Operations management (OM) has seen the many new management practices sum up total quality management and lean production. A new model in OM based on the assumption that the adoption of best practice in a wide range of areas leads to superior performance.
The universal value of these practices has frequently stemmed from unreliable case studies of ‘‘world class manufacturing’’ firms, which tend to be large and operate in global, modern and dynamic industries. These are the companies that make up the samples used in the practice–performance empirical studies.
even though supporters of the universal view of OM best practices would disagree that implementation difficulties are part of moving the organization towards excellence , an substitute details is that these difficulties result from too great a difference between the proposed form of best practice and the particular organizational context . Research in maturing OM best practices has typically anchored on acontingency approach and examines relationships between contextual variables, the use of practices and the associated performance outcomes that’s the research OM practice contingency research (OM PCR).
the key purpose of this paper is to inspect and analysis the current state of OM PCR i.e.
(i) add to a better definition of this body of research as an area of study in OM;
(ii)characterize and synthesize research to date and identify its limitations;
and (iii) identify a number of tasks that future research should undertake in order to provide more solid conceptual foundations on which to anchor rigorous research in this area.
- Contingency theory and its relevance to OM practice contingency research
OM problems have a cross-disciplinary nature and has led to the broadening of the scope of the OM field and the desirability of conducting interdisciplinary research . CT is a major theoretical lens used to view organizations and this theory holds that organizations adapt their structures in order to maintain fit with changing contextual factors, so as to attain high performance . Theoretical and practical contributions of this approach are achieved by
(i) identifying important contingency variables that distinguish between contexts;
(ii) grouping different contexts based on these contingency variables;
and (iii) determining the most effective internal organization designs or responses in each major group
CT has give up many insight and has received substantial empirical support.Seminal work in CT was the precursor of major OM contingency approaches such as Skinner’s (1969, 1974) notions of fit between the production system and the priorities of the organization These approaches resulted the manufacturing strategy contingency paradigm.
This all shows that OM field has been strongly rooted in a contingency paradigm Therefore, we take the perspective that CT can be a very useful theoretical lens to view OM issues, in particular in areas where OM theory is less well developed. OM PCR shares some of the theoretical assumptions of CT, such as an economic efficiency and normative mindset. Despite the growing importance of OM PCR, little application of CT has taken place in this area.
- Scope of the review of OM PCR
In the field of CT, contingency studies involve three types of variables. Contextual (or contingency) variables ,Response variables and Performance variables . Also analyzing the literature across these dimensions grouped along three axes:
(i) research variables and measurement;
(ii) research design;
(iii) employed form of fit. In this procedure, we identify a number of tasks that future research should undertake .
- OM PCR through the lens of CT: research variables and measurement
OM practice contingency model comprises three sets of variables: use of practices, contingency factors and performance. As table 1 shows the most mature sets of practices quality management and lean production received the most attention. In this section, we review OMPCR in terms of:
(i) the contingency and performance variables that have been addressed;
(ii) measurement issues across the three sets of variables.
4.1. Contingency variables
The contingency variables grouped into four broad categories: national context and culture, firm size, strategic context, and other organizational context variables
A first group of studies was one of the first areas of interest in OMPCR due to which the emergent best practices had their origin in Japan. The growth of globalization has encouraged additional cross-country/cultural research that’s all support the existence of contingency effects.
A second group of studies has examined the use of practices across firms of different sizes(applicability in smaller firms ). While studies addressing quality management found no evidence of firm size effects .
A third group of studies examined the use of OM practices across different strategic contexts, and is generally rooted in the manufacturing strategy contingency paradigm of the OM field.
A final group of studies addresses a number of loosely related factors associated with the general context of organizations. All of them provide support for contingency effects.
4.1.1. Future research
We identify a number of issues related to contingency variables that need to be addressed. Apart from national context and firm size, the studies analyzed in Table 1 have employed a wide variety of variables to characterize organizational contexts. Too many contingency variables may limit generalizability and hamper the comparison of results between different studies. Better research could be conducted if it were possible to
identify a limited set of contingency variables defined as relevant for the OM discipline and that distinguish between contexts, similar to what has been accomplished in the CT field. The challenge is to identify the contingencies that explain the greatest variance in performance. This identification process might be started with a thorough examination of the literature, drawing both on theoretical grounds (e.g., Sitkin et al., 1994 for strategic context effects) to generate a comprehensive list of factors, and existing empirical results to establish preliminary relevance. Table 1 also shows that several OM studies employ different contingency variables that could be expected to be highly correlated. For example, Sitkin et al. (1994) and Reed et al. (1996) use as their main contingency variable ‘‘organizational uncertainty’’, while Hendry (1998) addresses the ‘‘policy for satisfying customer demand’’(make-to-order vs. other policies).
