Beruflich Dokumente
Kultur Dokumente
Information Systems Department, Sam M. Walton College of Business, University of Arkansas, Fayetteville, AR 72701, USA
b Department of Information and Operations Management, Lowry Mays College of Business, Texas A&M University,
College Station, TX 77843, USA
Abstract
Research in the areas of both manufacturing and marketing/sales have advocated the integration of several important
interrelated decisions between the two functions (i.e. product development, process development, marketing/sales planning,
and manufacturing planning decisions). The process of managing the strategic alignment between a firms business strategy,
external environment, and the integration of manufacturing and marketing/sales decisions is very complex phenomenon that
requires a level of analysis that has not occurred previously. This study examined the moderating effects of business strategy
and demand uncertainty on the relationship between the integration of manufacturing and marketing/sales-based decisions
and organizational performance. The study found general support for the proposed model, suggesting that the impact of the
integration of manufacturing and marketing/sales decision on organizational performance is moderated by a firms business
strategy and demand uncertainty. 2002 Elsevier Science B.V. All rights reserved.
Keywords: Marketing/operations integration; Operations strategy; Empirical research
1. Introduction
The integration of key decision areas between manufacturing and marketing/sales is widely cited as a
means for gaining a competitive advantage in the marketplace (e.g. Shapiro, 1977; Wheelwright and Hayes,
1985; Nemetz and Fry, 1988; Konijnendijk, 1994).
Although there is anecdotal support that integration
of decisions between these two functions may lead
to increased organizational performance, there is little
empirical research to support this claim. In addition,
most anecdotal studies tend to ignore the substantial
Corresponding author. Tel.: +1-501-575-4035;
fax:+1-501-575-4168.
E-mail addresses: solearykelly@walton.uark.edu
(S.W. OLeary-Kelly), b-flores@tamu.edu (B.E. Flores).
1 Tel.: +1-409-845-4248.
0272-6963/02/$ see front matter 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 2 7 2 - 6 9 6 3 ( 0 2 ) 0 0 0 0 5 - 0
222
223
Table 1
Key decision areas between marketing/sales and manufacturing
Decision area
Functional domain
Manufacturing
Marketing/sales
Table 2
The integration of manufacturing and marketing/sales decisions: previous research
Decision areas
Type of research
Conceptual
Analytical
Blois (1980)
Fitzsimmons et al. (1991)
Hayes and Wheelwright (1979)
Hill (1988)
Konijnendijk (1994)
Nemetz and Fry (1988)
Shapiro (1977)
Stalk and Hout (1990)
Utterback and Abernathy (1975)
Wheelwright and Clark (1992)
Wheelwright and Hayes (1985)
Manufacturing and
marketing/sales planning
Crittenden (1992)
Damon and Schramm (1972)
Leitch (1974)
Empirical
224
225
226
its own goals. The more assertive each party is, the
more they tend to interact. In this sense, a high level of
integration is characterized by a high degree of both
cooperation and interaction. Paralleling prior definitions, in this study, the level of integration refers to
the extent to which separate parties work together in
a cooperative manner to arrive at mutually acceptable
outcomes. Accordingly, this definition encompasses
constructs pertaining to the degree of cooperation,
coordination, interaction, and collaboration.
This study will investigate the contingency relationship regarding the integration of these four manufacturing and marketing/sales decisions previously
discussed: manufacturing planning decisions, marketing/sales planning decisions, product development
decision, and process development decisions.
4.2. Business strategy
A business strategy represents a pattern of decisions
regarding how an organization will compete in its
market. Research has strongly supported the proposition that a firms business strategy is comprised of
multiple dimensions. Numerous studies have assessed
an organizations business strategy by measuring the
competitive emphasis placed on several dominant
strategic dimensions (e.g. Miller, 1988; Venkatraman,
1989b; Zahra and Covin, 1993).
The following strategic variables have been identified in the literature as comprising important components of a firms business strategy: differentiation via
product innovation, cost leadership, superior product
quality, on-time delivery, and breadth of product-lines.
With the exception of on-time delivery, these strategic variables/dimensions have been used extensively
to gauge a firms overall business strategy (Hambrick,
1983; Porter, 1985; Miller, 1988; Zahra and Covin,
1993). These variables, along with their relationship
to the different manufacturing and marketing/sales decisions, are detailed below.
