Professional Documents
Culture Documents
145–161
Abstract
In their review of the operations strategy literature, Anderson et al. wAnderson, J.C., Cleveland, G., Schroeder, R.G.,
1989. Operations strategy: a literature review. J. Operations Manage., 8Ž2.: 133-158x contend that the hypothesis that a
company will perform better if it links its operations strategy to the business strategy is intuitively appealing, but lacks
empirical verification. In light of this contention, this research attempts to: Ž1. define and measure the concept of fit as it
applies to operations strategy; Ž2. show how fit leads to better performance; and Ž3. investigate the interrelationships
between fit, business strategy, productivity, and performance. These objectives are investigated through field-based research
within a wholesale distribution service setting. Utilizing the classificatory framework of Venkatraman wVenkatraman, N.,
1989. The concept of fit in strategy research: toward verbal and statistical correspondence. Acad. Manage. Rev., 14Ž3.:
423-444x, fit is defined as the degree to which operational elements match the business strategy. This precise definition
closely resembles the concept of ‘external fit’ that began with the work of Skinner wSkinner, W., 1969. Manufacturing–mis-
sing link in corporate strategy. Harvard Bus. Rev., 47Ž3.: 136-145x. A conceptual model of business performance is used
with productivity as a mediating variable between the independent variables of business strategy and external fit and the
dependent variable of business performance. Path analysis is used to analyze the effect of external fit on performance and to
investigate the interrelationships between fit, business strategy, productivity, and performance. The results show that external
fit has a significant positive and direct effect on business performance. When coupled with the nonsignificant direct effects
of the strategy variables, this suggests that the fit of the operational elements with the strategy is of greater importance than
the particular choice of strategy. Although all three business strategies Žlow cost, a combination of low cost and high
customer service, and high customer service. had no significant direct effects on performance, a high customer service
strategy did have a significant positive effect on the intervening productivity variable. Finally, the particular design of the
research and the findings suggest that much of the conceptual work in operations strategy may be applicable to service
operations as well as manufacturing. q 1999 Elsevier Science B.V. All rights reserved.
Keywords: Empirical research; Measurement and methodology; Operation strategy; Productivity; Service operations; Statistical analysis
)
Corresponding author.
0272-6963r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.
PII: S 0 2 7 2 - 6 9 6 3 Ž 9 8 . 0 0 0 3 7 - 0
146 T.M. Smith, J.S. Reece r Journal of Operations Management 17 (1999) 145–161
One recent study has examined the fit between firm’s proach to fit described by Drazin and Van de Ven
manufacturing infrastructure practices and Just-in- Ž1985.. This approach defines fit in terms of consis-
Time manufacturing ŽSakakibara et al., 1997.. tency across multiple dimensions of organizational
The concept of external fit also began with the design and context. Bozarth and Berry Ž1997. utilize
work of Skinner Ž1969.. Skinner’s seminal work on the classificatory framework of Venkatraman Ž1989.
operations strategy describes the need for companies to evaluate the strategic fit between manufacturing
to have the proper external fit when developing and and marketing in support of the overall business
implementing a manufacturing strategy. He suggests strategy. The strength of the strategic fit is conceptu-
that companies should tailor their production systems alized as the degree of adherence for a specific unit
to perform the tasks which are vital to corporate of analysis with a multidimensional, ideal profile.
success and consistent with the corporate strategy. A Utilizing a different perspective from the classifica-
variety of authors claim that consistency between tory framework of Venkatraman Ž1989., this research
business strategies and manufacturing strategies, or a also attempts to address this problem by precisely
proper external fit, is an important component in the defining fit as the degree to which operational ele-
success of organizations ŽBuffa, 1984; Wheelwright, ments match the business strategy. Once defined and
1984; Fine and Hax, 1985; Schroeder et al., 1986; operationalized, this study also attempts to examine
Kotha and Orne, 1989; McDougall et al., 1992; how this external fit affects both productivity and
Miller and Roth, 1994.. Unfortunately, the evidence performance.
