Professional Documents
Culture Documents
Abstract
The issue of vendor choice has traditionally not been considered in determining the optimal set of products to produce.
That is, component purchase has typically been considered only at the MRP explosion stage, well after the set of products
to produce and their production quantities has been decided. This decoupling of roles may sometimes result in high
acquisition cost of components. To control costs, it is crucial that unit prices of component variants, fixed costs of doing
business with vendors, and their capabilities to supply different component variants be assessed before deciding which set
of models to include in a product line. In our analysis, we have explicitly provided for this coupling between product choice
and vendor selection, and have gone a step further by bringing in customers to determine the sales potential of the different
models of a product that may be produced. We use our analysis to study the following issues: (1) how can a manufacturer
decide which product models to offer and which vendors to select to offer components for these models, (2) under what
situations are general vendors (who offer a larger variety of component variants) preferred to more focused vendors, (3)
how much discount do focused vendors need to offer to compensate for their lack of versatility, (4) how does the fixed cost
of vendor relationships influence the vendor selection decision, (5) how does the intensity of market competitiveness
influence this decision, (6) how can a manufacturer work with a general vendor to exploit new market opportunities, and
(7) how much incentive can a manufacturer offer focused vendors to get them to become more versatile in their
capabilities. For each problem scenario, the output of our procedure tells us the set of models that should be chosen for a
product line, production quantities of each model, specific vendors from whom component variants needed to produce
these models should be procured, and profits to the manufacturer and each vendor selected.
r 2007 Elsevier B.V. All rights reserved.
0925-5273/$ - see front matter r 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.ijpe.2005.11.011
ARTICLE IN PRESS
724 N. Balakrishnan, A.K. Chakravarty / Int. J. Production Economics 112 (2008) 723–741
can a manufacturer offer a focused vendor to get 2. Product demands and component costs
him to become more versatile in his capabilities.
In our analysis, as in Balakrishnan and Chakra- The commercial success of a model of a product
varty (1996), we consider the problem of a market (referred to hereafter simply as model) depends on
leading OEM and base our analysis on life cycle how much of it can be sold at a given price, and at
demands and costs. Specifically, we study the what cost it can be produced. We therefore discuss
problem from the OEM’s perspective in terms of demand- and cost-related issues that may impact the
increasing their profit. However, we show how an product choice and vendor selection decisions.
increase in the OEM’s profit may be accompanied
by a concurrent increase in the profit for certain 2.1. Products and vendors
vendors. We also show how the commonality of
components used in product choice leads to using Let N denote the number of outsourced compo-
only a few of the possible vendors. Price discounts nents used in manufacturing the product. Each
by vendors (based on volume of sales) could component may have several variants depending on
obviously limit the chosen vendor set even further. the levels at which that component’s attributes are
However, since our primary focus is to study the set. The selection of 1 variant for each component
impact of component commonality on product and would, together, create a model denoted by i (iAI,
vendor choice, we do not include price discount where I is the set of all models). Each model i will be
effects in our analysis. The inclusion of price an N-vector corresponding to specific component
discounts adds an additional layer of complexity variant choices. For example, the product (food
to the model and we plan to address it in a future processor) in the Sunbeam case has 4 components
research project. (i.e., N ¼ 4). The motor assembly component has 27
This paper is organized as follows. In Section 2, possible variants (three attributes at three choices
we discuss demand and cost issues that could affect each), while the bowl assembly, feed-tube assembly,
the choice of specific models to produce and the and pouring spout components have 12, 4, and 2
selection of vendors to supply outsourced compo- possible variants, respectively. The total number of
nent variants for these models. Based on this models for Sunbeam is therefore 2592 ( ¼ 27
discussion, we derive an expression for calculating 12 4 1), although, in practice, this number will
the OEM’s profit. In Section 3, we describe an be smaller since some combinations of component
efficient procedure adapted from Chakravarty and variants may not be compatible with each other.
Balakrishnan (2001) that may be used to identify It is clear that the same component variant could
the optimal set of models to offer and the vendors be present in several models. For example, as shown
who should be used to facilitate their production. in Table 2, in the 5 models selected by Sunbeam for
In Section 4, we describe computational tests detailed in-depth analysis (Page and Rosenbaum,
designed to answer the research questions posed 1987), the same motor assembly variant (heavy duty
earlier, and discuss managerial implications of the capacity, 5 blades, 7 speed) is used in both models 3
results. We conclude this paper in Section 5 with a and 4.
summary of our findings and statements of our It is reasonable to expect that all possible variants
contributions. of a component may not be available from a given
The primary contributions of this paper are the vendor. Thus, if we choose to produce all 5 models
mathematical model that has been developed for in Table 2, the set of vendors chosen must, together,
this comprehensive problem and more importantly, possess the capability to supply all the component
the various managerial insights derived from the variants required. If there are (say) 4 vendors, we
computational results obtained by the efficient may choose to do business with any number (p4) of
solution of this model for different problem them, depending on their combined capabilities,
environments and parameters. These insights into costs, and models chosen.
how an OEM can work with different types of Gupta and Krishnan (1999) point out that for
vendors (general and focused) in different ways so every vendor who is chosen, a certain vendor-
that they can jointly exploit the marketplace are related fixed cost is incurred. These costs comprise
especially interesting, and could be very useful to issues such as vendor certification, contract agree-
other researchers working on supply chain design ment, and progress monitoring costs, and are
issues. independent of the volume and type of component
ARTICLE IN PRESS
726 N. Balakrishnan, A.K. Chakravarty / Int. J. Production Economics 112 (2008) 723–741
Table 2
Models selected by Sunbeam
Motor assembly Motor capacity Professional Regular Heavy duty Heavy duty Heavy duty
Number of blades 5 3 5 5 5
Number of speeds 1 1 7 7 2
Bowl assembly Bowl size 4 quarts 2.5 quarts 1.5 quarts 1.5 quarts 2.5 quarts
Bowl type Regular Regular Regular Side discharge Regular
Bowl shape Cylindrical Cylindrical Cylindrical Cylindrical Cylindrical
Feed-tube assembly Feed-tube type Large Regular Large Large Large
Feed-tube pusher Interchangeable Regular Interchangeable Interchangeable Regular
Pouring spout Absent Present Absent Absent Absent
variants supplied. Thus, the buyer would like to OEM, the total sales revenue may be expressed as
minimize the number of vendors used. Clearly, if the X
number of vendors chosen is small, these vendors Revenue ¼ Pi Qi , (4)
i2I^
have to be more general (i.e., capable of supplying
several different variants of several components). where Qi is the quantity of model i sold. It is clear
that
2.2. Product demand and revenue Qi pDi 8i 2 I. (5)
variant can be bought may be written as Theorem 1. The value of Qi that maximizes profit
^ U; Þ in expression (7), will be equal to Di if
PðI;
C jk ¼ Minimumfv2Ug C jkv . (6) PiXCiU, and equal to zero if PioCiU.