4.2.1. Differentiation via product innovation
Product innovation is widely recognized as a common approach that companies choose in order to compete in the marketplace. Companies that adopt this
approach tend to sell products that are on the cutting
edge of technology. They are characterized by their
extensive involvement in new product research and de-
227
H4. For firms that adopt a business strategy of superior product quality, there will be a positive association
between the level of decision integration with respect
to both product development and process development
decisions and firm performance. For firms that do
not adopt a strategy of superior product quality, there
will not be an association between integration and
performance.
228
5. Research method
5.1. Data collection
The firms included in this study were sampled
across multiple industries (i.e. five two-digit SIC
codes). These industries included: primary metal,
fabricated metal products, industrial machinery and
equipment, electronic equipment, and transportation
equipment. Firms were selected from the Wards
Business Directory of US Private and Public Companies, and included only those that were located in the
central US and had more than 200 employees.
For each SBU, data was collected from three key
informants (a CEO level executive, a marketing/sales
executive, and a manufacturing executive) via a mailed
survey following the guidelines suggested by Dillman
(1978). A two-stage process was used to collect the
data. First, companies were contacted by a letter sent
to the head of each firm. Interested respondents were
asked to complete and return participation forms providing the information required to mail the surveys
to the appropriate individuals within the firm. Next,
a survey, a pre-paid return envelope, and a brief description of the study were mailed to the three informants within each company. The initial mailing of the
surveys was followed-up by two reminder cards sent
to non-respondents after the second and fourth weeks.
After 3 weeks, the initial contact letter, a second mailing was made to all non-respondents.
A total of 846 firms were contacted, of which 35
were unable to participate for restructuring reasons
(e.g. they were recently bought or sold, or they were
being consolidated with another division) and 22
non-deliverable letters were returned, providing an effective sample size of 789 firms. A total of 121 firms
returned usable surveys from all three informants.
This represents an effective response rate of 15.3%
which is comparable to other studies in which multiple high level respondents were used (e.g. Phillips,
1981; Phillips and Bagozzi, 1986). The vast majority
of informants held top level positions in their respective areas and had been with the firm on average for
over 13 years.
In order to assess whether a response bias was
present, a Chi-square goodness-of-fit test was used
to compare the study sample to the original mailing sample with regard to: (1) the type of business
unit, (2) sales levels, and (3) number of employees. All three tests failed to reject the null hypothesis of equal distributions at the 0.10 level. A
KolmogorovSmirnov one-sample test was also used
to compare the study and mailing sample regarding
both annual sales and the number of employees. Both
tests failed to reject the null hypothesis of equal distributions at the 0.05 level. Although these results
help establish our confidence in a lack of response
bias, they do not preclude biases related to other factors (e.g. firm performance) that were not tested due
to the lack of available data in the original mailing
sample.
5.2. Sample
The general profile of the study sample is outlined
in Table 3. In order to limit the influence of factors
external to our study, we examined several factors
that might influence a firms financial performance:
34.1
4.7
31.8
28.7
<0.1
19.4
43.4
5.4
28.7
3.1
Sales range
$ 429 million (%)
$ 3050 million (%)
$ 5190 million (%)
$ 91158 million (%)
$ 159850 million (%)
17.8
23.2
20.2
20.2
18.6
Number of employees
200300 (%)
400500 (%)
600700 (%)
8001000 (%)
11004700 (%)
28.6
16.3
16.3
18.6
20.2
229
is not influenced by industry type. In addition, the effects related to both the type of products marketed by
firms and the type of firm also were examined using
an unbalanced ANOVA model. The results also indicate that perceived profitability is not influenced by
either product type or firm type.
5.3. The measures
The measures (except the self-reported ROI) used
to operationalize the constructs in the study were comprised of multiple items, each based on a seven-point
Likert scale. In addition, all constructs were measured
across multiple key informants. Key informants were
selected on the basis of their specialized knowledge
relevant to the study (Kumar and Stern, 1993). The
use of key informants to collect data is an effective
means for measuring organizational constructs and has
been used extensively in business strategy research
(Phillips, 1981; Kumar and Stern, 1993). In addition
to relying on previous research to ensure the content
validity of the measures, they also were reviewed by
several expert judges (e.g. current and previous managers with manufacturing experience) and adjustments
were made based on their recommendations.
5.3.1. Business strategy
Business strategy was operationalized via five dimensions: differentiation via product innovation, cost
leadership, superior product quality, on-time delivery,
and product breadth. The items used to measure a
firms business strategy are based on past research.
Specifically, differentiation via product innovation
captures the extent to which a firm competes on both
the number and rate of product innovations they introduce into the marketplace. The measurement items
used for this construct are based on the work of
Cooper (1987), Miller (1988), Capon et al. (1992),
and Zahra and Covin (1993).