to support this claim is tenuous at best ŽAnderson et This research also deals with two of the major
al., 1989; Miller and Roth, 1994; Williams et al., weaknesses in operations strategy research as pointed
1995.. The few studies that have examined this area out by Adam and Swamidass Ž1989.. The first short-
have concentrated instead on either Ža. providing a coming is the use of productivity as the exclusive
conceptual framework that links business strategy measure of performance. Several recent empirical
with operations strategy ŽKotha and Orne, 1989; studies in operations have focused on productivity
Williams et al., 1995; Kim and Arnold, 1996.; Žb. and the factors that have led to higher productivity
defining the link between business strategy and pro- ŽHayes and Clark, 1985; Chew et al., 1989; Lieber-
duction operations ŽCleveland et al., 1989; Vickery man et al., 1990; Schmenner, 1991; Istvan, 1992;
et al., 1993.; or Žc. explaining how operations strat- Ittner, 1994.. While these studies support the notion
egy directly impacts business performance ŽRichard- that productivity gains are a function of several
son and Gordon, 1980; Swamidass and Newell, factors, both within and outside the control of man-
1987.. agement, they also caution that productivity was
Although it is clear that the concept of fit is used as a surrogate measure due to the limited access
central to both the theoretical and empirical research to other performance measures. Skinner Ž1986. also
in operations strategy, the extensive use of this con- points out that improvement efforts focused exclu-
cept for a variety of applications has led to some sively on productivity may well end up with disap-
confusion and may have retarded some of the re- pointing results if they are not consistent with the
search in this area ŽDrazin and Van de Ven, 1985; overall business and operations strategy. Several re-
Venkatraman, 1989.. As Venkatraman points out, cent empirical studies have begun to alleviate this
‘‘the role of fit in strategy research has been severely situation by using a variety of financial and market
handicapped by the absence of appropriate links measures to test the effectiveness of operations strat-
between the concept and theory testing.’’ He goes on egy ŽSwamidass and Newell, 1987; Cleveland et al.,
to state that a leading cause for this has been the lack 1989; Vickery et al., 1993; Williams et al., 1995;
of a precise definition and operationalization of the Gupta and Somers, 1996; Ahmed et al., 1996.. The
fit variable. second shortcoming deals with the lack of research
Some recent studies have begun to address this that studies the interrelationship among variables,
concern. In trying to develop a model of fit that is particularly the effect of strategy on performance.
consistent with the fit assertions of configurational With the exception of the works by Swamidass and
theories, Doty et al. Ž1993. utilize a systems ap- Newell, and Gupta and Somers, the studies men-
148 T.M. Smith, J.S. Reece r Journal of Operations Management 17 (1999) 145–161
in their article on alternative research paradigms in the schedule of data collection activities—were ac-
operations. The particulars of this study, in terms of complished through meetings and correspondence in
the conditions suggested by Yin and in the explana- July 1992 with SIA management. The thirty branch
tory nature of the objectives, strongly suggest the visits were conducted from August through Novem-
case study as the most appropriate research method- ber of 1992.
ology. In addition, the testing of hypotheses within The heart of the protocol is the specific set of
the service setting of this study involves demonstrat- questions used for actual inquiry. These questions
ing a theory’s applicability under circumstances not are geared toward understanding the individual data
previously investigated ŽMcCutcheon and Meredith, items representing the study’s four main constructs.
1993.. Since a major source of data is through unstructured
Given the selection of the case-study methodol- interviews, the questions provide a general plan of
ogy and the objective of theory verification, the inquiry, but not a precise set of questions that must
design characteristic of single or multiple case stud- be asked in specific words or in a particular order. A
ies needs to be determined. The logic underlying the list of these questions is provided in Appendix A.
use of multiple case studies is similar to the use of Yin Ž1989. lists six sources of evidence that can
multiple experiments, that is, the use of ‘replication’ be the focus of data collection for case studies. These
logic. Each case is selected so that it either predicts sources are documentation, archival records, inter-
similar results Ža literal replication., produces con- views, direct observations, participant observation,
trary results but for predictable reasons Ža theoretical and physical artifacts. Three of these sources are
replication., or produces results that simply do not used in this study for direct access to information as
support the theory. However, if there are enough well as a means to corroborate information gathered
cases, some forms of inferential statistical analysis from other sources. Archival records in the form of
are possible ŽFlynn et al., 1990.. Therefore, for this financial reports, budgets, and operating reports are
study, a modest application of inferential statistics is used for all four variables. Unstructured interviews
feasible based on the number of cases that are used of a majority of the employees at each branch are
Žthirty branches.. used for the three predictor variables. Finally, direct
observation is used for the three predictor variables,
primarily to corroborate and validate the findings
3.3. Data collection methods from the first two sources. The use of multiple
sources helps with both the validity and reliability of
All observations and branch interviews were per- the study. In addition, a database containing the
formed by the first author. Financial and operating various field notes, documents, and narratives col-
data collection was done by SIA personnel and lected over the course of the study was maintained to
verified by the first author. A case study protocol is improve reliability. The particular framework for
used for increasing the reliability of research. The empirical research described in the preceding pages
major components of the protocol are the field pro- is proposed by Flynn et al. Ž1990.. Fig. 2 represents
cedures and the case-study questions. The two topics a summary of the research methodology used for this
for the field procedures—access to interviewees and study according to their framework.