As noted earlier, we do not consider the impact of In our current model, we assume that the OEM
price discounts based on volume of sales in our does not have a total capacity constraint with
current analysis. Therefore, by adding up the costs respect to all its models. As such, we can substitute
C jk for all component variants in model i, the lowest for Qi with Di in our model. Here again, while this
cost of acquiring the purchased components of that assumption may be rather restrictive in some
model using set U of vendors may be easily situations, it could be quite logical in other
determined. Let this cost be denoted by CiU. Then, situations depending on the availability of out-
for a given set I^ of models and set U of vendors sourcing capabilities for the OEM.
from whom to acquire component variants for these Substituting for Qi with Di in expressions (3) and
models, the total profit may be expressed as (7), we have
X X X ," #
PðI; ^ UÞ ¼ Pi Qi C iU Qi dv , (7) X X X
^ UÞ ¼
PðI; ðPi C iU ÞORi RC þ Ri dv .
i2I^ i2I^ v2U
i2I^ i2I^ v2U
To develop out search procedure, we make use of points. We establish this property, called the
Theorems 2 and 3 below. Letting, consecutiveness property, in Theorem 3.
xiU ¼ ðPi C iU ÞORi , (11) ^ I^ IÞ
Theorem 3. For a given vendor set U, the set Ið
we rewrite expression (9) as corresponds to an extreme point eAE if for any i 2 I^
^ xiU =Ri 4xjU =Rj .
and jeI,
," #
Y X X X
^ UÞ ¼
ðI; xiU RC þ Ri dv , Theorem 3 is extremely useful in our context, as it
i2I^ i2I v2U allows us to rank order the models i (iAI) in
(12) decreasing order of xiU/Ri values. Items from the
^ so as long
top end of this list will be placed in set I,
which, in turn, can be rewritten as as profit PðI; ^ UÞ as in expression (13) keeps
Y X
^ UÞ ¼ X ^ ½RC þ Y ^
ðI; dv , (13) increasing. For optimality,
IU I Y Y
P P
v2U
ðI^ þ j; UÞp ðI; ^ UÞ and xiU =Ri 4xjU =Rj
where XIˆU ¼ iAIˆ xiU and YIˆ ¼ iAIˆ Ri. Thus, the
8i 2 I^ and jeI.
^ ð17Þ
problem reduces to:
For model i in the optimal set I, ^ purchase
Program P1. quantities will be determined as
Y X
Maximize ^ UÞ ¼ X ^ =½RC þ Y ^
ðI; dv ,2 3
IU I X
v2U Qi ¼ ORi 4RC þ Rj 5 8i 2 I.
^ (18)
(14) j2I^
possible to improve the solution by swapping model problems of large size. However, in all our extensive
i in set I^ with some model j not in set I.
^ Such a tests even with problems where there were multiple
swapping procedure, in general, will be combina- vendors, we were unable to find a single instance
torial in nature. Using additional dominance where Procedure P yielded a profit that was more
properties, however, we show how the swapping than 0.1% away from the optimal profit (found
procedure can be simplified, at least for one-to-one using the completely enumerative procedure). This
swaps. This property is stated in Theorem 4, and the seems to indicate that our procedure will yield near-
proof of this theorem is given in Appendix A. optimal solutions for problems of practical size.
In what follows, we describe the results of several
Theorem 4. For a given vendor set U, for any models computational tests conducted to study the 7
i 2 I^ and jeI,
^ PðI^ þ j i; UÞ4PðI;
^ UÞ if and only if
research questions stated earlier. These are: (1)
xjUXxiU and RjXRi. how does the choice of product models the
manufacturer should offer in the market influence
Using the analysis described above, the procedure
the vendor choice decision, and vice versa, (2) under
to identify the set I^ of models that should be
what circumstances would general vendors be
selected by the OEM may be summarized as
preferred to more focused vendors, (3) how much
follows:
discount do focused vendors need to offer to
Procedure P. compensate for their lack of versatility, (4) how
does the fixed cost of vendor relationships influence
Step 1. Select current set U(CV) of vendors. the product line and vendor choice decisions, (5)
Compute CiU for each iAI. how does market competition influence decisions
Step 2. Compute xiU for each iAI, as defined in related to product choice and vendor selection, (6)
expression (11). how can a manufacturer induce a general vendor to
Step 3. Identify and sort the top f (pre-specified) offer a discount so that, together, the manufacturer
models in descending order of xiU/Ri. Set and the vendor can exploit new market opportu-
r ¼ 1. nities, and (7) how much incentive can a manufac-
Step 4. Let X rIU r turer offer focused vendors to get them to become
^ ¼ Si¼1 to r xiU and Y I^ ¼ Si¼1 to r Ri .
r
Determine profit PðI^ UÞ ¼ X rIU r more versatile in their capabilities? For each
^ =½RC þ Y I^ Þ.
r r1 r1 question that is studied, we also point out the
Step 5. If PðI^ UÞoPðI^ UÞ, I^ is optimal. Stop.
managerial implications of our findings.
