = 0. Thus,
t∗I = 0 if πI(t1) > 0 and t∗I = ∞ if πI(t1) < 0.
(ii) π′I = 0 has two roots t2 < t1. π′
I < 0 for t < t2, π′I > 0 for t2 < t < t1 and
π′I < 0 for t > t1. Therefore, πI(t2) is a minimum and πI(t1) is a maximum.
The global maximum is the larger of πI(0) and πI(t1): If πI(0) < πI(t1) <,
t∗I = t1, otherwise t∗I = 0.
(c) More general S and σ may have multiple local minima and maxima, and the
global maximum can be found by line search.
Proof of Lemma 2. Similar to the proof of Lemma 1.
Proof of Lemma 3.
ξ(z) =φ(z)
Φ(z)=
φ(−z)
1 − Φ(−z).
27
dξ(z)
dz=
d
dz
(
φ(−z)
1 − Φ(−z)
)
= − d
dz
(
φ(z)
1 − Φ(z)
)
.
The Normal distribution is an increasing failure rate distribution:
d
dz
(
φ(z)
1 − Φ(z)
)
> 0.
Therefore,dξ(z)
dz= − d
dz
(
φ(z)
1 − Φ(z)
)
< 0.
Proof of Proposition 2. Since t∗ is an interior maximum, it satisfies the first order
condition
S ′(t) − xσ′(t) = 0
and the second order condition
S ′′(t) − xσ′′(t) < 0,
where x = ξ(zo) for the OEM and x = ξ(zc)√M
for the CM.
(i) Define h(t, x).= S ′(t)− xσ′(t). Let t∗(x) denote the optimal time as a function of
the parameter x. Then t∗(x) satisfies
S ′(t∗(x)) − xσ′(t∗(x)) = 0.
Taking the derivative of this equality with respect to x, we obtain
S ′′(t∗(x))dt∗(x)
dx−
(
σ′(t∗(x)) + xσ′′(t∗(x))dt∗(x)
dx
)
= 0
dt∗(x)
dx(S ′′(t∗(x)) − xσ′′(t∗(x))) − σ′(t∗(x)) = 0.
The multiplier of dt∗(x)dx
is negative. Since σ′ is also negative by assumption, we
conclude that dt∗(x)dx
is positive. In our problem, we have x = ξ(zo) for the OEM and
x = ξ(zc)√M
for the CM, so t∗ increases in ξ(zo) and ξ(zc), and decreases in M .
(ii) Consider S1 and S2 such that S ′2(t) > S ′
1(t) ∀t. Let t∗1 and t∗2 be the optimal time
of investment under the two cases, respectively. Then S ′2(t
∗1) − xσ′(t∗1) > S ′
1(t∗1) −
xσ′(t∗1) = 0. Since S(t) − xσ(t) decreases in t, t∗2 > t∗1.
28
(iii) With a parallel argument we conclude that t∗2 < t∗1.
By assumption, S ′(t) < 0 and σ′(t) < 0. Hence, we can find a small ε s.t. S ′(t) <
εσ′(t)∀t. When x = ε, π′ < 0 ∀t, and t∗ = 0 if π(0) > 0.
Proof of Proposition 3. (i) From Proposition 2, there exists a threshold ξ0 for
ξ(·), below which t∗ = 0. If ξ(zo) < ξI0 and ξ(zc) < ξII
0 , then t∗o = t∗c = 0. (ii) The
optimal adoption times for the OEM and the CM are, respectively,
t∗I = arg maxt≥0
mo(S(t) − ξ(zo)σ(t))
t∗II = arg maxt≥0
mc(S(t) − ξ(zc)√M
σ(t))
For identical OEMs, the difference in timing is driven by the difference in the multi-
plier of the standard deviation in these expressions. Recall from the proof of Propo-
sition 2 that t∗ increases in x. Therefore, if ξ(zo) < ξ(zc)√M
, then t∗I < t∗II .
References
BASS, F. M. (1969), “A New Product Growth Model for Consumer Durables,” Man-
agement Science, 15, 1 215-227.
BAYUS, B. L. (1997), “Speed-to-market and New Product Performance Trade-offs,”.
Journal of Product Innovation Management, 14, 6, 485-497.
BHATTACHARYA, S., V. KRISHNAN, AND V. MAHAJAN (1998), “Managing
New Product Definition in Highly Dynamic Environments,” Management Science,
44, 11, S50-S64.
BOWMAN, D., AND H. GATIGNON (1996), “Order of Entry as a Moderator of the
Effect of the Marketing Mix on Market Share,” Marketing Science, 15, 3, 222-242.
BUSINESS WEEK, “Betting Big on Chips,” April 30 2001.
29
COHEN, M. A., J. ELIASHBERG, and T. H. HO (1996), “New Product Develop-
ment: The Performance and Time-to-Market Trade-off,” Management Science, 42, 2,
173-186.
CACHON, G. P. AND P. HARKER (2002), “Competition and Outsourcing with
Scale Economies,” Management Science, 48, 10, 1314-1333.
CAMERON, I. (2002), “Dubai Looks to be a Fab Player,” Electronic Engineering
Times, April 15, 2002.
CATALDO, A. (2002), “Time to put the brakes on fast process shifts, TSMC says,”
Electronic Engineering Times, April 15, 2002.
CHATTERJEE, R. AND Y. SUGITA (1990), “New Product Introduction under De-
mand Uncertainty in Competitive Industries,” Managerial and Decision Economics,
11, 1, 1-12. DICKIE, M. (2002), “Good Judgement and Deep Pockets Needed,”
Financial Times, April 17 2002.