3.4. Definition and operationalization of Õariables moderate service Žcombination. has an indirect
and significant effect on business performance
The four constructs found in Fig. 1 are discussed through its direct effect on productivity.
in the following section. Business Strategy is de- H3c: A business strategy of customer service has
fined as the basis upon which an organization an indirect and significant effect on business per-
achieves and maintains a competitive advantage formance through its direct effect on productivity.
ŽWheelwright, 1984.. The generic strategies of cost Two data items ŽS1 and S2. are used to generate a
leadership and differentiation of Porter Ž1980. are business strategy index ŽSIndex. and to categorize
used to polarize the business strategies employed at the business strategy for each branch. ŽAll data items
each branch. The other generic strategy proposed by are summarized in Table 1.. The first data item ŽS1.
Porter of focusing on a particular strategic target was is based on both the stated strategy of the branch
not applicable in this setting and was therefore not employees as well as the observed strategic behavior
considered. For the distribution branches observed in within the marketplace. For example, each branch
this study, a differentiation strategy takes the form of employee was interviewed and each branch was
superior customer service. observed on the specific approaches and tactics em-
The four stages of service firm compe- ployed in trying to secure new customers. This data
titiveness—available for service, journeyman, dis- item is computed on a three-point scale. Each branch
tinctive competence achieved, world class service was given a subjective rating upon completion of the
delivery—found in Chase and Hayes Ž1991. are also branch visit and interviews based on the criteria
used to further structure the distinction between the established by Chase and Hayes Ž1991.. A value of 1
two extreme strategies of low cost and high customer was given for low cost branches; a value of 2 for
service. Branches that best fit the description of combination branches; and a value of 3 for high
available for service were employing a low cost customer service branches.
strategy and branches that best fit the description of In this particular industry, branches providing
distinctive competence achieved or world class ser- higher customer service Ždifferentiation strategy.
vice delivery were employing a differentiation strat- should be able to receive a higher price. Therefore,
egy. A combination strategy of both low cost and the second data item ŽS2. is a three-point scale that
differentiation, or firms Porter Ž1980. dubs as ‘stuck attempts to measure the price premium received at
in the middle’, most closely resembled a journeyman each branch. A value of 1 was given for branches
description. Therefore, three distinct strategies—low whose premium was greater than one standard devia-
cost, moderate cost and moderate service Žcombina- tion below the mean price premium received at all
tion., and high customer service—describe the busi- branches. A value of 3 was given for branches whose
ness strategy employed in the marketplace at each premium was greater than one standard deviation
individual branch. As a result, the hypotheses regard- above the mean price premium received at all
ing business strategy ŽH2 and H3. are redefined into branches. The rest of the branches received a value
the following six hypotheses: of 2. This data item is then used to validate, and if
H2a: A business strategy of low cost has a direct applicable, adjust S1. If no adjustment is made, the
and significant effect on business performance. business strategy index ŽSIndex. is then simply the
H2b: A business strategy of moderate cost and S1 data item. An adjustment is made if the absolute
moderate service Žcombination. has a direct and difference between the S1 and S2 data items is
significant effect on business performance. greater than one. Thus, the business strategy con-
H2c: A business strategy of customer service has a struct is operationalized into three variables—Low
direct and significant effect on business perfor- Cost Strategy, Combination Strategy, and Customer
mance. SerÕice Strategy. The SIndex is found in Appendix
H3a: A business strategy of low cost has an B.
indirect and significant effect on business perfor- The concept of fit is a useful building block for
mance through its direct effect on productivity. theory construction, but the term has been used
H3b: A business strategy of moderate cost and inconsistently within the strategic management litera-
Table 1
151
152 T.M. Smith, J.S. Reece r Journal of Operations Management 17 (1999) 145–161
ture ŽDrazin and Van de Ven, 1985; Venkatraman, considered but not included for the reasons stated
1989.. Venkatraman Ž1989. develops a useful con- above are the sophistication of reports, ease of cus-
ceptual framework and identifies six perspectives of tomer and employee inventory inquiry, inventory
fit. Corresponding to one of these perspectives, and replenishment policy, location of purchasing activi-
within the context of this study, fit is viewed as a ties, and the tracking of backorders.