From an OM perspective, these two variables might be expected to be highly correlated and might be candidates for collapsing into a more general contingency variable, such as, for instance, ‘‘product-process matrix positioning’’ (e.g., high variety – low volume operations might be seen as having high degrees of uncertainty and more frequently employing make-to-order policies, while the reverse would be expected in their low variety – high volume counterparts).
Avoiding highly correlated variables provides opportunities to reduce the set of relevant OM contingency variables by consolidating such variables. The development of empirical taxonomies of contextual variables would be a useful avenue to identify a limited set of key variables.
4.2.1. Future research
We recognize the usefulness of examining different contingency models addressing different types of performance variables. However, we propose that, within the realm of OM PCR, special attention should be given to operational performance aspects, although other types of performance might be examined in addition. Implicit in this stance is the assumption that OM contingency theory should aim at producing prescriptive knowledge targeted at increasing an organization’s operational performance (which, in turn, may affect other types of performance variables, such as customer satisfaction). Traditional operational variables (also called competitive priorities or operations performance objectives) include cost, quality, delivery and flexibility (e.g., Schmenner and Swink, 1998; Ward et al., 1998).
OM PCR research has not addressed operational performance impacts in sufficient depth, and future research could benefit from increasing the examination of contingency models with multiple dimensions of operational performance. Of particular interest would be to examine whether the adequate match between OM practices and context differs according to the operational performance dimension in question (for example, is the set of OM practices appropriate for a small size operation the same whether we consider cost or flexibility performance?).
The comparability of different continghency studies and their contribution let us know the three sets of variables: use of practices, contextual variables and performance. There are many different measures and scales available for measuring the same performance variable
This general pattern is reflected in the contingency
studies in Table 1.
This diversity of measurement affects practice–context–
performance relationships and thus may be an
explanation for the conflicting findings observed across
some of the contingency studies in Table 1. For example,
while most studies found an impact of national context
and culture on quality management practices, the study by
Sila (2007) found none.
4.3.1. Future research
A consolidation and categorization effort is clearly
needed to foster sense making and generalizability.
Bringing together extant scales and metrics for general
OM research (Roth et al., 2007) is a strong contribution to
this. A particular challenge for contingency research is to
develop measures that are both valid and comparable
across different contexts. Increasing the generalizability of
a measure to encompass different contexts may reduce its
validity, because better data can be obtained by carefully
crafting measures for specific situations (Boyer and Pagell,
2000). Similarly, the use of objective measures is problematic
when different contexts are examined, as these
measures tend to be context-specific (Ketokivi and
We submit two possible ways forward. One is the
development of general perceptual measures and scales.
This would require research designs which address their
limitations, for example by having multiple respondents
and adopting appropriate and rigorous examinations of
validity (Ketokivi and Schroeder, 2004a). The other is to
employ research designs which control for as many
relevant factors as possible besides the contextual factors
under examination. Such designs increase the likelihood of
developing valid measures because such measures are
required to span less diverse contexts which only differ in
respect to the contextual variables under study. For
example, Sousa and Voss’s (2001) investigation of strategic
context effects was single-industry, which enabled the use
of industry-specific measures for relevant research variables,
some of them based on objective data. Drawing on
the output of the consolidation of contextual variables
proposed earlier, OM researchers could also develop
measures targeted to a limited set of particularly relevant
types of contexts (for example, measures that are valid for
a particular product-process matrix positioning that could
be used in contingency studies examining contextual
factors other than such positioning).
- OM PCR through the lens of CT: research design
the research design is categorized as noninferential, inferential-aggregate and inferential-detailed. Inferential designs are those that allow for the making of rigorous inferences as to the degree of applicability of practices across different contexts. Both conceptual and empirical studies may be considered inferential. Within inferential designs, we distinguish between two types of studies: (i) inferential-aggregate: studies which are designed to investigate the existence of differences in the use of practices at an aggregate level (typical format of hypotheses/conclusions: H0: There are differences in the use of a set of practices across different contexts); (ii) inferential-detailed: studies that go beyond the former, and are designed to investigate the existence of differences in the use of practices at a detailed level, specifying the effects of different contexts on individual practices
Of the 35 studies included in Table 1, close to two thirds
(24) employ an inferential research design. Of these, 18
studies are classified as inferential-detailed and 6 studies
as inferential-aggregate. The rest of the studies (11) remain
at the comparison level attempting to uncover differences
in the use of practices in different contexts, employing an
exploratory mode; the contingency stance in these studies
is only implicit, in that detected differences may indicate
the existence of contingency effects. The large majority of
the studies (30) have an empirical component and most of
them suggest the existence of relevant contingency effects,
which reinforces the importance of conducting contingency
5.1. Future research
While existing OM PCR shows a good degree of
inferential power, there is room for improvement. An
important aspect in designing contingency studies is the
choice of the point in time, relative to the initial adoption of
a given set of best practices, at which to empirically assess
fit. Discussion of this aspect is absent from most of the
studies in Table 1.We propose that the assessment of fit in
OM PCR should concern the match between context and
practices when these have reached a stable level of
development. This is for two reasons. The first is related
to the need for organizations to resort to experimentation
in adopting and selecting practices. In the present business
environment of fast diffusion and innovation in managerial
concepts, organizations may have no choice but to
experiment with many untried innovative practices while
searching for a few appropriate ones, because the costs of
these experimentations may be much lower than the
returns from using the surviving practices (Abrahamson,
1991). The second reason is the generally accepted view
that there are time lags between the implementation of
practices and their performance effects (e.g., Reed et al.,
1996; Hendricks and Singhal, 1997).