Else, go to Step 6.
r r1
Step 6. If PðI^ UÞXPðI^ UÞ, Set r’r+1. If,
4. Computational results
rpf, go to Step 4. Else, go to Step 7.
f
Step 7. Swapping procedure. Identify i 2 I^ and
f
Procedure P is coded in Visual Fortran and
jeI^ such that xjUXxiU and RjXRi. If no implemented on a PC. We first describe the
f
such (i,j) pair exists, or if PðI^ þ j procedure used to randomly generate the data for
f f a given problem scenario.
i; UÞoPðI^ UÞ for all such (i,j) pairs, I^ is
optimal. Stop. Else, go to Step 8.
f 4.1. Generation of problem data
Step 8. Find the (i,j) pair that causes ½PðI^ þ j
f
i; UÞ PðI^ UÞ to be the maximum. Set In all our tests, we assume the product consists of
f f
I^ I^ þ j i. Go to Step 7. 5 components (i.e., N ¼ 5), all of which are
purchased from vendors. For each component, we
Since the swapping procedure (Step 7) considers first randomly generate its number of variants
only one-to-one swaps in the improvement stage of (discrete uniform distributed between 4 and 7).
the procedure, it does not guarantee that the Next, for each component variant, we randomly
solution obtained is a global optimum. Therefore, generate its base cost (discrete uniform distributed
to test the level of sub-optimality of our approach, between $10 and $30). This base cost is the unit
we also developed a completely enumerative proce- purchase cost paid by the OEM to a general vendor
dure that examined all possible combinations of capable of supplying that item. A focused vendor
models for the set I. ^ Clearly, such an enumerative capable of supplying that item would discount this
procedure becomes computationally intractable for base cost by a specified factor. To compute vendor
ARTICLE IN PRESS
730 N. Balakrishnan, A.K. Chakravarty / Int. J. Production Economics 112 (2008) 723–741
profits, we set the unit manufacturing cost (for the vendor for a component variant to compensate
vendor) of a component variant at 80% of its base for the lack of versatility in his capabilities, when
cost. compared to a general vendor. In our tests, we set
Using the list of variants available for each this factor at levels of 5%, 10%, and 15%.
component, we create set I of all possible product Clearly, as the discount rate increases, we expect
models i (note that each model i is an 5-element focused vendors to be able to easily compensate
vector). In the extreme case where each component for their lack of versatility, and become more
variant is fully compatible with every variant of all competitive.
other components, the number of possible models is Competition ratio (b). The second factor specifies
simply the product of the number of variants the level of market competition for the product
available for each component. However, in most models offered by the OEM. We define the
practical situations, the actual number would be competition ratio as the ratio of RC, the total
much less since several component variants may be attractiveness factor of competitors’ potential
incompatible with other variants. Without loss of products, to Si2I^ Ri , the total attractiveness factor
generality, we assume full compatibility of all of all the models the OEM offers for sale (i.e.,
component variants in our tests. b ¼ RC =Si2I^ Ri ). Note that a competition ratio of
Using this set of product models I, for each model 0.5 implies that total model rating of the
i, we randomly generate its rating Ri (uniform competition is only 50% that of the OEM. That
0.7–1.0 of the total base cost of the component is, the OEM dominates the market. In contrast, if
variants in the model) and selling price Pi (uniform the ratio is 1.5, the OEM’s competitors are
1.05–1.30 of the total base cost). significantly more dominant in the market than
Finally, we specify the capabilities of each the OEM. In our tests, we set this factor at levels
vendor. We assume that there are 7 vendors, of of 0.5, 1.0, and 1.5.
whom 3 are general vendors and the remaining 4 are Number of product models (f). The third factor is
focused vendors. For each vendor, we randomly the number of product models that the OEM
generate the actual number of component variants plans to offer. As noted earlier, we specify this
offered, and the actual identity of these variants. To parameter in our analysis to ensure that the
do so, we assume in our tests that each general resulting solution is practical and can be im-
vendor is capable of supplying between 70% and plemented by the OEM. The objective of the
90% of the total set of component variants procedure would be to then optimally select this
(randomly identified), and that each focused vendor specified number of models from the entire set I
is capable of supplying between 20% and 40% of of models. Clearly, as the number of models to be
this set (also randomly identified). selected increases, the number of component
In our analysis, we do not permit price competi- variants needed (and hence, the vendors neces-
tion between vendors of the same type. That is, for sary to supply these variants) would also
example, we assume that all general vendors capable increase. In our tests, we set this factor at levels
of supplying a specific component variant charge of 10, 15, and 20
the same price. Hence, vendors of the same type Vendor fixed cost (d). The final factor is the fixed
compete against each other only based on the cost of developing and maintaining relationships
variety of their offerings. Likewise, we assume that with vendors. As this cost increases, we would
the fixed costs of vendor relationships are the same expect the OEM to prefer working with general
for all vendors (i.e., dv ¼ d for all vendors v). vendors in order to keep the vendor set down. In
our tests, we set this factor at levels of $50,000,
4.2. Selection of model factors, and their $100,000, and $150,000. As noted earlier, we
experimental values assume that this fixed cost is the same for all
vendors. That is, the vendors do not compete
To study the problem under different problem against each other based on differentials in the
scenarios, we identify 4 experimental factors as fixed cost of vendor relationships.
follows:
Since there are 4 factors with 3 levels each, the
Discount factor (a). The first factor is the total number of experimental combinations is 81
purchase price discount offered by a focused ( ¼ 3 3 3 3). In the remainder of this section,
ARTICLE IN PRESS
N. Balakrishnan, A.K. Chakravarty / Int. J. Production Economics 112 (2008) 723–741 731
we first describe our computational tests to examine Although we compute several output measures
question nos. 1–5 posed earlier. To determine the using our procedure (as presented in Table 4), we
statistical significance of these results, we also focus our discussion here on the following key
analyze the experimental results using the ANOVA performance measures: (1) OEM’s profit, (2) total
procedure. Finally, we describe additional results number of general vendors selected, (3) total
conducted on a specific problem environment to number of focused vendors selected, (4) total profit
study question nos. 6 and 7. for all general vendors selected, and (5) total profit
for all focused vendors selected.