matching concept—a match between two related Each data item is converted to a three-point scale,
variables without reference to a criterion variable. In but the conversion process involves two steps. The
this study, the measure of fit between the two vari- first step requires the identification of a branch oper-
ables Žbusiness strategy and operational elements. is ating characteristic for each of the eight operational
developed independent of any performance anchor. elements; the second step involves the determination
The distinction between business strategy and exter- of how well this operating characteristic matches the
nal fit lies in where they are implemented andror intended business strategy. A proper match for each
observed. Business strategy describes behavior in the data item is hypothesized subsequent to the pilot
marketplace and how a branch attempts to attract and study but prior to the branch visits. To illustrate,
retain customers. The operational elements describe locating the order processing activities outside of the
characteristics and particular decisions made within branch in a centralized location is hypothesized to
the branch; the external fit then describes how well support a low cost strategy; locating these activities
these characteristics complement the marketplace inside the branch is hypothesized to support a cus-
strategy. Therefore, External Fit is defined as the tomer service strategy. A literature review of the
consistency between the business strategy and the hypothesized matches in these infrastructure decision
decision categories Žoperational elements. that con- areas provided little direction, especially in a service
stitute the overall operations of the branch. firm setting. For example, the work by Hayes and
The major decision categories that comprise these Wheelwright Ž1984. concentrates on long-term struc-
distribution branches include warehousing, materials tural decisions, while the infrastructure decisions
management Žpurchasing and inventory., order pro- investigated by Van Dierdonck and Miller Ž1980.
cessing, organization structure and control, and sales and Cleveland et al. Ž1987. are more applicable to a
management ŽLambert and Stock, 1982.. These deci- manufacturing setting. Therefore, the hypothesized
sion categories are roughly akin to the infrastructure match for each of the fit data items is primarily the
decisions of workforce management, inventory and result of interviews with experts in the distribution
logistics systems, and organization proposed by field, discussions with the SIA management team,
Wheelwright Ž1978. and expanded by Hayes and and the researchers’ knowledge. The hypothesized
Wheelwright Ž1984. and Fine and Hax Ž1985.. proper fit for each data item is provided in Table 1.
In keeping with the typology of operational deci- Operating and financial data provided by SIA was
sions suggested by Hayes and Wheelwright Ž1984., used to identify the operating characteristic for Fit3
eight data items are used to operationalize the exter- and Fit7. The operating characteristic for the remain-
nal fit variable. The data items are found in Table 1. ing six fit data items ŽFit1, Fit2, Fit4, Fit5, Fit6, and
Of the eight data items, three pertain to inventory Fit8. was identified by the researcher through direct
and logistics decisions ŽFit1, Fit3, and Fit4., three observation. The operating characteristic for each of
pertain to workforce issues ŽFit2, Fit6, and Fit7., and the eight operational elements was rated on a three-
two pertain to organization ŽFit5 and Fit8.. These point scale. A characteristic supporting a low cost
particular operating elements are chosen because they strategy was given a score of 1; a characteristic
are under the direct control of the branch manager supporting a customer service strategy was given a
and there are differences in the elements among the score of 3; and a characteristic that supported both
branches. For example, the vendor selection decision strategies, but to a lesser degree, was given a score
is an important one for fit considerations. But, be- of 2.
cause the branch managers in several locations did Once identified and scored, the operating charac-
not have control over this decision, this particular teristic score was compared to the value for the
data item was not included in the index. Other items branch’s business strategy. A ‘poor’ fit Žvalue s 1.
T.M. Smith, J.S. Reece r Journal of Operations Management 17 (1999) 145–161 153
was given if the absolute difference between the two are standardized after they have been calculated. The
measures was greater than one; an ‘average’ fit productivity index ŽProIndex. for an individual
Žvalue s 2. was given if the absolute difference be- branch is then simply the standardized average of the
tween the two measures was equal to one; and a three data items. An equal weighting scheme is used
‘good’ fit Žvalue s 3. was given if the absolute because there is no prior knowledge as to the relative
difference between the two measures was zero. For importance of each data item and interviews with
example, Branch 411 performs its order processing SIA management provided no additional insight. In
activities outside the branch. This operating charac- addition, it accurately reflects the proportion of the
teristic is hypothesized to support a low cost strategy branch expenses devoted to human resources and
and therefore receives a score of 1. Since the busi- warehouse space. The productivity construct is thus
ness strategy for Branch 411 is a combination strat- operationalized as the variable called ProductiÕity of
egy ŽSIndexs 2., the order processing activities for Operations. The ProIndex is found in Appendix B.