The cost leadership variable stems from the work
of Porter (1980, 1985), which highlighted low cost as
a central strategic position that could be adopted by
a firm. The items used in this study were primarily
adopted from Zahra and Covins (1993) cost leadership scale. The items used to measure the superior
product quality variable are based on eight different
sub-dimensions associated with quality management
as identified by Saraph et al. (1989). The items used
230
must first be established before the internal consistency of a measure may be assessed. Because the
constructs were measured across multiple informants
within each firm, the added burden of establishing that
the measures are the same (i.e. have the same factor
weights) across all informants also must be met.
Unidimensionality was assessed using a confirmatory factor analysis (CFA) model. In order to establish
standard-unidimensional measures across all three
informants, it was necessary to re-specify several
of the measures. In order for a construct to be unidimensional, the factor loadings for all items associated
with the construct must be significant and the model
must be correctly specified, as indicated by acceptable
goodness-of-fit values (Gerbing and Anderson, 1988;
OLeary-Kelly and Vokurka, 1998). Following the
suggestions of Anderson and Gerbing (1982, 1988),
items that did not have significant factor loadings
across all three informants were eliminated. Although
this approach yields a set of standard measures for
analysis, the limitation is that the set of measures may
be less generalizable.
The next step involved establishing that the constructs were based on identically weighted items
(same item factor loadings) across all informants. This
was accomplished by testing the hypothesis that factor
items were equal across informants, using LISRELs
multi-sample analysis (Jreskog and Sorbom, 1989).
The results of this analysis indicated strong support
for equal factor weights across all informants and further demonstrates that the informants are providing
ratings of the same construct.
Cronbachs (1951) coefficient and Werts et al.
(1974) composite coefficient of reliability ( c ) were
used to assess the reliability of the measures. Although
there are no definitive rules regarding reliability, scales
with of 0.70 or higher (Nunnally, 1978), and c of
0.50 or higher are generally considered to be reliable
(Werts et al., 1974). Expect for the cost leadership
scales, all of the measures had a Cronbachs (1951)
and a composite reliability above 0.7 (values ranged
from 0.71 to 0.92). For the cost leadership scales, both
of the reliability coefficients ranged from 0.49 to 0.55.
Although these values are low, Van de Ven and Ferry
(1980) have argued that for broader measures, those
comprised of several conceptually distinct terms (as
in the case of the cost leadership measure), values
around 0.50 are acceptable.
231
232
Table 4
Partitioning of variance due to trait, method (informant), and error
Measure (informant)
Variance components
Trait
Method
Error
Moderating
Product innovation (CEO)
Product innovation (marketing)
Product innovation (manufacturing)
Cost leadership (CEO)
Cost leadership (marketing)
Cost leadership (manufacturing)
Superior quality (CEO)
Superior quality (marketing)
Superior quality (manufacturing)
On-time delivery (CEO)
On-time delivery (marketing)
On-time delivery (manufacturing)
Product breadth (CEO)
Product breadth (marketing)
Product breadth (manufacturing)
Demand uncertainty (CEO)
Demand uncertainty (marketing)
0.42
0.61
0.40
0.53
0.30
0.42
0.28
0.14
0.35
0.58
0.35
0.33
0.64
0.25
0.21
0.38
0.51
0.06
0.09
0.07
0.01
0.00
0.00
0.37
0.53
0.46
0.22
0.22
0.18
0.00
0.01
0.01
0.12
0.03
0.52
0.30
0.53
0.46
0.70
0.58
0.35
0.33
0.19
0.20
0.43
0.49
0.36
0.74
0.78
0.50
0.46
Integration variables
Product development decisions (marketing)
Product development decisions (manufacturing)
Marketing planning decisions (marketing)
Marketing planning decisions (manufacturing)
Process development decisions (marketing)
Process development decisions (manufacturing)
Manufacturing planning decisions (marketing)
Manufacturing planning decisions (manufacturing)
0.03
0.76
0.07
0.61
0.30
0.02
0.46
0.05
0.48
0.20
0.65
0.26
0.45
0.70
0.40
0.66
0.49
0.04
0.28
0.13
0.25
0.28
0.14
0.29
Performance variables
Perceived profitability (CEO)
Perceived profitability (marketing)
Perceived profitability (manufacturing)
0.51
0.86
0.54
0.07
0.05
0.07
0.42
0.09
0.39
variablesa
As noted by Campbell and Fiske (1959), only traits with the same informants should be tested together. Because the demand
uncertainty variable was not measured by the manufacturing informant, a separate set of models were tested using only the CEO and
marketing/sales informant. The results reported for demand uncertainty are based on this separate set of tests (it should be noted that the
variance for the other constructs were all within one standard error of the models ran for the strategy variables using all three informants).
