towards a taxonomy of search patterns of manufacturing flexibility in small and medium-sized firms

Towards a taxonomy of search patterns of manufacturing¯exibility in small and medium-sized ®rms

Marcello Bragliaa, Alberto Petronib,*aDipartimento di Ingegneria Meccanica, Nucleare e della Produzione, UniversitaÁ di Pisa, Via Bonanno Pisano 25/B, 56126 Pisa, Italy

bDipartimento di Ingegneria Industriale, UniversitaÁ di Parma, Viale delle Scienze, 43100 Parma, Italy

Received 1 November 1998; accepted 1 July 1999

Abstract

Manufacturing organizations are experiencing a wide range of forces which make it necessary for them tocontinuously adapt their operations and consider the consequences of their production activities on customerexpectations, supplier constraints and changes in the law. This paper presents an empirical veri®cation of knowledge

levels and the perceived strategic relevance of manufacturing ¯exibility. A questionnaire survey was conducted witha sample of 89 manufacturing ®rms located in the area around Brescia, one of the most industrialized areas inEurope. Initial results gave a better understanding of which type of ¯exibility is preferred when facing di�erent

environmental challenges. In addition, insights into managerial approaches were brought out through clustering anddiscriminant analysis which have made it possible to emphasize some signi®cant di�erences between ®rms in thesearch patterns of manufacturing ¯exibility. # 2000 Elsevier Science Ltd. All rights reserved.

Keywords: Manufacturing ¯exibility; Operations management; Small and medium-sized enterprises; Mechanical industry

1. Introduction

In general terms, manufacturing ¯exibility has been

de®ned as the ability of a manufacturing organization

to deploy and redeploy its resources e�ectively in re-

sponse to changing environmental and internal con-

ditions [1]. Manufacturing ¯exibility has been

emphasized as a major competitive concern in oper-

ations management [2] and its improvement is fre-

quently an important issue on the manufacturing

manager's agenda [3]. The manufacturing ¯exibility

choices made by ®rms are aimed at improving compe-

titiveness when faced with rapid changes [4,5]. With

globalization the company must consider international

markets and still maintain a certain threshold of ¯exi-

bility to adjust quickly and cost-e�ectively to local

markets requirements, evolving technical standards

and changes in the law [6]. Manufacturing ¯exibility

can also be directed towards speeding up the introduc-

tion of products, which is critical in situations of

increased uncertainty where products have short mar-

ket lives. But even when product innovation capability

is not critical (in established industries where fewer

product changes occur), manufacturers often have to

face considerable environmental uncertainty (due, for

instance, to safety and environmental legislation or

compliance to quality certi®cation requirements).

Despite the vast amount of research carried out on

manufacturing ¯exibility over the last decade, common

de®nitions and shared concepts still remain elusive.

Omega 28 (2000) 195±213

0305-0483/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.

PII: S0305-0483(99 )00044-4

www.elsevier.com/locate/orms

* Corresponding author. Tel.: +39-0521-905-861; fax: +39-

0521-905-705.

E-mail addresses: [email protected] (M. Braglia), petro-

[email protected] (A. Petroni).

Many di�erent classi®cations of the concept are pro-

posed (e.g.[7±10]), and it is clear that approaches tooperational ¯exibility have not yet reached a paradig-matic stage. Furthermore, some empirical evidence

seems to con®rm that, despite the wide proliferation of¯exibility types and de®nitions proposed in the litera-ture, ¯exibility tends to be unidimensional when opera-

tionalized [6]. Even when adopting a speci®ctaxonomy, the basic types of ¯exibility are not unre-

lated [5]: new product introduction and mix ¯exibilityinteract in a fairly strict manner, tend to reinforce eachother and in¯uence common elements in operational

domains such as the possibility of reusing componentsand the involvement of personnel. It also appears tobe the case that mix ¯exibility can reduce the need for

volume ¯exibility by decreasing volume swings [11].The research illustrated below addresses the issue of

whether it is possible to track meaningful uniformityin the search patterns of manufacturing ¯exibilityadopted by small and medium-sized enterprises. The

underlying assumption is that the organizational adap-tive behavior, when faced with external contingencies,

is largely ®rm-speci®c (depending on the ®rm's physicaland managerial structure and organizational resourcesand capabilities) and path-dependent (depending on its

previous choices and past history) [12]. After morethan a decade of research into manufacturing ¯exi-bility, it is now recognized that ¯exibility and the

measures used to express it are user- and/or situation-speci®c [13].

This paper is divided into three parts. The ®rst triesto identify the nature of environmental changes that®rms are currently facing. The second part provides a

measurement of the perceived e�ectiveness of eachdimension of manufacturing ¯exibility when facingmajor environmental challenges. The third section il-

lustrates an attempt to build a taxonomy of small andmedium-sized enterprises (SMEs) based on their

approach to manufacturing ¯exibility. The data referto a sample of 89 SMEs located in the province ofBrescia, one of the most industrialized area of northern

Italy. The basic research model relies on the assump-tion that, by identifying speci®c environmental sourcesof uncertainty and the responses of the ®rms, one can

a priori identify critical dimensions of manufacturing¯exibility.

Previous work in this area, in assuming that man-agers face the task of balancing manufacturing ¯exi-bility and external uncertainty, have generally aimed at

building operational frameworks for the analysis ofmanufacturing ¯exibility, where managers are allowed

to delineate each of the types of ¯exibility with whichthey may be concerned.Both quantitative and qualitative approaches have

been followed when dealing with the measures of man-ufacturing ¯exibility types [14]. Quantitative physical

measures are based on actual production data (relatedto product, machines, processes, etc.). The relevant lit-

erature is extensive. Excellent reviews are presented[9,15,16].Two types of qualitative research can be identi®ed

based on whether perceptual scales are used or not. Inthe former case, major international surveys aimed atmeasuring ¯exibility investment and performance are

presented, for instance, in Refs. [17] and [18]. On theother hand, qualitative studies without perceptualscales have usually been placed in a framework in such

a way that ®rst the dimensions of change and the timehorizon are identi®ed, and then the critical elements of¯exibility are derived [2]. Dynamic equilibrium modelshave been used to visualize changes in external posi-

tioning, assessing the balance between ¯exibility andexternal uncertainty, and establishing long-term andshort-term action plans. These models have been used

together with audit checklists of external uncertaintyand manufacturing ¯exibility [19±21]. The practicalbene®ts of these studies include the provision of a less

ambiguous frame of reference for discussing the var-ious ¯exibilities the ®rm needs to manage.