Branch 411 receives an ‘average’ fit value ŽFit1 s 2.. Business Performance is defined as the ability of
Finally, after the scores have been computed, the the operations to satisfy the desires of the company’s
eight individual fit data items for each branch are major stakeholders. The major stakeholders in this
standardized and used to calculate an index for each study are shareholders, employees, and customers.
individual branch ŽFitIndex.. This index is simply The shareholders of this company are limited to
the standardized average of the eight data items. An employees of the company and range from top cor-
equal weighting scheme is used because there is no porate management to salaried branch personnel.
prior knowledge as to the relative importance of each Therefore, to a large extent the desires of the share-
data item. Interviews with SIA management pro- holders mirror those of the employees.
vided insufficient reasons to change this weighting Two different measures of business performance
scheme. The external fit construct is thus opera- were initially conceived: return on capital employed
tionalized as the variable called Operational Fit. The ŽROCE. and two-year growth rate in sales. ROCE is
FitIndex is found in Appendix B. a proxy measure for how well the shareholders are
ProductiÕity is defined as the ratio of real output being satisfied. This measure is appropriate in this
produced to real resources consumed ŽKearney, study for two reasons: Ž1. it is the measure upon
1978.. Inherent in this definition are two elements of which branch management is evaluated, and Ž2. since
productivity: the efficiency of the transforming re- each branch has leased warehouse space and equip-
sources and the utilization of transforming resources. ment, no distortion arises from depreciation of non-
In this study, the productivity variable considers both current assets. The two-year growth rate in sales is
aspects. For this particular industry, the primary used as a surrogate measure of customer satisfaction.
inputs used in the transformation process of whole- This measure is appropriate in light of the presence
sale distribution are human resources and physical of competitors and the frequency of repeat pur-
space. Therefore, this variable will consider the effi- chases, as well as the nonexistence of internal or
ciency of the human resources and the utilization of external customer satisfaction measures.
warehouse space. As a result of the pilot study and in a further
Three data items ŽPro1 to Pro3. are used to review of the factors within the control of branch
generate the productivity index ŽProIndex.. The data management, several problems arose in the use of
items are listed in Table 1. The first two data items these business performance measures. First, informa-
ŽPro1 and Pro2. are selected because they provide tion regarding the age of the 30 branches reveals the
meaningful measures on the efficient use of the most range to be 6 months to 15 years. Eight branches
significant Žin terms of monetary expenditures. input have been in operation less than two years. The sales
—human resources. The third data item ŽPro3. is a growth rate experienced by these ‘younger’ branches
measure of the utilization of the warehouse space. far exceeds those of the larger and more established
Operating and financial data provided by SIA was branches. Therefore, due to the wide variation in the
used to calculate the three productivity data items. age of the branches and with several branches having
Each of the three individual productivity data items started within the past two years, two-year sales
154 T.M. Smith, J.S. Reece r Journal of Operations Management 17 (1999) 145–161
growth rate is an inappropriate and unreliable mea- US$0.50 per square foot to a low of US$0.13 per
sure of relative business performance among the 30 square foot. Since this analysis is focused on branch
branches. management and branch operations, some adjustment
Second, ROCE is a good measure of relative is necessary to account for the differences resulting
branch performance if ROCE is positiÕe for all 30 from the location decision. Consequently, the
branches, because it properly reflects positive gains profitrloss statement for each branch is adjusted to
in both profit margin and capital turnover ŽROCEs reflect the company-wide average cost per square
profit margin= capital turnover.. However, if ROCE foot. The resulting analysis explores the question,
is negatiÕe for more than one branch, then ROCE is ‘‘How well do branches perform and what deter-
a poor measure of relative branch performance. In mines this performance given that a certain location
fact, ROCE is an unsuitable measure of relative to operate has been established and the rental cost
performance any time the profit margin is negative. per square foot of warehouse space has been equal-
Since 13 of the 30 branches had a negative profit ized across all branches?’’ This adjustment facilitates
margin in 1992, ROCE is not a good choice for the making the distinction between the economic perfor-
relative performance comparisons needed in this mance of the branch and the managerial performance
study. of the manager.
The decision of how many distribution branches Therefore, the business performance measure
to operate and where to locate them is made by ŽBp1. chosen as the dependent variable is defined as
corporate management. The most obvious result of adjusted profit after tax percentage ŽAdj PAT%..
this decision, in terms of branch performance, is the This variable is simply a branch’s profit after tax
difference in the cost per square foot of warehouse percentage calculated from the profitrloss statement
space at the different branches. The average rental after it has been adjusted to equalize the cost per
cost per square foot of warehouse space for the 30 square foot of warehouse space across all branches.
branches is US$0.31. The cost ranges from a high of Fig. 3 illustrates the Adj PAT% for all 30 branches
during the year 1992. The values for this variable are presumed that these four variables cause variation in
also provided in Appendix B. the endogenous variables Ž ProductiÕity of Opera-
tions, Business Performance., but variations in the
3.5. Data analysis methods exogenous variables are not to be explained by the
model.