4.3. Product model choice and vendor selection To verify the statistical significance of the output
measures for the 4 experimental factors, we first
For each of the 81 combinations of factor values analyze the computational results using the ANO-
studied, 20 replications (each with a different VA procedure. Prior to interpreting the ANOVA
random number seed) are performed. The total results, the necessary diagnostic checks are per-
number of replications is therefore 1620. For each formed using standard graphical analysis of the
replication, Procedure P is used to select the residuals to verify homogeneity of error variances.
specified number (f) of product models for inclu- The residual plots are found satisfactory.
sion in the optimal set. The ANOVA results are summarized in Table 5.
Selected details of a sample problem are shown in For each of the 4 main factors and the 6 two-way
Table 3. In this example, there are a total of 3360 interactions, the table shows the F-ratios and the
product models possible, of which the procedure corresponding p-values for the 5 output measures
has to select f models. For each model, the table listed earlier.
shows its rating, total base cost (which is the sum of Examination of the ANOVA results indicates
base costs of all component variants in the model), that with the exception of 1 main effect, all of the
and selling price. For each component variant, the main effects involving the 4 factors are statistically
table shows the purchase cost from each vendor significant at the 7% level. In fact, most are
capable of supplying that item. A blank entry significant even at the 1% level. With the exception
implies the vendor does not have the capability to of some 2-way interactions involving the factor a,
supply that item. Note that focused vendors (vendor other 2-way interactions are not significant for all
nos. 1–4) discount the unit cost of a variant by a the output measures indicating that interpretation
factor a, when compared to general vendors (vendor of the main effects is valid.
nos. 5–7). The only non-significant main effect in the
A typical output (for a given set of factor levels) is ANOVA table is the total profit of general vendors
shown in Table 4. The output lists the factor levels with regards to factor b. To explore the reason why
for the current problem, the OEM’s profit (after this is the case, we present (in Table 6) the least-
paying the total fixed cost of vendor relationships), squared means for all the output measures for each
the vendors (focused and general) selected and their level of each experimental factor.
respective profits, the product models that are to be The table reveals that unlike other output
included in set I, ^ and the specific vendors from measures that monotonically either increase or
whom the component variants required for each decrease as the factor level increases, the total profit
model should be sourced. For each product model, of general vendor first increases as b increases from
the output also includes its production quantity, as 0.5 to 1.0, and then decreases as b increases further
well as the sales revenue and total purchase cost of from 1.0 to 1.5. A possible explanation for this
all component variants in that model. For example, phenomenon is as follows. As we will illustrate
the first model selected includes variants 4, 2, 2, 3, shortly, when the market becomes more competitive
and 5 of components 1, 2, 3, 4, and 5, respectively. there is a shift away from focused vendors toward
The variants required for components 2, 4, and 5 the general vendors. Hence, as b increases from 0.5
are supplied by vendor no. 1, and the variants to 1.0, the general vendors see an increased level of
required for the other 2 components are supplied by business and hence, increased profits. However, as b
vendor no. 4. The total sales quantity of this model increases further from 1.0 to 1.5, the OEM’s market
is 6399 units, yielding a total revenue of $844,720 share suffers even more. Under this situation, even
for the OEM (unit price is $132). The total purchase with the increased level of business, the general
cost is $645,059. vendors experience a decline in their total profit.
ARTICLE IN PRESS
732 N. Balakrishnan, A.K. Chakravarty / Int. J. Production Economics 112 (2008) 723–741
Table 3
Sample problem data
C1 C2 C3 C4 C5
1 91 97 102 1 1 1 1 1
2 77 88 93 1 1 1 1 2
3 81 82 89 1 1 1 1 3
^
3,358 73 86 96 4 6 4 5 5
3,359 74 84 94 4 6 4 5 6
3,360 70 73 80 4 6 4 5 7
Component Variant Unit purchase cost from vendors (blank implies not available)
1 1 28.0 28.0
1 2 9.5 9.5 10.0 10.0
1 3 21.9 23.0 23.0
1 4 26.6 28.0 28.0 28.0
2 1 16.1 17.0 17.0 17.0
2 2 20.9 20.9 22.0 22.0 22.0
2 3 19.0 19.0
2 4 29.0 29.0
2 5 17.1 18.0 18.0 18.0
2 6 12.4 12.4 13.0 13.0
3 1 11.4 11.4 12.0 12.0 12.0
3 2 18.0 18.0 19.0 19.0 19.0
3 3 10.0 10.0 10.0
3 4 9.5 9.5 10.0 10.0 10.0
4 1 13.0 13.0 13.0
4 2 23.0 23.0
4 3 17.1 17.1 17.1 18.0
4 4 29.0 29.0 29.0
4 5 9.5 10.0
5 1 27.0 27.0 27.0
5 2 17.1 17.1 18.0 18.0 18.0
5 3 11.4 12.0
5 4 19.0 19.0 20.0
5 5 23.8 25.0 25.0 25.0
5 6 21.9 23.0
5 7 11.4 12.0 12.0 12.0
Detailed examination of the F-ratios in Table 5 measure. While the effect of the vendor fixed costs
indicates the following. The OEM’s profit is affected (measured by the factor d) is also dominant for the
most by the competitive conditions in the market number of focused vendors, it has a much smaller
(measured by the factor b), followed by the discount effect on their profits.