2. Research framework and methodology

Two major objectives were pursued via the researchoutlined below. First, to gain an understanding of

which dimensions of manufacturing ¯exibility are per-ceived as being most e�ective and, as a consequence,are pursued by successful small and medium-sized®rms. Second, to explore the di�erent patterns of ¯exi-

bility search followed by the ®rms and relate these pat-terns to the characteristics of the ®rms and the natureof the industry where they operate.

The results presented in this paper are derived froma broad research program performed in three mainstages: initial survey design, questionnaire development

and data collection and analysis.

2.1. First stage: initial survey design

The ®rst stage was aimed at the identi®cation ofmajor dimensions of manufacturing ¯exibility and thegeneration of items able to capture the speci®c domain

of interest. An inductive approach was used to identifyitems for inclusion in the scales, and the available lit-erature was used to provide the theoretical foun-

dations.The ®rst step was aimed at splitting the overall con-

struct (manufacturing ¯exibility) into its component

subconstructs (dimensions). The attempt to classifymanufacturing ¯exibility has involved many research-ers (for an exhaustive review, see [15]). The problem

M. Braglia, A. Petroni / Omega 28 (2000) 195±213196

with many of these studies is that they either fail toprovide operational de®nitions of manufacturing ¯exi-

bility or consider the di�erent ¯exibility types at anaggregate level (de®ning a generic property of the man-ufacturing system rather than a speci®c attribute of it).

Since no conventionally accepted de®nition for manu-facturing ¯exibility types has been established, it is im-portant to summarize the operational de®nitions for

each type of ¯exibility.The dimensions of manufacturing ¯exibility con-

sidered for this study rely on a combination of the

manufacturing types as proposed in [8,13], since theseauthors have attempted to provide operational de®-nitions of a wide set of manufacturing ¯exibility types.The de®nitions of the subconstructs that were utilized

in the survey together with their speci®c measures arereported in the copy of the questionnaire which isshown in the appendix.

2.2. Second stage: questionnaire development

In the second stage, a population consisting of suc-

cessful small and medium-sized ®rms operating inmany industries was chosen. As concerns the size ofthe ®rms, the criterion was to include manufacturing

®rms with less than 250 employees and 20 millionsECU of aggregate sales. As for the `successful' attri-bute, the basic criteria for selecting the sample wasthat eligible ®rms should have had performances well

beyond the average industry-speci®c ®gures. Morespeci®cally, there were three basic requirements thathad to be met:

. +50% with respect to the yearly variation pro®tindex for the manufacturing sector;

. +20% with respect to the yearly variation pro-

duction index for the manufacturing sector;. +20% with respect to the yearly variation domestic

orders index for the manufacturing sector.

These cut-o� levels were agreed upon after indi-cations provided by two experts employed by theChamber of Commerce of Brescia. These ®gures were

drawn from the 1998 Joint Report of the samechamber of commerce.A ®nal population of 542 ®rms was obtained from

the database and 95 of these were approached through

personal contacts and following individual knowledgeof the experts. Six of them refused to be involved inthe research project and a ®nal sample of 89 ®rms was

obtained. The ®nal sample represented a wide spec-trum of activities within the mechanical industry (elec-tromechanical, components, machinery-building,

machining jobs), manufacturing practices (assembly,discrete parts manufacturing and machining), pro-duction type (one-of-a-kind, batch and mass pro-

duction) and commercialization type (catalogue, stock,job order).

In addition, in order to provide more complete in-formation for the hypothesis and render the prelimi-nary version of the questionnaire valid, `round table'

meetings were organized with eight CEOs from thesample. This stage proved to be particularly usefulsince it provided a direct insight into how speci®c

phenomena which came to the fore during the litera-ture survey found actual application in the industrialpractice within the context under investigation. This

activity pursued two distinct objectives: building aframework for a structured research hypothesis con-cerning determinants, modes and the e�ects of environ-mental changes in manufacturing ¯exibility practices

and, secondly, to decide upon which data should becollected in a subsequent questionnaire survey. Theresults of this process were edited for overlaps and rep-

etitions to produce a questionnaire with a ®nal list ofmore than 50 items. The structure of the questionnaireconsisted of three parts: the ®rst referred to the collec-

tion of general information, the second was aimed atinvestigating the basic characteristics of the competitivebase and the environmental features in which ®rms are

engaged, and the third concerned the perception of thee�ectiveness of the di�erent dimensions of manufactur-ing ¯exibility (subconstructs) in countering externaland internal stimuli and their relative level within the

®rm. Speci®cally, in the questionnaire, seven types ofmanufacturing ¯exibility were de®ned and describedand examples and explanations were provided to illus-

trate each particular ¯exibility type. A full samplecopy of the questionnaire is reported in the appendix.As for the questionnaire, some methodological notes

should be mentioned. Measurement for both Section 2and 3 of the questionnaire was made using discreteLikert-type scales. The number of points on the Likertscales was considered carefully, since reliability

increases as the number of scale points increases to ®veand continues to increase at a much smaller rate foradditional points above ®ve. Our choice to use 9-point

Likert scales followed Saaty's approach [22] used toestablish hierarchical priorities among perceptual state-ments thus making it possible to carry out further

analysis using the analytic hierarchy process.

2.3. Third stage: data collection and analysis

Finally, based on these two preliminary stages, a

questionnaire survey was presented to executives (pre-dominantly belonging to production functions, butalso including other technical functions) who were

identi®ed as being responsible for the strategic and op-erational management of the manufacturing processes.In the administration of the questionnaire, respondents

M. Braglia, A. Petroni / Omega 28 (2000) 195±213 197

were always assisted by at least one interviewer who,given the prevailing perceptual nature and the purpose

of the research, took the maximum possible care toavoid exerting any in¯uence on the respondent. Thepresence of an individual (who received thorough

training prior to the interview) helped avoid misinter-pretation of questions.Following this design, the study is illustrated in such

a way that, ®rst, a general description of the manufac-turing ¯exibility types considered and a classi®cationof environmental changes are provided. Then, overall

correlation between determinants and responses arepresented and discussed. Finally, a somewhat higher-level perspective will be taken which makes it possibleto pinpoint some signi®cant di�erences between ®rms

in the search patterns of manufacturing ¯exibility.