Significant at the 0.05 level.
Significant at the 0.01 level.
measures, the criterion for convergent validity (significant trait factors across informants for a given measure) was not met, thus, indicating a significant lack
of agreement between the two informants (manufacturing and marketing/sales). The one exception is with
respect to the marketing/sales planning decision integration measure, where the trait factors are significant
for both informants. However, the trait factor for the
marketing/sales respondent is only marginally signifi-
sales-based decisions (product development and marketing/sales planning) are significant. Similarly, with
regard to the marketing/sales respondent, the trait
variances for the two manufacturing-based decisions
(process development and manufacturing planning)
are significant. Based on these strong results, we used
the manufacturing-based responses to measure the
level of integration regarding product development
and marketing/sales planning decisions. Likewise, we
used the marketing/sales-based responses to measure
the level of integration regarding process development
and manufacturing planning decisions.
The results for the integration variables are not altogether unexpected. Other studies have found similar results for marketing/sales and R&D functions,
leading researchers to conclude that members of these
functions often do not share the same perceptions regarding the level of integration between the two functions (Gupta et al., 1985; Saghafi et al., 1990; Souder
and Sherman, 1993). For example, because marketing/sales tends to have greater control over the outcome of product development decisions, they may
have more of a biased view regarding their own willingness to cooperate and involve (i.e. integrate) manufacturing into the decision process.
Finally, the analysis of the perceived performance
measures strongly support the attainment of convergent validity. Specifically, all trait variances across informants were significant at the 0.005 level, while the
methods variances were all near zero, strongly indicating a lack of informant bias. Based on these results,
it was deemed appropriate for the perceived performance measures to be combined across informants.
5.4.2. Discriminant validity
Discriminant validity refers to the extent to which
the different measures are unique. Two methods were
used to assess discriminant validity at both the monoand multi-method level. First, discriminant validity
was assessed by measuring the correlation between
all constructs and evaluating whether they were different from 1.0 (OLeary-Kelly and Vokurka, 1998).
All variables included in the study were determined
to be significantly different from 1.0. This model was
then compared to one in which the correlation coefficient ( ij ) was constrained to 1.0 (representing
unity between constructs). A Chi-square difference
test between the values obtained for the constrained
233
and unconstrained models was used to assess discriminant validity. A significantly lower Chi-square
value for the unconstrained model provides support for
discriminant validity (Anderson and Gerbing, 1988).
All Chi-square difference tests were significant at the
0.005 level, indicating strong support for discriminant
validity at the mono-method level. Second, a more
conservative test of discriminant validity requires that
the average variance extracted for each construct exceed the squared correlation between that construct
and all others in the model (Fornell and Larcker, 1981).
All of the constructs met this criteria, providing additional evidence of acceptable discriminant validity
among the constructs.
In summary, all construct measures were determined to exhibit acceptable levels of convergent and
discriminant validity at the mono-method level of analysis. With the exception of the integration variables,
all independent variable measures were determined
to have sufficient levels of convergent and discriminant validity at the multi-method level of analysis.
Based on these results, the average summated scores
(i.e. the average of the item scores for each variable)
were used as a composite measure for each construct
(Hair et al., 1992).
With regard to the integration variables, which did
not meet the convergent validity criteria at the multimethod level of analysis, the composite measures for
each integration variable were based on a single informant. That is, the measures of process development
and manufacturing planning decision integration were
derived from the data provided by the marketing/sales
informant. Likewise, the measures of product development and marketing/sales planning decision integration were derived from the data reported by the
manufacturing informants. Summary statistics for the
final variables are reported in Table 5.
5.5. Data analysis
Moderator median split regression analyses (MMSRA) were used to analyze the data (Arnold, 1982).
MMSRA involves splitting the sample into two
sub-samples (high and low) with respect to the median value of the moderating variable. The criterion
variable (perceived profitability) is then regressed on
the predictor variable (e.g. product development decision integration or manufacturing planning decision
234
235
6. Results
The results for the moderated median split regression analyses are reported in Table 6. Included are the
coefficients of the integration variable (i.e. predictor variable) pertaining to each hypothesis, the significance of the coefficient, and the P-values associated
with the difference t-test of the coefficients in each
sub-sample. Support for a hypothesis exists when the
analysis provides evidence of both a significant coefficient in the predicted direction and a significant
difference t-test. Partial support for a moderated relationship exists when one of the coefficients is significant, but the difference between the two sub-samples
is non-significant.