a offered by focused vendors. With regard to all The preceding discussion indicates that with
other measures, the discount a is the most influential regard to research question (1) listed earlier, all 4
ARTICLE IN PRESS
N. Balakrishnan, A.K. Chakravarty / Int. J. Production Economics 112 (2008) 723–741 733
Table 4
Solution for a given problem environment
C1 C2 C3 C4 C5
Table 5
ANOVA results
Factor OEM’s profit No. of focused No. of general Focused vendors’ General vendors’
vendors vendors profit profit
F p F p F p F p F p
a 1610.53 0.0001 456.03 0.0001 190.87 0.0001 605.43 0.0001 433.27 0.0001
b 3296.11 0.0001 50.17 0.0001 2.79 0.0617 412.01 0.0001 0.07 0.9308
f 978.73 0.0001 19.00 0.0001 2.89 0.0561 112.33 0.0001 3.25 0.0391
d 307.85 0.0001 349.55 0.0001 27.84 0.0001 49.67 0.0001 29.94 0.0001
ab 36.16 0.0001 2.14 0.0736 0.56 0.6892 35.61 0.0001 0.17 0.9540
af 9.86 0.0001 0.48 0.7494 0.67 0.6115 9.64 0.0001 0.12 0.9766
ad 1.56 0.1814 2.15 0.0724 3.38 0.0092 30.52 0.0001 26.08 0.0001
bf 4.05 0.0029 0.61 0.6545 0.13 0.9714 0.39 0.8175 0.07 0.9919
bd 0.43 0.7848 0.32 0.8620 0.35 0.8430 0.55 0.7020 0.60 0.6595
fd 0.36 0.8343 0.19 0.9455 0.09 0.9848 0.07 0.9919 0.05 0.9949
experimental factors seem to impact the choice of of these factors, the tables show the average (based
product models and the vendor choice decision. To on 20 observations) for the 5 output measures.
address research questions (2)–(5), we now examine Examination of the means reveals several inter-
the main factor for each output measure for all esting observations. In response to question (3) that
combinations of factor levels. These results are focuses on the discount offered by focused vendors,
summarized in Table 7(a) (for d ¼ $50,000), Table we note that as the discount increases (all other
7(b) (for d ¼ $100,000), and Table 7(c) (for factors held constant), the OEM’s profit increases
d ¼ $150,000), for all combinations of the other 3 appreciably. This increase in profit is accompanied
factors (namely, a, b, and f). For each combination by an increase in the number of focused vendors
ARTICLE IN PRESS
734 N. Balakrishnan, A.K. Chakravarty / Int. J. Production Economics 112 (2008) 723–741
Table 6
Least-squared mean values
Factor Level OEM’s profit No. of focused No. of general Focused vendors’ General vendors’
vendors vendors profit profit
selected, and a corresponding decrease in the nent variants decrease by a larger amount (in
number of general vendors selected. However, the absolute terms) when compared to cheaper variants.
total profit for both types of vendors typically Hence, product models that include these expensive
decreases. While the total profit for focused vendors component variants (which are unprofitable at
decreases by around 50% (as a increases from 5% current prices) now become attractive. This is an
to 15%), the magnitude of the decrease in profits for important insight since it clearly links vendor choice
general vendors drops in excess of 70% in most to the product mix offered by the OEM.
cases. As noted earlier, while the OEM’s profit increases
Interestingly, the tables also reveal that as the as a increases, the focused vendors’ profit typically
discount factor a increases, the OEM may find it decreases. For example, for the factor levels
optimal to change the mix of product models itself. considered in the preceding discussion, while the
For example, consider the solution values in OEM’s profit increases by $327,187 as a increases
Table 7(a) when a ¼ 5%, b ¼ 0.5, f ¼ 10, and from 5% to 10%, the focused vendors see their
d ¼ $50,000. The profits in this case for the OEM, profit decrease by $202,677. In many cases, the
focused vendors, and general vendors are OEM may not have sufficient influence to demand
$1,583,766, $890,850, and $344,944, respectively. this additional discount from these vendors. In such
Now assume the focused vendors increase their cases, it may even be possible for the OEM to offer
discount from 5% to 10% (with no change from the to compensate these vendors for their losses from
general vendors). If the OEM chooses to keep the the OEM’s increased profit. That is, the OEM and
same mix of product models, their profit would focused vendors can cooperate with each other and
increase by 5% of the focused vendors’ total jointly exploit the market conditions to reap
revenue (which is $5,642,051; not shown in Table increased profit for both parties. This increase in
7(a)). Hence, the OEM’s new profit would be both their profits would typically be at the expense
$1,583,766+0.05 $5,642,051 ¼ $1,865,868. How- of the general vendors. In such cases, it is of course
ever, we note that the OEM is actually able to likely that the general vendors would also offer
realize a profit of $1,910,953 in this scenario, as some discounts to retain their share of business.