3. Phenomena observed

After collecting a large quantity of informationabout the ®rm's size, its market and productioncharacteristics, together with descriptions of the main

environmental changes that have occurred over the lastten years, respondents were asked to rate (on a 9-pointscale) the overall relevance and e�ectiveness of each

type of manufacturing ¯exibility for the ®rm as awhole.The ®rst important evidence is that, on average,

®rms look at manufacturing ¯exibility as an importantcompetitive tool (see the mean values of Table 1), andthat, moreover, there is a good knowledge of and

acquaintance with ¯exibility-related concepts. Whatemerges from the survey is that all dimensions of man-ufacturing ¯exibility are being increasingly exploited(except layout ¯exibility) and are becoming a matter of

survival for a host of ®rms even of limited size.As concerns the impact of each speci®c ¯exibility

type on the overall performance of the ®rm (measured

on a 9-point scale), ®gures vary according to the indus-try considered (Table 1). In particular, empirical ®nd-ings suggest that product ¯exibility is widely

acknowledged as an important performance factor inall four subindustries (reaching the highest mean scorefor the electromechanical sector), while machine, rout-ing and process ¯exibility seem to be largely recognized

as critical for machining ®rms as well as for com-ponents manufacturers. Volume and expansion ¯exi-bility are assigned a relatively high consideration by

machinery-building ®rms, while layout ¯exibility stilldoes not appear to draw much attention and concernfrom operations managers.

As far as the strategic use of manufacturing ¯exi-bility is concerned, even if beyond the scope of thisstudy, some ®ndings should be mentioned.

Among the major external and internal changes

which require a response in terms of manufacturing

¯exibility (Table 2), one should note those induced by

demand ¯uctuations (10.7% out of the overall number

of changes mentioned), customer satisfaction (12.4%),

competitors (10.5%), suppliers (6.8%) and institutions

(regulatory bodies and unions) (8.4%).

As far as the relationships between organizational

factors and manufacturing ¯exibility are concerned,

intangible factors, such as high involvement of workers

in problem-solving activities, close relationships with

suppliers, and ¯exible wage schemes, are associated

with greater process, volume, and product ¯exibility.

Finally, the ®eld analysis has also provided some

insights into the relationships between ¯exibility types,

and between them and internal uncertainty.

. Internal uncertainty is partly a result of external

uncertainty.

. More automation is associated empirically with less

¯exibility.

Table 1

Perceived strategic e�ectiveness of manufacturing ¯exibility

types

Industry N Mean Std. Deviation

Elettromechanical Machine 13 4.46 3.20

Routing 13 3.69 3.45

Process 13 6.85 2.76

Product 13 8.69 .63

Volume 13 7.08 2.53

Expansion 13 5.46 2.70

Layout 13 3.08 2.53

Components Machine 29 7.86 1.81

Routing 29 7.03 2.26

Process 29 6.31 2.78

Product 29 7.10 2.21

Volume 29 5.90 2.82


Layout 29 3.76 2.64

Machining jobs Machine 10 8.70 .67

Routing 10 5.70 3.02

Process 10 7.50 1.65

Product 10 5.70 3.50

Volume 10 5.20 3.16


Layout 10 3.60 2.50

Machinery-building Machine 34 5.47 3.35

Routing 34 3.12 2.58

Process 34 4.79 2.83

Product 34 7.53 1.74

Volume 34 7.21 2.33


Layout 34 2.68 2.21


. Process and product ¯exibility are mutually reinfor-

cing and tend to be supported by similar factors.

. Process ¯exibility may reduce volume ¯uctuations

which, in theory, could reduce the need for volume

¯exibility.

. The possibility of reusing a component is signi®-

cantly correlated with high product mix require-

ments and product ¯exibility.

. Production reorganization prefers process reengi-

neering rather than layout redesign.

. Machine ¯exibility is shown to have a moderating

e�ect on process ¯exibility, but one which does not

necessarily cancel out the bene®ts of process ¯exibility.

4. Manufacturing ¯exibility: towards a taxonomy of

®rms

The third aim of the study, as pointed out earlier,was that of carrying out some meaningful classi®cationof ®rms based on their perception of the relevance and

Table 2

Environmental changes

Changesa Frequency Percent

Increased demand 34 5.5

Competitors' technological progress 32 5.2

Safety legislation 28 4.5

Increased product quality demanded by customers 22 3.6

Reduced delivery times demanded by customers 21 3.4

Job devision according to ISO norms 19 3.1

Increased product range demanded by customers 19 3.1

Lack of personnel competencies 18 2.9

Meaningful product/process innovation 18 2.9

Price competition 15 2.4

Product customization 14 2.3

Enhancement of controlling, testing and inspection systems 14 2.3

ISO registration 13 2.1

Reorganization of production 13 2.1

New production planning and control techniques 12 1.9

Re-organization of logistics 12 1.9

Increase of non-shop¯oor personnel (quality, inspection) 11 1.8

Reorganization of sales and post-sales 10 1.6

Market globalization 10 1.6

Demand reduction 9 1.5

Suppliers selection (reliability analysis) 9 1.5

Externalization of production phases 8 1.3

New materials/components 8 1.3

Impact on physical structures of quality program 8 1.3

Increase of customer number 7 1.1

Change of product mix 7 1.1

Stock reduction 7 1.1

Price ¯uctuations of raw materials 7 1.1

Suppliers collaboration/partnership 7 1.1

Reduction of delivery times from suppliers 6 1.0

Demand seasonality 6 1.0

Increase of production turns 6 1.0

Increase of shop-¯oor personnel 6 1.0

Tax and credit inducements for investment purposes 6 1.0

Smaller batches and more frequent delivery 5 .8

Union pressure 5 .8

Increase of suppliers number 5 .8

a Changes with more than ®ve counts have been included.


e�ectiveness of the various types of manufacturing

¯exibility. With this purpose in mind, hierarchical clus-

tering analysis was conducted in order to identify the

critical features of homogeneous ®rms and intergroup

di�erences1. Distances among clusters were originally

measured using di�erent methodologies in order to

select the most appropriate one. Only two methods

(Ward's method and within-group linkages) proved to

be particularly e�ective for di�erentiation. Finally, the

most suitable method was found to minimize the incre-

ment of internal inertia within the clusters by Eucli-

dean distances. In choosing the number of clusters a

set of criteria were carefully considered: the analysis of

the rescaled distances in the dendrogram and of the

icicle (which are not reported in the interest of brevity)

and the impact of cluster con®gurations on the succes-

sive discriminant analysis. More speci®cally, the var-

iance explained by the (n ÿ 2) discriminant functions

were considered, where n is the number of clusters cho-

sen. With four clusters the ®rst two discriminant func-

tions explained 93.0% of variance, with ®ve clusters

three discriminant functions explained 95.6% of the

variance and with six clusters four discriminant func-

tions explained 91.8% of variance. Thus ®ve clustersof a homogeneous size were identi®ed (the number of®rms included in each cluster ranges from 16 to 20).