According to Asher Ž1983., the technique of path Further, in the path analytic model shown in Fig.
analysis is basically concerned with estimating the 4, D1 and D2 are disturbance terms associated with
magnitude of the linkages between variables and the two endogenous variables. These disturbance
using these estimates to provide information about terms are used to account for the variations in the
the underlying causal processes. It also enables one two endogenous variables that are attributable to
to measure the direct and indirect effects that one causes not included in the formal structural model.
variable has on another and to decompose the corre- Two injunctions were adhered to in using a path
lation between any two variables into a sum of analytic model. First, the number of variables in-
simple and compound paths; some of which may be cluded in the model were kept to a minimum. Sec-
meaningful and others which may not. Given these ond, only those relationships that had theoretical
advantages, path analysis is the technique used in support were included in the model. The arrows in
this study to explore the hypothesized relationships Fig. 4 indicate the causal order assumptions deduced
among the four variables. Fig. 4 is the conceptual from the literature.
model of business performance using path analysis Missing data is not a major problem in this study,
conventions. but all the information for each data item for each
As was previously stated in Section 3.4, business branch is not available. Data is missing either be-
strategy is operationalized into three variables—Low cause it is not currently tracked at a particular branch,
Cost Strategy, Combination Strategy, and Customer the information cannot be separated in a meaningful
SerÕice Strategy, and external fit is operationalized way, or not enough data points are available to
as the variable called Operational Fit. As shown in construct the measure. In total, 27 out of 30 branches
Fig. 4, and according to path analysis conventions, have complete data available. Missing data is han-
these four variables are exogenous variables. It is dled in a pairwise fashion. This technique computes
each correlation coefficient by using cases with com- gressed on the intervening variable times the path
plete data for the pair of variables correlated. Be- coefficient obtained from regressing the intervening
cause of the relatively small sample size, this ap- variable on the final dependent variable. For exam-
proach is used because it utilizes the maximum ple, the indirect effect of a low cost strategy on
amount of data available without inferring data or business performance is obtained by multiplying the
dropping entire cases if some data is missing. path coefficient for low cost strategy obtained from
the regression equation using productivity as the
dependent variable with the path coefficient for pro-
4. Interpretation of results ductivity from regressing on business performance
Žy0.12 = 0.54 s y0.07.. The total effect is equal to
4.1. Results the addition of the direct and indirect effects. Unfor-
tunately, since the direct effect of the three strategy
In order to obtain estimates of the main path variables are nonsignificant and the indirect effect of
coefficients, one simply regresses each endogenous the operational fit variable is nonsignificant, the total
variable on those variables that directly affect it effect numbers presented in the table provide little
ŽAsher, 1983.. No constant terms were included in meaning. On the other hand, other results provided
the regressions since all three strategy variables were by the path analysis do allow for some meaningful
included in the regression runs. Table 2 reports the interpretation and potential contribution.
standardized Beta coefficient, the t-ratio, and the
p-value resulting from the regression analyses for 4.2. Analysis of results
each of the arrows Žrelationships. depicted in Fig. 4.
Fig. 5 then shows the path coefficients Žstandardized Productivity has a direct and significant effect on
regression coefficients. and the residual path coeffi- business performance in light of the standardized
cients within the path analytic model. The residual regression coefficient of 0.54 and a p-value of 0.003
path coefficient is simply the square root of the Žsee Table 2.. This result is neither unexpected or
unexplained variation in the dependent variable in remarkable, but it does lend some credence to the
question. It is calculated as the square root of 1 y R 2 , data collection and measurement methods. This find-
where R 2 is the square of the appropriate multiple ing also adds some degree of confidence to those
correlation coefficient. studies that use productivity as a surrogate measure
Table 3 shows the direct and indirect effects of for business performance.
the four exogenous variables and the one intervening Of the six hypothesized relationships between the
variable on business performance. To calculate the business strategy variables and business performance
indirect effect of the exogenous variable on the final ŽH2a–H3c., only one ŽH3c. was significant. A cus-
dependent variable, one multiplies the path coeffi- tomer service strategy indirectly affects business per-
cient obtained for the exogenous variable when re- formance through its significant effect on productiv-
Table 2
Standardized regression coefficients and their significance
Variable ProductiÕity as the dependent variable Performance as the dependent variable
Beta coefficient t-ratio p-value Beta coefficient t-ratio p-value
Low cost strategy y0.12 y0.63 0.532 0.13 0.84 0.408
Combination strategy y0.19 y1.02 0.320 0.20 1.35 0.190
Customer service strategy 0.37 a 1.96 0.062 y0.04 y0.28 0.782
Operational fit 0.13 0.65 0.525 0.45 b 2.96 0.007
Productivity of operations nra nra nra 0.54 b 3.37 0.003
R 2 s 0.228 2
R s 0.563
a
Significant at the 0.10 level.
b
Significant at the 0.01 level.