shown in Table 7(a). That is, by altering the mix of Next, we address research question (5) that
product models to take advantage of the additional focuses on the impact of market competition. Table
discount offered by the focused vendors, the OEM 7(a)–(c) indicate that as the competition ratio b
is able to realize an increased profit. We see a similar increases (i.e., the OEM’s models face more
phenomenon for several other combinations of b, f, competition in the market), the OEM’s profit
and d values. A possible reason for this phenomen- decreases rapidly as expected. Interestingly how-
on is that when focused vendors increase their ever, there is a shift away from focused vendors to
discount percentage, the costs of expensive compo- general vendors in this problem scenario. That is, in
ARTICLE IN PRESS
N. Balakrishnan, A.K. Chakravarty / Int. J. Production Economics 112 (2008) 723–741 735
Table 7
Summary of results
OEM’s profit No. of focused No. of general Focused vendors’ General vendors’
vendors vendors profit profit
(a) d ¼ $50,000
5 0.5 10 1,583,766 2.10 0.80 890,850 344,994
15 1,790,039 2.30 0.85 1,017,805 359,538
20 1,911,765 2.25 0.90 1,080,295 394,362
1.0 10 1,128,177 1.90 0.80 636,682 284,616
15 1,369,230 2.05 0.80 763,335 338,234
20 1,530,731 2.15 0.90 862,889 356,454
1.5 10 865,661 1.70 0.85 478,730 264,937
15 1,102,372 1.85 0.80 605,024 312,463
20 1,271,927 1.90 0.85 696,342 347,426
10 0.5 10 1,910,953 2.80 0.60 688,173 140,754
15 2,157,186 2.95 0.60 773,128 152,264
20 2,302,958 2.90 0.60 817,264 176,284
1.0 10 1,360,104 2.60 0.60 495,808 129,359
15 1,649,294 2.75 0.65 596,386 148,178
20 1,842,860 2.75 0.80 654,589 185,434
1.5 10 1,043,887 2.35 0.50 389,333 104,669
15 1,327,232 2.65 0.70 486,848 140,891
20 1,530,678 2.60 0.80 546,460 176,841
15 0.5 10 2,261,228 2.95 0.45 356,661 77,271
15 2,551,284 3.15 0.35 401,072 82,103
20 2,722,228 3.20 0.35 428,153 80,676
1.0 10 1,615,939 2.85 0.40 259,464 77,251
15 1,957,326 2.95 0.45 311,149 90,005
20 2,181,920 3.00 0.45 344,851 99,471
1.5 10 1,244,730 2.65 0.40 204,450 67,936
15 1,579,129 2.95 0.45 256,269 77,811
20 1,816,500 2.95 0.45 291,314 89,007
(b) d ¼ $100,000
5 0.5 10 1,454,660 1.30 0.75 732,722 544,569
15 1,656,524 1.65 0.80 930,108 466,843
20 1,775,381 1.75 0.75 1,013,785 470,090
1.0 10 1,016,456 1.15 0.80 491,306 477,243
15 1,248,544 1.25 0.80 621,732 518,210
20 1,403,371 1.30 0.80 705,970 551,573
1.5 10 766,005 0.85 0.85 311,566 484,121
15 991,878 1.15 0.80 473,947 483,999
20 1,153,616 1.20 0.80 564,377 512,836
10 0.5 10 1,767,287 2.20 0.35 674,002 138,704
15 2,006,452 2.25 0.35 759,027 149,077
20 2,148,015 2.35 0.40 804,041 173,336
1.0 10 1,225,622 2.15 0.35 488,390 119,953
15 1,506,029 2.15 0.40 580,882 148,747
20 1,692,611 2.15 0.45 639,571 177,911
1.5 10 917,987 1.90 0.40 373,798 120,310
15 1,190,731 1.95 0.45 464,974 151,695
20 1,386,753 2.00 0.50 527,815 181,771
15 0.5 10 2,110,493 2.45 0.15 351,546 62,827
15 2,394,638 2.60 0.25 394,508 87,832
ARTICLE IN PRESS
736 N. Balakrishnan, A.K. Chakravarty / Int. J. Production Economics 112 (2008) 723–741
Table 7 (continued )
OEM’s profit No. of focused No. of general Focused vendors’ General vendors’
vendors vendors profit profit
(c) d ¼ $150,000
5 0.5 10 1,357,084 1.15 0.70 677,874 609,450
15 1,553,441 1.20 0.70 794,354 635,499
20 1,668,005 1.25 0.70 853,222 672,076
1.0 10 931,753 0.70 0.80 335,030 682,771
15 1,154,178 0.95 0.80 503,657 676,988
20 1,304,449 1.05 0.80 610,723 678,806
1.5 10 696,017 0.35 0.85 146,983 699,132
15 906,159 0.70 0.80 310,027 698,722
20 1,060,888 0.85 0.80 421,800 703,784
10 0.5 10 1,646,014 2.05 0.20 678,933 99,961
15 1,881,523 2.10 0.30 752,555 145,470
20 2,020,858 2.10 0.30 796,084 167,520
1.0 10 1,114,244 1.60 0.35 459,827 144,489
15 1,385,622 1.80 0.45 557,716 180,737
20 1,570,348 1.90 0.45 620,654 204,756
1.5 10 819,051 1.40 0.40 334,393 181,933
15 1,078,630 1.60 0.45 432,820 200,407
20 1,268,719 1.70 0.50 505,523 212,433
15 0.5 10 1,990,844 2.10 0.15 347,664 65,673
15 2,265,053 2.20 0.20 387,583 96,786
20 2,427,355 2.20 0.25 410,846 115,613
1.0 10 1,361,288 2.00 0.15 252,473 54,729
15 1,682,682 2.10 0.20 302,327 75,560
20 1,897,081 2.15 0.25 333,512 98,886
1.5 10 1,007,102 1.80 0.15 196,507 43,132
15 1,315,532 2.05 0.15 247,513 52,008
20 1,539,630 2.05 0.30 278,131 99,504
a highly competitive market, general vendors seem in the purchase of component variants. Hence, the
to have an edge over more focused vendors. As a OEM gravitates toward using fewer vendors to
consequence, while the total profit for both types of offset the vendor fixed costs. Clearly, these fewer
vendors tends to eventually decrease as b increases, vendors need to be general vendors so that they can
the magnitude of the decrease is much less provide the OEM with the versatility needed to
pronounced for general vendors when compared produce the entire product line that it decides to
to the focused vendors. We note that we had used offer.
this issue earlier in our explanation for the non- As the number of models the OEM plans to offer
significance of b for the general vendors’ profit. As (f) increases (all other factors held constant), all the
noted during that discussion, when competition 5 output measures tend to increase. This is not
increases, the overall reduced demand for its surprising given that by offering more models, the
products causes the OEM to lose scale economies OEM can hope to capture a larger share of the
ARTICLE IN PRESS
N. Balakrishnan, A.K. Chakravarty / Int. J. Production Economics 112 (2008) 723–741 737
market. Note that as the level of market competi- To study this issue in detail, we consider the
tion increases, the demand per product model specific problem for which the input was given in
decreases. Under such situations, the OEM is forced Table 3. Although we study this problem for all 81
to offer a larger number of product models to combinations of factor levels, we present and
achieve the same level of profits. discuss the results here only for the specific
We now discuss research question (4) that focuses environment for which the detailed results were
on the fixed cost of vendor relationships. As this presented in Table 4. Results for the other factor-
cost increases, Table 7(a)–(c) indicate that the OEM level combinations exhibit a similar behavior to the
tends to use fewer vendors of either type, and also results discussed below.
realizes decreased profits. Due to the decreased As shown in Table 4, the specific problem
overall number of vendors, the OEM is forced to scenario considers the situation where a ¼ 10%,
use the general vendors to a greater extent to supply b ¼ 0.5, f ¼ 10, and d ¼ $100,000. The current
the required component variants. This is evidenced optimal solution is to use focused vendor nos.
by the fact that while the total profit for focused 1 and 4, and general vendor no. 5. We now assume
vendors decreases as d increases, the profit for that the OEM is able to induce general vendor no. 5
general vendors tends to increase. to offer a discount on the unit purchase costs of
Finally, we discuss research question (2) that the component variants supplied by the vendor.
focuses on the types of vendors used. Based on the We further assume that this discount is 50% of
preceding discussion, the analysis indicates that the the discount offered by the focused vendors.