5. Description of clusters

Once each observation was ®tted into a cluster, the

complex phase of interpretation was carried out.Firstly in a more descriptive manner, and, then, morethoroughly through discriminant analysis. Fig. 1 illus-

trates the mean values of ¯exibility types as reportedby each cluster.The two horizontal lines have been drawn to indi-

cate the three value ranges:

(a) less than 3.5: low relevance;(b) 3.5 to 6.5: medium relevance;(c) more than 3.5: high relevance.

5.1. Cluster 1: ®rms responsive to all dimensions ofmanufacturing ¯exibility

Firms in this cluster are sensitive to all aspects ofmanufacturing ¯exibility, except for process and layout

¯exibility, and present the highest standard deviations.It should be emphasized that almost all cluster 1 ®rmshave attained UNI/EN 9000 quality certi®cation. This

indicates that compliance to normative regulationsoften implies, at least for more structured organiz-ations, a greater attention to managerial and organiz-

Fig. 1. Perceived e�ectiveness of ¯exibility types.

1 The unit of analysis was, then, the ®rm, which has been

distinguished according to the relevance granted to the seven

types of ¯exibility which have been outlined above. The i-th

observation corresponds to an ordered vector of the seven rel-

evance values (perceptual scores). Hierarchical cluster analysis

was used since this methodology is to be preferred when no

preliminary cluster de®nition and knowledge is attained.


ational aspects and, consequently, an increasing aware-

ness of competitive challenges such as ¯exibility.The cluster can be further split into two distinct sub-

groups.

The ®rst, consisting of component-manufacturersworking on a job order or production-for-inventory

basis, almost totally ignores process ¯exibility. Despitethe fact that their average size is highly variable, a sec-ond common feature is that they tend to establish

long-lasting relationships with many customers andsuppliers. In addition, such ®rms operate in an en-vironment which is de®ned as highly turbulent (in

terms of competitive pressure) but with medium levelsof uncertainty (the proxy for this variable is the degree

of demand predictability). Correlation analysis con-®rmed that, for this group of ®rms, the importancegranted to machine and routing ¯exibility increases as

annual sales variation increases. There also appears tobe a strong correlation between the degree of salesinternationalization and the perceived relevance of pro-

duct ¯exibility. This may be interpreted as the con®r-mation that innovation capabilities are acknowledged

as a critical condition for accessing new internationalmarkets. Summing up, ®rms of this group are ¯exi-bility-oriented, but production is largely product-

specialized and thus little e�ort is generally devoted tosigni®cantly modifying the production mix. The largenumber of customers together with the prevailing pro-

duction-for-inventory perspective adopted may par-tially justify a manufacturing strategy aimed at

curtailing customer contractual power and the ®rm'sdegree of dependence on customer expectations.A second group of cluster 1 ®rms is particularly

keen on process ¯exibility. This group is less homo-geneous than the ®rst group but, nonetheless, presents

some obvious common features. The ®rms are mainlymedium-sized machinery-manufacturers, but, unlikethe previous group, production is exclusively of the

one-of-a-kind type. Products are generally of highercomplexity and the competitive environment is verytough. These companies operate on the global market

and tender-based competition means that their evalu-ation by clients is not only based on the quality of the

®nal product (which is generally customized) but alsoupon the process followed to develop it. This meansthat ®rms in this group generally establish closer re-

lationships with customers and take care to continuallyupdate their production mix (this variable has beenproxied by the nature of demand change). It is also

interesting to note that these ®rms often reportdemand and customer satisfaction-related changes, but

seldom a meaningful product/process innovation. Sum-ming up, it is clear that the competitive environment ischaracterized to a signi®cant extent by incremental

rather than radical technological adjustments, requir-ing rapid responses.

5.2. Cluster 2: machine ¯exibility-oriented ®rms

Firms belonging to cluster 2 distinguish themselves

primarily for the importance given to machine ¯exi-bility, which, in fact, presents the lowest standard devi-ation. Another feature is the negligible weight assigned

to routing ¯exibility. Firms are mainly of small sizeand declare that they operate in a very tough competi-tive environment (average score: 8.07). It was ascer-

tained that most of ®rms have made substantialreplacement investments, mainly aimed at equippingtheir machinery with multipurpose tools. Furthermore,not a single ®rm of cluster 2 reported any changes re-

lated to production reorganization. This may be due toboth their reduced size and production saturation con-ditions. The former aspect implies that scarce attention

is given to the possibility of process and layout ration-alization, since it is clearly impossible for these ®rmsto establish appropriate bottleneck redundancies and

duplications of operation assignments. The latteraspect, which is due to the need to have a full returnon the investment, reduces the possibility of placingdi�erent jobs on di�erent machines (which are usually

already overloaded).Cluster 2 can also be subdivided into two groups.

The ®rst includes ®rms reporting high machine and

volume ¯exibility and low routing and expansion ¯exi-bility. This is in keeping with a situation wherebyannual sales ¯uctuations are not particularly signi®cant

(hence the low relevance of expansion ¯exibility), eventhough there are considerable monthly swings of bothvolume and production mix. Among the ®rms which

are machine ¯exibility-inclined one ®nds those thatconsider technology as the most important lever whenfacing a rather unstable market.The second group of ®rms all nominated, in addition

to machine ¯exibility, both product and expansion¯exibility as top priorities. Generally these organiz-ations operate on an international scenario and their

production is either of the one-of-a-kind type or cata-logue order-to-assembly type. It is worth noting thatthe vast majority of these ®rms are machinery manu-

facturers. New product development capability is criti-cal for international competition and is favored by thefact that the machinery set-up times and costs are rela-tively low. It is interesting to observe that the percep-

tion of the importance of process ¯exibility rises as thenumber of customers increases. This is because the suc-cess of these ®rms is strictly dependent on meeting cus-

tomers' technical speci®cations: the higher the latterare, the greater the product versatility.