T.M. Smith, J.S. Reece r Journal of Operations Management 17 (1999) 145–161 157
ity. As shown in Table 2, the standardized regression flexible workforce as opposed to a more specialized
coefficient is 0.37 and the p-value is 0.062. All other one. But, even though it was hypothesized that a
coefficients involving the strategy variables are not customer service strategy should employ a flexible
significant. In trying to understand the significantly workforce, the low cost strategy branches in this
positive coefficient, it is helpful to explore the corre- study were just as likely to have a flexible workforce
lations between the individual data items that make as the customer service strategy branch. Therefore,
up these two variables and those of the operational this does not provide an explanation for the signifi-
fit variable. The productivity variable has a high cant indirect effect of a customer service strategy.
positive correlation Žsignificant at the 0.05 level. Nevertheless, this positive correlation is an intriguing
with two operating characteristics—the breadth of finding in that it goes against the traditional wisdom
employee activitiesrresponsibilities and the exper- that greater specialization leads to higher productiv-
tise level of employees. ity.
The first correlation is interesting because it sug- On the other hand, the second significant correla-
gests that higher productivity may result from a more tion does provide a reasonable explanation, albeit not
a full one, for the indirect effect of a customer
service strategy. Eight out of the ten customer ser-
Table 3 vice branches have employees of high expertise level.
The direct and indirect effects of the four exogenous variables and
the intervening variable on business performance
This high expertise level seems to lead to higher
productivity, which in turn leads to higher perfor-
Variable Effect
mance. This phenomenon is also consistent with
Direct Indirect Total observations performed during the branch visits. This
Low cost strategy 0.13 y0.07 0.06 would suggest that the higher salaries paid to these
Combination strategy 0.20 y0.10 0.10 employees with high expertise are justifiable in terms
Customer service strategy y0.04 0.20 0.16
Operational fit 0.45 0.07 0.52
of bottom line improvements.
Productivity of operations 0.54 y 0.54 The operational fit variable has a significant direct
effect on business performance. As shown in Table
158 T.M. Smith, J.S. Reece r Journal of Operations Management 17 (1999) 145–161
2, the standardized regression coefficient is 0.45 and venient, but not necessarily the most appropriate. In
the p-value is 0.007. Coupling the strong significant addition, the hypothesis that it is the underlying
direct relationship with the nonsignificant direct ef- superiority of the operating characteristic, and not its
fects of the strategy variables seems to imply that match with the business strategy, that yields superior
choosing operating characteristics that fit a particular performance is a possibility that requires further
strategy is a far more important determinant of busi- attention.
ness performance than choosing the ‘right’ strategy. Third, this study demonstrates that external fit has
In other words, there were both successful and un- a significant positive and direct effect on business
successful low cost branches, combination branches, performance. This finding is the first to lend support
and customer service branches. But, one of the main to the intuitively appealing claims of Skinner Ž1969.
distinguishing factors of the successful branches was and Wheelwright Ž1984. and comes at a time when
that the operating characteristics fit the branch’s many are beginning to question the effectiveness and
particular strategy, regardless of the type of strategy usefulness of the traditional manufacturing strategy
they employed. paradigm ŽClark, 1996; Hayes and Pisano, 1996..
Finally, based on the R 2 for the entire model, 56 Coupled with the nonsignificant direct effects of the
percent of the variation in business performance is strategy variables, it suggests that the fit of the
explained by the three strategy variables, the opera- operational elements with the strategy is of greater
tional fit variable, and the productivity of operations importance than the particular choice of strategy. As
variable. This is an interesting finding given the with all studies in which many of the key extraneous
model used for the study was fairly simple and the variables are controlled for gathering data for one
objective was to understand interrelationships as op- particular industry, and for one company in this case,
posed to trying to ‘explain’ business performance. the generalizable nature of these findings must be
tempered to a certain extent. However, the strong
statistically significant results for this relationship
5. Contributions and conclusions suggest that the findings are quite reliable. Once
again, these findings provide evidence and support
This study provides several contributions to the for others to build upon in trying to ascertain emerg-
field of operations strategy. First, by utilizing the ing patterns within this field.