This favors the study of mature OM practice settings.
Most studies in Table 1 have not controlled for implementation
maturity, although their reliance on typically
large samples has reduced potentially adverse effects of
this lack of control. Ideally, future studies, especially if
employing smaller samples, should control for practice
maturity. This could be assessed, for example, by
estimating the typical length of time for different sets
of practices to achieve maturity in an organization or by
developing actual measures (or indicators) of maturity.
Some research has been conducted in this area for quality
management practices (e.g., Ahire, 1996; Dale and
Lascelles, 1997), but more research is needed for other
- OM PCR through the lens of CT: employed form of fit
In conducting contingency research, different forms of
fit can be employed (Doty et al., 1993). Two prominent
classifications of forms of fit in CT have been those of
Drazin and van de Ven (1985) and Venkatraman (1989).
Drazin and van de Ven (1985) consider three distinct forms
of fit, based on the configuration of the relationships
between contextual, response and performance variables
that are examined (selection, interaction and system
approach). Venkatraman (1989) puts forward six different
forms of fit based on the degree of precision of the
functional form of fit and the number of variables
considered in the fit equation (moderation, mediation,
matching, gestalts, profile-deviation and co-variation).
In this section, we examine OM PCR studies according
to the form of fit that they employ and discuss their
different roles in generating contingency knowledge. Due
to its parsimony, we employ Drazin and van de Ven’s
(1985) classification. Fig. 1 summarizes this classification
and its correspondence with Venkatraman’s (1989)
In the selection approach, fit is seen as a basic
assumption underlying congruence propositions between
the organizational context and response variables. This
approach does not examine whether the proposed context-
response relationships affect performance. Table 1
shows that there has been substantial use of the selection
approach (24 studies), fairly well spread out across the four
groups of contextual variables. The interaction approach
sees fit as the interaction of pairs of organizational contextresponse
variables which affects performance. Table 1
shows that the use of the interaction approach has been
lower (7 studies).
The selection and interaction approaches tend to focus
on how single contextual factors affect single response
variables. Advocates of the system approach argue that the
understanding of context–response relationships must
address simultaneously the many contingencies, response
alternatives and performance criteria that must be
considered holistically to understand organizational
design. Fit is seen as the internal consistency of multiple
contingencies and multiple response variables which
affects performance characteristics. The system approach
has recently incorporated the concept of equifinality by
interpreting fit as feasible sets of equally effective
alternative designs, with each design internally consistent
in its structural pattern and with each set matched to a
configuration of contingencies facing the organization
(e.g., Doty et al., 1993; van de Ven and Drazin, 1985). In
simple terms, the equifinality argument states that there
are multiple, equally effective ways in which an organization
can achieve fit.
In OM, a system view of practices has been adopted by a
number of authors. In the practice–performance stream a
number of studies have found evidence of strong interactions
between severalOMpractices (e.g., Cua et al., 2001;
Flynn et al., 1999; Kaynak, 2003; Shah and Ward, 2003,
Fig. 1. The selection, interaction and system forms of fit (Drazin and van
de Ven, 1985) and correspondence with Venkatraman’s (1989) six forms
- Sousa, C.A. Voss / Journal of Operations Management 706 26 (2008) 697–713
2007), suggesting that besides their individual effects,
their mutual interactions significantly affect performance.
However, the use of the system approach in OM PCR has
been limited. Table 1 shows that only four studies
employed this approach. Of these, two have adopted a
full system approach, considering bundles of practices and
contextual variables (Sousa, 2003; Sousa and Voss, 2001),
and two other have adopted only a partial system
approach, considering bundles of practices, but examining
contextual variables individually (Koufteros et al., 2002,
2005). Within the system approach, we found the
equifinality argument to be absent from the OM PCR
studies in Table 1.