5.3. Cluster 3: Innovation-oriented ®rms

Taking a closer look at Fig. 1, product ¯exibility


emerges as the factor which characterizes cluster 3®rms.

Fig. 3, however, shows the speci®c structure for thiscluster, consisting of a relatively homogeneous set ofcases close to its centroid and of a more decentralized

scatter plot of cases. Restricting the analysis to the for-mer cases, the resulting mean values show that bothvolume and expansion ¯exibility are the most import-

ant dimensions.The majority of these ®rms belong to the electrome-

chanical and machinery-building industries (all those

not included in cluster 2) and have to face major chal-lenges related to the ability to rede®ne their productranger over time and carry out product innovations(average score: 7.2). The picture that emerges is of

extremely dynamic ®rms with rapid and continuouslyincreasing growth. It is signi®cant that most of these®rms state that they have introduced radical organiz-

ational changes (leaner structures, functional redesign,job division and process re-engineering) in order tomore e�ectively adjust to growth. At the core of their

distinctive competencies lie problem-solving capabilitiesrather than the distinctiveness of the manufacturingprocess itself. The reason why machine ¯exibility is

almost totally neglected by these ®rms can be tracedback to the prevailing labor-intensive assembly natureof their operations (only conventional tools are used).This circumstance is indirectly supported by the fact

that not a single ®rm stated that it had performed mas-sive investments in machinery or technical equipmentto date. Also routing ¯exibility is given a low prefer-

ence rating and this is due to di�culty in changing thelocation of assembly stations. In this case, the absenceof large plants is an element of manufacturing versati-

lity enabling the ®rms to switch rapidly from one pro-duct type to another without jeopardizing innovationperformances. This is con®rmed by the nature of themore frequently mentioned environmental stimuli,

which relate to product, process or componentimprovements by either competitors or suppliers.

5.4. Cluster 4: customer-oriented ®rms

The most important discriminating variables of clus-

ter 4 are assigned to product as well as to process ¯exi-bility. Volume ¯exibility is important too but with ahigher standard deviation. The analysis of the competi-

tive environment for these ®rms indicates that satisfy-ing the ever changing requirements and expectations of

customers is a critical factor. A more detailed analysisshows that the cluster is fairly heterogeneous for indus-

try, production and sale typology. A feature that iscommon to all ®rms is the constant concern for inno-vation both in manufacturing processes and the func-

tional and structural characteristics of the product. A®rst group of ®rms can be seen which distinguishesitself by giving considerable importance to volume

¯exibility and to routing and layout ¯exibility. There isa strong positive correlation between product ¯exibilityand the frequency of process/product innovation (as

measured on the same perceptual scale). It should alsobe noted that the perceived suitability of expansion¯exibility increases as competitive pressure and salesvariability (demand uncertainty) increase. This con-

®rms the idea that speci®c attention is paid to techno-logical progress and growth dynamics.

5.5. Cluster 5: ®rms focusing on product range changes

Although cluster 5 is the largest, its dispersionmeasure from the centroid is the lowest one. Firms in

cluster 5 tend to give priority to the elements of ¯exi-bility which have major internal rather than externalimpact, particularly machine and routing ¯exibility.

The facility to periodically revise the production mix isa peculiar element identifying this cluster: machine androuting ¯exibility are perceived as e�ective ways to

gain a competitive advantage, associated with the abil-ity to produce, but without having to modify pro-ductions set-ups. Almost all of the ®rms in cluster 5

®rms operate in the machining and components-build-ing sectors. Production is of the one-of-a-kind typewhich aims at satisfying a job order or catalogue-baseddemand. This cluster consists of ®rms and smaller ma-

chine workshops equipped with highly versatile, inter-changeable and programmable work stations. Themajor strengths of these ®rms lie in their technical abil-

ity to manufacture and process di�erent part typesaccording to customer speci®cations. A great deal ofe�ort is devoted to satisfying the heterogeneity of cus-

tomer orders. These conclusions are supported by theobservation that almost all cluster 5 ®rms mentionedmore internal changes (directly related to the manufac-turing system) than environmental stimuli, and that all

of the ®rms have made huge investments in technol-ogy.In the research stage that followed, the aim was to

gain a more detailed interpretation and description ofthe clusters, by identifying what are the most discrimi-nating factors between clusters. Discriminant analysis

was conducted with a twofold purpose: to interpretinterclusters di�erences and to classify cases withineach cluster2.

2 The units of analysis cases are the 89 ®rms, each of which

is characterized by seven variables (predictor variables) corre-

sponding to the seven types of manufacturing ¯exibility. The

number of discriminant functions is 4 (one less than the num-

ber of clusters).


6. Interpreting interclusters di�erences

A study of the way clusters di�er was conducted in

order to (i) discriminate between groups using a

speci®c set of characteristics, (ii) establish to what

extent actual di�erences are included and (iii) decide

which discriminant functions are most e�ective.

A four-dimension graphic representation of the situ-

ation would be impossible.

Eigenvalues and cumulative function discriminant

power were, therefore, used to select the more repre-

sentative functions (Table 3).

As illustrated, function 1 and 2 count for almost

87% of the overall discriminant power. If function 4 is

also considered, nearly all of the di�erences between

the clusters can be explained (more than 95%).

In this case, a two-dimension chart is su�cient to

reveal meaningful information (Fig. 2).

A closer look at Fig. 2 reveals that along function 1

(the most discriminating function) cluster 5 is posi-

tioned at the right end of the horizontal axis while

cluster 3 is at the opposite end. The cluster 3 centroid

is 3.5 standard deviations o� the axis origin (i.e. the

mean of the whole sample) while the cluster 5 centroid

is positioned almost 3.5 standard deviations o� the

overall centroid. Both clusters are, therefore, quite dis-

tant each other. A closer look at the di�erences

between cluster 5 and cluster 3 may make it possible

to learn more about the meaning of function 1.

As far as function 2 is concerned, it is evident that

in the vertical direction all clusters are positioned very

close to the origin. Their distances from the origin is

less than 1 standard deviation, except for cluster 2

which has an abnormal behavior because it is posi-

tioned more than two standard deviations o� the over-

all centroid. Further investigation would be helpful in

order to understand what determines this peculiarity.