OEM tends to gravitate towards using general That is, while vendor no. 5 is cheaper than
vendors in preference to focused vendors when other general vendors, their unit costs are still
either the competition in the market increases higher than that of focused vendors for the same
(measured by b) or the fixed cost of vendor variant. The revised optimal results (with the 5%
relationships (measured by d) increases. discount from vendor no. 5) are presented in
The discussion so far has focused on an existing Table 8.
set of vendor capabilities and conditions. That is, we Comparing the results in Table 8 with that in
assume that vendors are static in terms of the Table 4, it is clear that the optimal course of action
component variants they can offer, and the price for the OEM is to divert business away from the
at which they can offer them. However, what focused vendors towards general vendor no. 5.
happens if a general vendor decides to compete While the optimal solution is this specific example is
with focused vendors on price, or if a focused to divert all business to the general vendor,
vendor decides to compete with general vendors on solutions in other scenarios recommend only partial
their range of offerings? Therefore, we now explore diversion. Interestingly, this diversion of business
situations where the OEM is able to induce 1 or typically results not only in increased profits for the
more vendors of a given type to work with it (OEM) OEM but also for the vendor. That is, the OEM and
in a manner that could potentially be beneficial to vendor no. 5 together are able to exploit the market
both parties. to their mutual benefit. In fact, the percentage
increase in vendor no. 5’s profits far exceeds that
4.4. Impact when a general vendor offers a discount realized by the OEM. Also, we note that the identity
of the product models that the OEM should offer
In this section, we explore the impact on the changes appreciably (only 3 of the original 10
OEM’s and vendors’ profits when a specific general models remain in the optimal set).
vendor can be induced (using incentives, as dis- The preceding discussion illustrates that a general
cussed below) to offer a discount on unit prices. vendor’s versatility in variant offering has compe-
Note that while a focused vendor may wish to titive value, both for the vendor and for the OEM.
match discounts offered by this vendor, it may not A general vendor can leverage its component variety
be profitable for him to do so given his already offering to increase profit even after reducing unit
discounted prices. When a general vendor offers prices. Observe that the OEM can, if needed,
such a discount, the attractiveness of focused provide additional incentives to a general vendor
vendors would be expected to decline further as to provide this discount by offering to share some of
now the general vendor becomes more competitive their increased profits with the vendor. That is, the
in all variants offered by focused vendors. OEM can essentially encourage a specific vendor to
ARTICLE IN PRESS
738 N. Balakrishnan, A.K. Chakravarty / Int. J. Production Economics 112 (2008) 723–741
Table 8
Results when general vendor no. 5 offers a discount
C1 C2 C3 C4 C5
offer discounts that other vendors may not be able component 4 is the one that is used most by the
to match. At the extreme, it may be possible for the OEM. We now assume that the OEM can induce
OEM to offer incentives and work closely with just focused vendor no. 4 to develop the capability to
1 or 2 vendors to make them its only vendors, while supply this variant. We further assume that vendor
simultaneously increasing profits for all concerned no. 4 can set the purchase price of this variant at the
parties. non-discounted price charged by vendor no. 5. That
is, vendor no. 4 does not compete with vendor no. 5
4.5. Impact when a focused vendor develops new on price for this additional variant. The revised
component variants optimal solution under this problem environment is
presented in Table 9.
Now that we have studied the impact when a Comparing the results in Table 9 with that
general vendor offers a discount, we next study the shown in Table 4, it is apparent that the new
impact when the situation is reversed. That is, what capability of focused vendor no. 4 makes it
happens when a focused vendor can be encouraged beneficial for the OEM to divert business away
by the OEM to develop capabilities for supplying from general vendor no. 5. While this specific
additional variants that the vendor does not example recommends complete diversion of busi-
currently offer. As in the preceding discussion, we ness from vendor no. 5, solutions in other scenarios
consider the specific problem presented in Table 3, recommend partial diversion. This diversion is
and focus our discussion on the specific environ- accompanied by an increase in the OEM’s profit
ment presented in Table 4. compared to the current situation, as well as an
Examining the results in Table 4, it appears that increase in profits for the focused vendors. As an
of all the component variants currently being added incentive, if necessary, the OEM can even
supplied by general vendor no. 5, variant 4 of subsidize development costs of additional variants
ARTICLE IN PRESS
N. Balakrishnan, A.K. Chakravarty / Int. J. Production Economics 112 (2008) 723–741 739
Table 9
Results when focused vendor #4 develops a new variant
C1 C2 C3 C4 C5
of component variants, albeit at higher unit costs), competitive in terms of building additional capabil-
and focused vendors (who can supply fewer ities, although their unit prices may be comparable.
variants, but possibly at lower unit costs). First, as While many companies have expressed the need
market competition (measured by b) increases, we to involve vendors in the product choice process,
find that the OEM shifts more of its business to not many have figured out the best approach for
general vendors increasing their profit. The shift doing so. Our model does make 2 rather restrictive
occurs because the reduction in demand of the assumptions in that we assume there are no price
OEM’s products (with increasing b values) de- discounts based on volume, and that the OEM as
creases the volume of business it can award to well as the vendors are not capacity constrained. As
focused vendors. The fixed cost of using a focused such, while we cannot claim to have solved the
vendor is no longer justifiable in relation to the problem completely, we believe that our research is
volume of business handled by it. an important step in the right direction. In our
Second, at some high value of b, the focused future research we intend to use the framework
vendors’ business gets reassigned in its entirety, as developed here to analyze important issues such as
the OEM experiences a large drop in its revenue. demand uncertainties, pricing discounts, capacity
With no additional business to be had from focused constraints, and logistics.