classificatory framework of Venkatraman Ž1989., this The fourth contribution involves understanding
study more precisely defines the concept of external the effects of strategy on performance. It appears that
fit first proposed by Skinner Ž1969.. Fit is defined as certain types of strategies influence performance in
the degree to which operational elements match the an indirect manner. In a simplistic model, one may
business strategy. This is the first study to examine jump to the conclusion that a customer service strat-
this critical concept utilizing empirical methods egy has very little effect on performance or that
within a field-based setting. other strategies Žlow cost or combination. have a
Second, by utilizing the infrastructure decision relatively greater positive effect. But, the significant
categories proposed by Wheelwright Ž1978., this effect of customer service on the intervening variable
study operationalizes the concept of fit according to of productivity produces the result that customer
the previous definition. Infrastructure decisions that service has a significant indirect effect on perfor-
provide a good fit with a particular business strategy mance. This result may lend support to Deming’s
are hypothesized prior to data collection. This is the notion that improvements focused on customer satis-
first study to operationalize this concept. In doing so, faction, though often hard to quantify, are ultimately
others can begin to build, challenge, and improve beneficial to the long-term health of the organization
upon the definition and the conception of this vari- ŽDeming, 1986..
able. The operationalization of the fit concept re- Finally, this study has utilized empirical methods
quires additional perspectives and the consideration of analysis and field-based research to examine con-
of additional operating characteristics. For example, cepts that seem reasonable, yet lack empirical verifi-
the characteristics examined in this study were con- cation. The particular design of the research and the
T.M. Smith, J.S. Reece r Journal of Operations Management 17 (1999) 145–161 159
findings suggest that much of the conceptual work in Who are their major customers? Why?
operations strategy may be applicable to service What makes customers choose your company?
operations as well as manufacturing. What major things have changed in the past two
years?
What is the education level of your employees?
Acknowledgements How much do you pay your employees?
empirical analysis of the product-process matrix. Manage. Sci. Thompson, J.D., 1967. Organizations in action. McGraw-Hill,
42 Ž11., 1576–1591. New York.
Sakakibara, S., Flynn, B.B., Schroeder, R.G., Morris, W.T., 1997. Van Dierdonck, R., Miller, J.G., 1980. Designing production
The impact of Just-in-Time manufacturing and its infrastruc- planning and control systems. J. Operations Manage. 1 Ž1.,
ture on manufacturing performance. Manage. Sci. 43 Ž9., 37–46.
1246–1257. Venkatraman, N., 1989. The concept of fit in strategy research:
Schmenner, R.W., 1991. International factory productivity gains. Toward verbal and statistical correspondence. Acad. Manage.
J. Operations Manage. 10 Ž2., 229–254. Rev. 14 Ž3., 423–444.
Schroeder, R.G., Anderson, J.C., Cleveland, G., 1986. The con- Vickery, S.K., Droge, C., Markland, R.E., 1993. Production com-
cept of manufacturing strategy: An empirical study. J. Opera- petence and business strategy: Do they affect business perfor-
tions Manage. ŽAugust., 405–415. mance. Decision Sci. 24 Ž2., 435–455.
Skinner, W., 1969. Manufacturing—missing link in corporate Wheelwright, S.C., 1978. Reflecting corporate strategy in manu-
strategy. Harvard Business Rev. 47 Ž3., 136–145. facturing decisions. Business Horizons 21 Ž1., 57–66.
Skinner, W., 1974. The focused factory. Harvard Business Rev. Wheelwright, S.C., 1984. Manufacturing strategy: Defining the
52 Ž3., 113–121. missing link. Strategic Manage. J. 5 Ž1., 77–91.
Skinner, W., 1986. The productivity paradox. Harvard Business White, R.E., Hamermesh, R.G., 1981. Toward a model of busi-
Rev. 64 Ž4., 55–59. ness unit performance: An integrative approach. Acad. Man-
Skinner, W., 1988. Report of production and operations manage- age. Rev. 6 Ž2., 213–223.
ment research needs committee. Operations Manage. Rev. 7 Ž1 Williams, F.P., D’Souza, D.E., Rosenfeldt, M.E., Kassaee, M.,
and 2., 17–23. 1995. Manufacturing strategy, business strategy and firm per-
Stobaugh, R., Telesio, P., 1983. Match manufacturing policies and formance in a mature industry. J. Operations Manage. 13,
product strategy. Harvard Business Rev. 61 Ž2., 113–120. 19–33.
Swamidass, P.M., Newell, W.T., 1987. Manufacturing strategy, Yin, R.K., 1989. Case study research: Design and methods. SAGE
environmental uncertainty and performance: A path analytic Publications, Newbury Park, CA.
model. Manage. Sci. 33 Ž4., 509–524.