We observed that OM PCR studies have not explicitly
considered the existence of distinct forms of fit. This is
consistent with the fact that many of these studies did not
position themselves as contingency studies. In addition,
the literature review seems to show a natural progression
of knowledge-building along time. Studies employing the
selection approach have been the earliest to appear, and
many have employed non-inferential designs. In contrast,
all empirical studies employing the interaction and system
approaches have been published after 2000 and all employ
inferential designs. Finally, we did not find any OM PCR
studies which performed triangulation between the three
approaches to fit.
6.1. Future research
We identify two main research needs. The first is to
increase the use of the system view in OM PCR. Two
possible reasons may explain the sparse use of the system
approach. One may be the reductionistic approach that is
dominant in OM empirical research, whereby organizations
such as manufacturing plants are studied by breaking
them into their constituent parts (Ketokivi and Schroeder,
2004b, p. 64). Another reason may be the difficulties that
are involved in addressing complex forms of interactions
among variables. We put forward several suggestions to
increase the use of the system view in OM PCR.
First, OMPCR scholars should consider the application of configurational research methods (Meyer et al., 1993), an endeavor that has already been embraced by research in operations strategy (Bozarth and McDermott, 1998; Boyer et al., 2000). Second, OM PCR scholars may readily draw on work such as Venkatraman’s (1989), which provides anoverview of analytical methods that can be used to test system forms of fit, including profile deviation approaches (Venkatraman and Prescott, 1990), which have been employed in other areas ofOM(e.g., Ahmad and Schroeder, 2003; Das et al., 2006; da Silveira, 2005). These analytical methods, based on several statistical techniques, are ideally suited to the survey methodology, which has been the most frequently used method in OM PCR. Third, OM scholars may wish to increase the use of methodologies oriented towards theory building, such as case research.
Whilst survey research is excellent for identifying contingency effects, case research can be a better method for building explanations for the observed effects, an important requirement for system approaches. Associated data analysis methods such as ‘‘causal networks’’ (Miles and Huberman, 1994) can be especially useful to analyze
networks of causal relationships between the research variables (the system studies by Sousa (2003) and Sousa and Voss (2001) in Table 1 are examples of the application of this method in OM). Finally, the previously mentioned need for the development of taxonomies of contextual variables would also facilitate the use of system fit approaches.
The second main research need is to recognize and combine different approaches to fit, on a journey of cumulative theory building. Different forms of fit are not mutually exclusive and can provide unique and complementary information (Drazin and van de Ven, 1985; Venkatraman, 1989).
We next discuss the insights that the use of each of the three main forms of fit singly and in combination can provide, and give examples of how they have been and/or could be applied in OM PCR. The selection approach can be useful for exploring important relationships between context and OM practices. This information can then be used for the generation of contingency propositions for future tests incorporating the performance dimension (Drazin and van de Ven, 1985).
For example, Rungtusanatham et al.’s (1998) empirical results using a selection approach raised the possibility of national culture affecting workforce management practices. These findings could be used to generate appropriate contingency propositions for future tests incorporating performance.
The interaction approach can be used to identify the most critical context–practice relationships. If the use of such approach detects fit, but only among certain pairs of context-practice relationships, such findings would indicate that those context–practice matches are more relevant predictors of performance than others (Drazin and van de Ven, 1985). Such findings would be of great practical utility, implying that limited resources should be allocated to the most critical context–practice relationships.
For example, Kathuria and Partovi (1999) found that the degree of fit between the context variable emphasis on manufacturing flexibility and two types of HRM practices (relationship-oriented practices and participative leadership and delegation practices), but not a third type of HRM practice (work-oriented practices), had a significant effect on managerial performance. Hence, managers aiming at achieving contextual fit of HRM practices should focus their efforts on the first two practices and might ignore the third.
Whenever the contingency theory in question is based on configurations of variables, it is recommended that interaction results be compared with system results (Drazin and van de Ven, 1985). If the interaction results are not significant, but the system results are, then it can be reasonably concluded that fit does not occur at the level of any individual variable alone but rather at the level of deviation from an overall pattern of variables (i.e., the effects of fit are present at a holistic level). The system approach can also be useful when it is possible that conflicting contingencies are present (e.g., one contextual factor specifies a high use of a practice as the fit and a second contextual factor specifies a low use of the same practice as the fit).
Finally, the system approach could be
- Sousa, C.A. Voss / Journal of Operations Management 26 (2008) 697–713 707 used to address equifinality in OM PCR. Of particular interest would be for future research to ascertain whether there are multiple, equally effective ways of achieving fit between the set of OM practices to adopt and an organization’s context.