The third row of Table 7 suggests that, along function

3, cluster 1 presents an exceptional behavior (even

though to a lesser extent than in the former cases, its

Table 3

Eigenvalues and discriminant power of functions

Canonical discriminant function Eigenvalue Relative percentage Cumulative percentage

1 5.380 69.7 69.7

2 1.311 17.0 86.7

3 0.686 8.9 95.6

4 0.338 4.4 100.0

Fig. 2. Representation of clusters.


distance from the origin being only 1.5 standard devi-

ations). The deviation of function 4 is even less accen-tuated.

7. The discriminant functions

In order to determine the degree of similarity

between each single variable and each discriminantfunction, product±moment correlation between thetwo (total structure coe�cients3) were calculated

(Table 4).

7.1. Function 1: e�ciency of plants

Table 4 shows a strong positive correlation (84%)between discriminant function 1 and the perceived im-portance of machine ¯exibility and routing ¯exibility

(approximately 70%). This indicates that these two¯exibility dimensions are the major di�erentiating fac-tors with regard to the ®rst discriminating function.

The joint analysis of the position of centroids alongthe axis, the values of Table 4 and the evidence of Fig.2, produces the pooled values of the ®ve clusters for

the two dimensions, and these are reported in Table 5.By comparing the two clusters positioned at the

extremities of the horizontal axis of Fig. 2 it can beconcluded that in cluster 3 both dimensions receive

low scores, while in cluster 5 the importance assignedto them is far higher. The analysis of the mean scoresof ¯exibility con®rm such a trend (last row of Table

5): positive values of discriminant function 1 revealthat substantial importance is given to machine androuting ¯exibility (clusters 1 and 5). Values close to

zero indicate nonsigni®cant perceived importance (clus-

ter 2) and negative values mean low levels of import-ance (cluster 4 and 3).In order to characterize the discriminant function, it

may be useful to have another look at the descriptionsof clusters 3 and 5. The former consists of ®rms with alow level of advanced manufacturing technologies andmachinery and they focus their attention on satisfying

customer expectations. The latter, on the other hand,contains those ®rms which aim at bringing togetherhigh levels of productivity and ¯exibility within an

almost ®xed range of products (generally mechanicalcomponents). As a consequence, we described discrimi-nant function 1 as `e�ciency of plants'.

7.2. Function2: organizational and managerialdevelopment

The strongest correlations for function 2 are again

with machine and routing ¯exibility, though values aresigni®cantly lower than for the previous case. Here,however, the routing ¯exibility coe�cient is negative.

The algebraic sum of the two ¯exibility dimensions hasbeen tabulated for function 2 (Table 6).In general terms, machine ¯exibility is recognized as

being a dimension which is a little more important

than routing ¯exibility except for cluster 2 ®rms which,as noted before, behave in a very di�erent fashion.This di�erence in behavior may be explained by the

relative small size of the manufacturing systems ofcluster 2 ®rms (see the last row of Table 7). It is worthnoting that the remaining four clusters can be thought

of as homogeneous when it comes to ®rm size. Eventhough their number of employees range between 62and 90, these ®rms have all completed the evolutionfrom craftsman type structures to more developed and

complex organizational modes and designs. The samecannot be stated for cluster 2 ®rms.In synthesis, discriminant function 2 can be

explained in terms of elevated perceived importance ofmachinery ¯exibility and rigidity of production partsrouting. Such a situation is found particularly in ®rms

which are still in a phase of outdated organizationaland managerial development.The level of organizational development has a strong

Table 4

Total structure coe�cients

Variable

Function

1 2 3 4

Machine ¯exibility 0.839 0.470 0.060 0.044

Routing ¯exibility 0.692 ÿ0.561 0.387 0.053

Process ¯exibility 0.545 ÿ0.311 ÿ0.281 0.361

Product ¯exibility ÿ0.538 0.123 0.513 0.498

Volume ¯exibility ÿ0.100 0.365 0.369 0.068

Expansion ¯exibility ÿ0.154 0.152 0.817 ÿ0.444Layout ¯exibility 0.257 ÿ0.251 0.324 0.391

Table 5

Tabulation of clusters based on discriminant function 1

Cluster

3 4 2 1 5

Function 1 ÿ3.544 ÿ0.615 0.073 1.439 2.896

Machine ¯exibility 2.4 5.06 8.65 8 8.63

Routing ¯exibility 2.25 4.94 2 7.88 7.47

Mean (machine, routing) 2.325 5 5.325 7.94 8.05

3 As correlations, they can be considered as the cosines for

the angles formed by the variables and the function. Thus, by

knowing these coe�cients, we know the geometric structure

of the data space.


impact on the ¯exibility features of an organization.More speci®cally, if we accept the perspective of manu-

facturing ¯exibility as being a set of capabilities, the re-lated choices, in order to be e�ective, have to beembodied in `action-producing' forms, in other wordsin speci®c physical assets, managerial capabilities, indi-

viduals skills and social and cultural values. The ®rstdimension deals with the apparatus of physical assetswhose combination produces a `¯exible manufacturing

system' (FMS): software (for production planning andcontrol, CAD, CAM, etc.) and hardware (facilities,material handling devices, etc.). The dimension related

to managerial system represents the formal and infor-mal organizational mechanisms through which knowl-edge and capabilities are created (collective learning

programs, partnership, ®rm±suppliers/customers re-lationships), codi®ed and controlled (i.e. rewards andincentives, information services, manpower mobility).The last dimension is associated with the values

assigned, within the organization considered as a socialsystem, to the knowledge base, its structure and themeans for its acquisition, accumulation and control: it

is, in other words, an embodiment of the organiz-ational culture which consequently becomes a consti-tuting element of ¯exibility and, as seen, a

discriminating factor between ®rms.

7.3. Function 3: orientation to innovation

Even though it presents little discriminatory power,

discriminant function 3 deserves some consideration.As from Table 7, it is strictly related to expansion

¯exibility (82%) and, to a lower extent, to product¯exibility (51%). Table 7 reports the tabulation of

clusters based on canonical discriminant function 3.Even though the trend of values assumed by each

¯exibility type does not match perfectly with discrimi-nant values, their mathematical means do so. In con-

sideration of the meaning given to expansion andproduct ¯exibility, the canonical discriminant function3 may be described in terms of a ®rm's concern for in-

novation, both in terms of time-to-market and productquality (as perceived by customers), and the ability to

adapt and adjust manufacturing resources to such in-novations.