6.1. Effects of integrating marketing/sales-based
decisions
Marketing/sales-based decisions include both product development and marketing/sales planning (highlighted by the shaded areas in Table 6). The results
involving the marketing/sales based decisions mostly
support the hypotheses, i.e. a moderated relationship
between the level of decision integration and performance. Specifically, four of the six hypotheses were
strongly supported, indicating that the effect of marketing/sales based decision integration on performance
is dependent on the business strategy and demand uncertainty faced by the firm.
For example, for firms pursuing a strategy of either
product innovation (H1a) or superior quality (H4a),
increasing the level of integration with regard to
product development decisions yielded improved performance. Likewise, for firms that pursued a strategy
of on-time delivery (H3a), higher levels of product
Manufacturing-based decisions involve both process development decisions and manufacturing planning decisions. These results also tend to support
a moderated relationship between the level of decision integration and performance, however, they were
not in the predicted form. For example, the results
demonstrated that for firms emphasizing a strategy
of product innovation (H1b), higher levels of process
development decision integration were not associated
with improved firm performance. Instead, we found
that for the those firms that did not focus on product
innovation, a higher level of process development decision integration was associated with improved firm
performance. Similarly, for firms that competed via
on-time delivery (H3b), or that emphasize a strategy
of product breadth (H5b), higher levels of integration
regarding manufacturing planning decisions were not
associated with higher firm performance; instead, we
again found that for those firms that did not focus
on either on-time delivery or product breadth, higher
levels of integration were associated with higher firm
performance.
Finally, there was partial support in the opposite
direction regarding the moderating effects of demand
uncertainty (H6b) on the relationship between the
integration of manufacturing planning decisions and
firm performance. In this case, for those firms that
faced lower levels of demand uncertainty, the level of
integration was significantly associated with higher
firm performance, but higher levels of integration
were not associated with higher firm performance
for those firms facing higher levels of demand uncertainty. However, there was only partial support
for an integrationperformance relationship, in that
the coefficients across the two sub-samples (high
236
7. Discussion
The results of this study support the principal argument guiding this paper, that is the relationship
between the integration of manufacturing and marketing/sales decisions and firm performance is
moderated by a firms business strategy and by environmental uncertainty. An interesting finding of this
study was the differentiated results regarding the form
of the moderated relationship. Specifically, the results were found to vary with respect to both the type
of decision being analyzed (marketing/sales-based
versus manufacturing-based decisions), and the key
informant used to measure integration (manufacturing
respondent versus marking/sales respondent).
In regard to the former, the results supporting the
hypothesized form of the moderating relationships
all involved marketing/sales-based decisions (product
development and marketing/sales planning), and those
that were in the opposite form than hypothesized in
the study all involved manufacturing-based decisions
(process development and manufacturing planning).
One possible explanation is the time differential that
exists between marketing/sales and manufacturing
decisions, in that marketing/sales-based decisions are
typically a source of input for the manufacturing-based
decisions. For example, in a typical marketing/sales
manufacturing planning cycle, the marketing/sales
planning decisions serve as a primary input for the
manufacturing planning decisions which then follow
(see Fig. 2) (Vollmann et al., 1997). Here, it is more
advantageous for decision integration to occur at the
marketing/sales planning stage. If manufacturing has
a greater opportunity to review and identify potential
237
238
8. Concluding remarks
Overall, this study has challenged conventional
thinking regarding the impact of decision integration
between manufacturing and marketing/sales on organizational performance. The results clearly demonstrate that the effect of decision integration between
manufacturing and marketing/sales on firm performance is much more complex than has been suggested
previously. Second, the results indicate that marketing/sales and manufacturing respondents have very
different perceptions regarding the level of integration
for decision areas that they traditionally control; this
has important implications for both researchers and
managers who desire to assess the level of decision
integration. Finally, this study found very different
forms of moderating effects associated with decision
integration for the two functional areas. Although, it
presently is impossible to attribute these differences
to a specific factor, some suggestions for a potential
starting point for future research into this interesting
phenomenon were presented.
References
Adler, P.S., 1995. Interdepartmental interdependence and coordination: the cast of the design/manufacturing interface. Organizational Science 6 (2), 147167.
239
240