vendors, the profits of general vendors decrease with
further increases in b. These 2 issues cause the Appendix A
general vendors’ profit not to be significantly related
to the OEM’s profit when b increases. Theorem 4. For a given vendor set U, for models
Third, we note that while a price discount i 2 I^ and jeI,
^ PðI^ þ j i; UÞ4PðI;
^ UÞ if and only if
percentage (a) decreases the focused vendors’ profit, xjUXxiU and RjXRi.
the general vendors gain from such a discount—the
gain even exceeds the gain of the OEM. Note that Proof. Let X 0IU
^ ¼ Sk2I;kai
^ xkU and
0
with a larger discount percentage, the price of an Y I^ ¼ Sk2I;kai
^ Rk . Hence,
X
expensive component drops by a larger amount in PðI;^ UÞ ¼ X 0^ =½RC þ Y 0^ ^
dv if ieI. (A.1)
IU I
absolute terms. The OEM would like to exploit this v2U
by redesigning its product mix. The general vendors, ^ it follows that:
However, since i 2 I,
in contrast to focused vendors, are in a better
position to satisfy the OEM’s revised needs because X 0IU 0 0
^ =½RC þ Y I^ oðX IU
^ þ X IU
0
^ Þ=½RC þ Y I^ þ Rj
they possess a large component variety, and so their (A.2)
profit increases appreciably. This clearly establishes,
which simplifies to
at least for this example, that economy of scope of
general vendors is more valuable than the economy ðX iU =Ri Þ4 þ X 0IU 0
^ =½RC þ Y I^ . (A.3)
of scale of focused vendors. Now, since i 2 I^ and jeI,
^ we have (xjU/Rj)oxiU/
It is interesting that developing new capabilities Ri, which simplifies to:
for a focused vendor does not necessarily bring high
dividends to the OEM. Since there is no price ðxe^U xaU
^ Þoðye^ ya^ ÞðxaU
^ =ya^ Þ. (A.4)
advantage in using the new capability, the only way Consider a swap of i 2 I^ with jeI.
^ If the swap is
the OEM would gain is by saving the fixed cost of to be made, the profit must increase, i.e.,
the general vendor. To do so, the OEM needs to
ðX 0IU 0 0
^ þ xjU Þ=½RC þ Y I^ þ Rj 4ðX IU
0
^ þ X iU Þ=½RC þ Y I^ þ Rj .
redesign its product mix substantially, as the
component variety possessed by the focused vendor (A.5)
is not broad enough even when the new capability is The above expression can be simplified to
included. This reduces the OEM’s profit margin,
½RC þ Y 0I^ ðxjU xiU Þ4X 0I^U ðRj Ri Þ þ ðxiU Rj Þ ðxiU Ri Þ.
although saving the fixed cost increases its total
profit by a small amount. As more capabilities are (A.6)
added and the component variety of the focused From expression (A.4), it follows that (xiU
vendors increase, the OEM may be able to increase Rj)(xjU Ri)40. Hence, expression (A.6) can be
its profit margin, and its total profit may exceed rewritten as
what it was before adding new capabilities. Note
that not all focused vendors would be equally ðxjU xiU Þ4X 0IU 0
^ ðRj Ri Þ=½RC þ Y I^ . (A.7)
ARTICLE IN PRESS
N. Balakrishnan, A.K. Chakravarty / Int. J. Production Economics 112 (2008) 723–741 741
After multiplying both sides of expression (A.4) Cooper, L., Nakanishi, M., 1988. Market Share Analysis:
by 1, and from expression (A.7), it follows that: Evaluating Competitive Marketing Effectiveness. Kluwer
Academic Publishers, Dordrecht.
ðRj Ri Þ½X 0IU 0
^ =ðRC þ Y I^ Þ ðxiU =Ri Þo0. (A.8) Farlow, D., Schmidt, G., Tsay, A., 1996. Vendor Management at
Sun Microsystems (A), Case Study OIT-16A. Stanford
Since X 0IU 0
^ =ðRC þ Y I^ ÞoðxiU =Ri Þ from expression
University, Palo Alto, CA.
(A.8), this implies that RjXRi. Using this relation- Fisher, M., Ramdas, K., Ulrich, K., 1995. Using parts sharing to
manage product variety: a study in the automobile industry.
ship in expression (A.7), it follows that xjUX- Working Paper, University of Texas at Austin, Austin, TX.
xiU. & Gupta, S., Krishnan, V., 1999. Integrated component and vendor
selection for a product family. Production and Operations
Management 8 (2), 163–182.
Ishii, K., Juengel, C., Eubanks, C.F., 1995. Design for Product
References variety: key to product line structuring. In: ASME Design
Technical Conference Proceedings, September, Boston, MA.
Balakrishnan, N., Chakravarty, A.K., 1996. Managing engineer- Lillien, G., Rangaswamy, A., 1998. Marketing Engineering.
ing change: market opportunities and manufacturing costs. Addison-Wesley, Reading, MA.
Production and Operations Management 5 (4), 335–356. Macduffie, J.P., Sethuraman, K., Fisher, M.L., 1996. Product
Chakravarty, A.K., Balakrishnan, N., 2001. Achieving product variety and manufacturing performance: evidence from the
variety through optimal choice of module variations. IIE international automotive industry. Management Science 42
Transactions 33 (7), 587–598. (3), 350–369.
Chakravarty, A.K., Baum, J., 1992. Coordinated planning for Page, A., Rosenbaum, H., 1987. Redesigning product lines with
competitive products and their manufacturing operations. conjoint analysis: how sunbeam does it? Journal of Product
International Journal of Production Research 30, 2293–2311. Innovation Management 4, 120–137.