Discriminant analysis was therefore used to explorethe di�erences among shared patterns of the preference

structures of ®rms classi®ed in distinct clusters. Inorder to analyse the degree of reliability of the ®ndings

and ensure that misinterpretation of the questions wasavoided, the results of the analysis were partially con-veyed to the participants in the study. The major aim

was to investigate whether or not the participating®rms agreed with how they were perceived in the

study. For each cluster two ®rms were contacted ran-domly and given a report consisting of ®ve statements

that were designed to capture the essential picture ofthe ®rm (one item for the competitive base, one for the

nature of the changes, one for each of the three discri-minant functions). The degree of agreement with eachstatement was measured on 5-point ordinal scales, (1

Table 6

Tabulation of clusters based on function 2

Cluster

4 5 3 1 2

Function 2 ÿ0.860 ÿ0.668 ÿ0.425 ÿ0.188 2.244

Machine ¯exibility 5.06 8.63 2.4 8 8.65

Routing ¯exibility 4.94 7.47 2.25 7.88 2

Machine ¯exibilityÿrouting ¯exibility 0.12 1.16 0.15 0.12 6.65

Average number of employees 90 62 62 66 35

Table 7

Tabulation of clusters based on function 3

Cluster

5 4 2 3 1

Function 3 ÿ0.699 ÿ0.593 ÿ0.300 0.071 1.555

Product ¯exibility 4.74 7.94 7.53 8.5 8.12

Expansion ¯exibility 4.21 3.25 5.53 6.7 7.82

Mean (product, expansion) 4.475 5.595 6.53 7.6 7.97


for `very low agreement', 5 for `very high agreement').The mean scores for the ten ®rms are:

®rm 1 4.0®rm 2 4.2

®rm 3 4.0®rm 4 4.2®rm 5 4.4

®rm 6 3.4®rm 7 4.4®rm 8 4.0

®rm 9 4.0®rm 10 4.0

The results are quite satisfactorily in all cases exceptfor ®rm 6 of cluster 3 which has reported a low level

of agreement.

8. Classifying cases within each cluster

The successive goal was to get a deeper understand-

ing of how discriminant functions are related to thevariables within the groups. This information wasobtained from the pooled within group correlations.

By analysing the within structure coe�cients (whichare not reported in the interests of brevity), it can benoticed that they are smaller than the total structure

coe�cients but the rankings from the largest absolutemagnitude to the smallest are similar in both cases(although not identical). For example, routing ¯exi-

bility in function 3 has a total coe�cient of 0.387 anda within coe�cient of 0.5. This is a typical result butnot a necessary condition. These two sets of coe�-cients report on di�erent kinds of structures, so we

should not expect them to lead to the same interpret-ation.For the ®rst two canonical discriminant functions,

however, relations between variables (particularly ma-chine and routing ¯exibility) almost re¯ect those pin-pointed by total structure coe�cients. The

interpretations of discriminant functions sketched pre-viously are, therefore, applicable to within groupsanalysis. In the following paragraph the behavior ofthe ®ve clusters is depicted more analytically.

As far as cluster 1 is concerned, one should note itslimited dispersion (Fig. 3) if few outliers positionedalong the vertical axis (organizational development)

are not considered.Cluster 2 is rather compact, as can be seen in Fig. 3,

especially along function 2. One must not be deceived

from the mean values of machine and routing ¯exi-bility which are due to the combination of high valuesof the former and low values of the latter.

In a sense, ®rms belonging to cluster 2 rank betweenthose that are highly mechanised and those that are

mainly labor-intensive, since they can count on theavailability of advanced manufacturing technologiesand machinery, which are, because of their limited

number, permanently overloaded.Also cluster 3 is fairly condensed, despite the evi-

dence from Euclidean distances analysis. The centroid

is excessively right-shifted due to the presence of out-liers: this implies that the behavior of the most repre-sentative cases is actually even more distant from that

exempli®ed by the centroid (this fact was supported bythe descriptive analysis). It is important to note thatthis cluster has been described as `innovation-oriented'according to both descriptive and discriminant analy-

sis: cluster 3 assumes high but not top values withregard to discriminant function 3 (orientation to inno-vation ) with cluster 1 scoring approximately twice as

much. The reason for this is to be found in the factthat the innovative approach of these ®rms is only par-tial since they seem to privilege demand±pull adjust-

ments of operations (mainly product-related), in otherwords those which are directly perceived by customers,rather than internal ones.

Firms belonging to cluster 4, on the other hand,may be considered to be in an intermediate positionwith regard to function 1 and function 2 since theyshow a massive dispersion around the origin.

Finally, the behavior of cluster 5 may be bestcharacterised in view of the properties of discriminantfunction 1.

9. Summary of results

This study has helped to get a better insight into the

patterns of manufacturing ¯exibility in SMEs. Havinggiven a general description of changes which haveoccurred in the external environment of companies

and the ¯exibility responses they have used to counterthese, it is now possible to take a higher-level perspec-tive and try to sum up the determinants for the di�er-

ences in the behavior of ®rms. First, the pro®le of thecompanies in the sample has been discussed on thebasis of their original membership to di�erent groups.Each group was characterised by a unique combi-

nation of manufacturing ¯exibility levers which wererelated to the basic characteristics of the competitivebase and the environment in which the ®rms are

engaged. These ®ndings have, in other words, sup-ported the hypothesis that ®rms behave in a ®rm-speci®c and situation-speci®c fashion when combining

resources and distinctive capabilities to provide e�ec-tive responses to environmental stimuli. The ®ndingsobtained when the determinants for such di�erent


Fig.3.Clusters'sdispersion.


behaviors were investigated via discriminant analysiswere much more general. An important ®nding is that

the strongest determinant is operations e�ciency. Itseems as if most of the companies in the industriesconsidered are mainly concerned with reaching cost

di�erentials which cannot be found in workforce ¯exi-bility due to the peculiar characteristics of the Italianlabor market. A further determinant refers to the

degree of maturity of ®rms in terms of managerialcompetence and organisational development. This factsuggests the existence of low-structured `interstitial'

®rms that are probably spin-o�s from major compa-nies and tend to exploit commercial ¯exibility andmarket prompt adaptation rather than operations ¯exi-bility on the shop ¯oor.