EISENDHARDT, K. M. AND B. N. TABRIZI (1995), “Accelerating Adaptive Pro-
cesses: Product Innovation in the Global Computer Industry,” Administrative Science
Quarterly, 40, 84-110.
ELECTRONIC NEWS (2001), “Size Matters-in MPU Manufacturing,”, November
19, 2001.
GRIFFIN, A. AND J. H. Hauser (1993), “The Voice of The Customer,” Marketing
Science, 12, 1, 1-27.
HOPPE, H. C (2000), “Second-Mover Advantages in the Strategic Adoption of New
Technology under Uncertainty,” International Journal of Industrial Organization, 18,
315-338.
30
JENSEN, R. (1982), “Adoption and Diffusion of an Innovation of Uncertain Prof-
itability,” Journal of Economic Theory, 27, 182-19.
KAMIEN, M. I. AND N. L. Shwartz (1982), Market Structure and Innovation, Cam-
bridge University Press, New York, NY.
KERIN, R. A., P. R. VARADARAJAN, AND R. A. PETERSON (1992), “First
Mover Advantage: A synthesis, Conceptual Framework, and Research Propositions,”
Journal of Marketing, 56, 33-52.
LIEBERMAN, M. B. AND D. B. MONTGOMERY (1988), “First-Mover Advan-
tages,” Strategic Management Journal, 9, 41-58.
LIEBERMAN, M. B. AND D. B. MONTGOMERY (1998), “First-Mover (Dis)Advantages:
Retrospective and Link with the Resource-based View,” Strategic Management Jour-
nal, 19, 1111-1125. LEACHMAN, R. C., J. Plummer, AND N. Sato-Misawa (1999),
“Understanding Fab Economics,” CSM Working Paper, U.C. Berkeley.
LILIEN, G. L. AND E. YOON (1990), “The Timing of Competitive Market Entry:
An Exploratory Study of New Industrial Products,” Management Science, 36, 5,
568-585.
MACCORMACK, A., R. VERGANTI, AND M. IANSITI (2001), “Developing Prod-
ucts on ‘Internet Time’: The Anatomy of a Flexible Product Development Process,”
Management Science, 47, 1, 133-150.
MACHER, J. (2001), “Vertical Disintegration and Process Innovation in Semicon-
ductor Manufacturing: Foundries vs. Integrated Device Manufacturers,” Working
Paper, Georgetown University.
MANNERS, D. (2002), “Foundries Get Upper Hand,” Electronics Weekly, February
13, 2002.
31
MOORE, M. J., BOULDING, W., AND GOODSTEIN, R. C. (1991), “Pioneering
and Market Share: Is Entry Time Endogeneous and Does It Matter,” Journal of
Marketing Research, 28, 97-104.
MCCARDLE, K. (1985), “Information Acquisition and the Adoption of New Tech-
nology,” Management Science, 31, 11, 1372-1389.
MAMER, J. W. AND K. MCCARDLE (1987), “Uncertainty, Competition and the
Adoption of New Technology,” Management Science, 33, 2, 161-177.
MAHAJAN, V. and MULLER, E. (1996), “Timing, Diffusion, and Substitution of
Successive Generations of Technological Innovations: The IBM Mainframe Case,”
Technological Forecasting and Social Change, 51, 109-132.
MORGAN, L. O., R. M. MORGAN, AND W. L. MOORE (2001), “Quality and Time-
to-Market Trade-offs When There Are Multiple Product Generations,” Manufacturing
& Service Operations Management, 3, 2, 89-104.
NETESSINE, S. AND N. RUDI (2001), “Supply Chain Choice on the Internet,”
Working Paper, University of Rochester, Simon Graduate School of Business Admin-
istration.
PLAMBECK, E. L. and T. A. TAYLOR (2001), “Sell the Plant? The Impact of
Contract Manufacturing on Innovation, Capacity and Profitability,” Working Paper,
Stanford Graduate School of Business.
REINGANUM, J. F. (1989), “The Timing of Innovation: Research, Development
and Diffusion” in Handbook of Industrial Organization, vol.1, R. Schmalensee, R. D.
Willig (eds.), North Holland, Amsterdam.
ROBINSON, W. T. and S. MIN (2002), “Is the First to Market First to Fail? Empir-
ical Evidence for Industrial Goods Businesses,” Journal of Marketing Research, 39,
120-128.
32
SMITH, P. G. and D. G. REINERTSEN (1991), Developing New Products in Half
the Time, Van Nostrand Reinhold Books, New York, NY.
SOLECTRON WEBSITE http://www.solectron.com/about/index.html. Accessed:
5/15/2002.
TELLIS, G. J. and P. N. GOLDER (1996), “First to market, First to fail? Real
Causes of Enduring Market Leadership,” Sloan Management Review, 41, 4, 65-75.
TIROLE, J. (1988) The Theory of Industrial Organization, MIT Press, Cambridge,
MA.
VAN MIEGHEM, J. (1999), “Coordinating Investment, Production and Subcontract-
ing,” Management Science, 45, 7, 954-971.
WARD, A., J. K. LIKER, J. J. CRISTIANO, AND D. K. SOBEK II (1995), “The
Second Toyota Paradox: How Delaying Decisions Can Make Better Cars Faster,”
Sloan Management Review, 36, 3, 43-61.
33