The relevance of product innovation search in deter-mining the di�erences in the patterns of manufacturing¯exibility is very modest, and this was rather a sur-

prise. This may be partially due to the fact the vastmajority of the ®rms in the sample are componentsmanufacturers supplying a few customers. Tradition-

ally many of them (except for ®rms in cluster 5) didnot even have, until recently, true product developmentresponsibilities. They usually received detailed draw-

ings from the customers' designers and their main con-cern was to manage shop ¯oor operations.

10. Conclusions

Manufacturing ®rms world-wide are investing in¯exibility to e�ciently and e�ectively adapt to external

forces. Manufacturing ¯exibility allows ®rms to quicklyput into use novel technologies, counter new consumerdemands and deal with legislative and social climates

that may di�er in particular industrial and geographi-cal areas.The perception of the increasing importance of man-

ufacturing ¯exibility and its actual use has focused

attention on the need for measures which indicate towhat extent the related concepts are acknowledged inan organizational context.

The roots of manufacturing ¯exibility performancein an organization is based on its capacity to respondto environmental changes. The environment in which

an organization must operate is fraught with uncer-

tainty and turbulence. The sources of uncertainty and

turbulence are related to changes in customers' require-

ments and tastes, increased competition, technological

change and socio-economic issues. Some of the turbu-

lence is also due to internal factors as ®rm-speci®c

resources, organizational capabilities change and

evolve over time.

In this paper the capacity of an organization to con-

sistently achieve high performance levels has been

linked to its ability to anticipate or respond to such

changes which will centre, in the manufacturing

domain, on the ®rm's ability to modify its use of tangi-

ble a well as intangible resources.

This paper illustrates an attempt to address the issue

of building a taxonomy of ®rm behaviors in terms of

manufacturing ¯exibility choices. For this reason a

questionnaire survey was conducted on 89 small and

medium-sized ®rms which had a threefold aim: investi-

gating the prevailing nature of environmental changes

which require suitable responses, perceptually measur-

ing the relative appropriateness of some opportunely

operationalized manufacturing ¯exibility concepts and,

®nally, building a taxonomy of ®rms which could pro-

vide some indirect insights into the ®rm-speci®c search

patterns for manufacturing ¯exibility.

From the analysis two factors emerge as the most

discriminating among the organizational approaches to

manufacturing ¯exibility. These are, the importance

given to machine as well as to routing ¯exibility and,

to a lesser extent, the perceived relevance of expansion

and product ¯exibility.

Discriminant analysis has helped to re®ne these con-

clusions, bringing to some further factors. The two

most discriminating functions are represented by the

e�ciency of plant operations and the organizational

and managerial development of a ®rm. It follows that

the analysis emphasizes peculiar relationships; on the

one hand, between manufacturing ¯exibility and

machining productivity and, on the other hand,

between ¯exibility and innovation (both technological

and organizational). This leads to the conclusion that

exogenous determinants of ¯exibility, adoption of

innovative technologies, organizational development

and managerial practices interact in a fairly complex

fashion and adapt themselves to a changing competi-

tive environment.


Appendix A


References

[1] Gerwin D. Manufacturing ¯exibility: a strategic perspec-

tive. Manage Sci 1993;39:395±410.

[2] Swamidass PM, Newell WT. Manufacturing strategy, en-

vironmental uncertainty and performance: a path ana-

lytic model. Manage Sci 1987;EM-34:85±91.

[3] Slack N. Flexibility as a manufacturing objective. Int J

Operations Production Manage 1983;3:4±13.

[4] Das SK. The measurement of ¯exibility in manufacturing

systems. Int J Flexible Manufacturing Sys 1996;8:67±93.

[5] Suarez FF, Cusumano MA, Fine CH. An empirical

study of ¯exibility in manufacturing. Sloan Manage Rev

1995;37:25±32.

[6] McGraith, Hoole. Manufacturing's new economies of

scale. Harv Business Rev 1992;70:94±102.

[7] Boyer KK, Leong GK. Manufacturing ¯exibility at the

plant level. Omega 1996;24:495±510.

[8] Browne J, Dubois D, Rathmill K, Sethi SP, Stecke KE.

Classi®cation of ¯exible manufacturing systems. The

FMS Mag 1984;2:114±7.

[9] Sarker B, Krishnamurthy K, Kuthethur SG. A survey

and critical review of ¯exibility measures in manufactur-

ing systems. Production Planning Control 1994;5:512±23.

[10] Slack N. The ¯exibility of manufacturing systems. Int J

Operations Production Manage 1987;7:35±45.

[11] Cusumano MA, Fine CH, Suarez FF. An empirical

study of manufacturing ¯exibility in printed circuit board

assembly. Operations Res 1996;44:223±40.

[12] Nelson RR, Winter SG. In: Nelson RR, Winter SG, edi-

tors. An evolutionary theory of economic change.

Cambridge: Harvard University Press, 1982.

[13] Gupta D, Buzacott JA. A framework for understanding

¯exibility of manufacturing systems. J Manufacturing

Sys 1989;8:89±97.

[14] Brill PH, Mandelbaum M. On measures of ¯exibility in

manufacturing systems. Int J Production Res

1989;27:747±56.

[15] Gupta YP, Goyal S. Flexibility of manufacturing sys-

tems: concepts and measurements. Eur J Operational Res

1989;43:119±35.

[16] Sethi AK, Sehti SP. Flexibility in manufacturing: a sur-

vey. Int J Flexible Manufacturing Sys 1990;2:289±328.

[17] Whyrback DC, Wastag G, editors. Global manufacturing

practices. New York: Elsevier, 1993.

[18] Manufacturing features survey. Boston Manufacturing

Roundtable, Boston University School of Management,

1994.

[19] Abdel-Malek L. A methodology to design ¯exible pro-

duction solutions. In: Tenth International Working

Seminar on Production Economics. Igls/Insbruck,

Austria, February. 1998.

[20] Newman WR, Hanna M, Ma�ei MJ. Dealing with the

uncertainties of manufacturing: ¯exibility, bu�ers and in-

tegration. Int J Operations Production Manage

1993;13:19±34.

[21] Upton DM. The management of manufacturing ¯exi-

bility. California Manage Rev 1994;36:72±89.

[22] Saaty P. Fundamentals of decision making and priority

theory with the analytic hierarchy process. Pittsburgh,

PA: RWS Publication, 1994.


towards a taxonomy of search patterns of manufacturing flexibility in small and medium-sized firms

Documents