Weeek-5
int. j. prod. res., 1999, vol. 37, no. 6, 1403± 1426
An empirical examination of quality tool deploy ment patterns and their
impact on performance
R . HA NDFIELD² *, J. JA Y A R A M ³ and S. GHOSH§
A lthough research suggests that quality management initiatives often fail to meet managers’ expectations, few studies consider that an inappropriate choice of quality tools may adversely a� ect the results. This paper analyses the pattern of quality tool deployment and its impact on performance using a sample of 313 North American and European ® rms. The analysis reveals that four primary types of quality tool applications occur: Human R esource (HR ) tools, Design tools, Discipline tools and Measurement tools. Several signi® cant relationships between these dimensions and quality performance were found, suggesting that successful tool deployment of ten depends on competitive conditions and internal strategies.
1. Introduction
A recent academic study f ound that very f ew quality initiatives have consistent and universal e� ects on ® rm perf ormance (A merican Quality Foundation and Ernest & Y oung 1992). Several reasons have been o� ered by researchers to explain this paradox. One group of researchers suggests that indices of ® rm perf orm ance such as market share, return on assets, etc., are di� cult to measure especially in diversi® ed organizations ( Pennings 1984, Kaplan and Norton 1992). Consequent ly, multi- industry empirical studies of ten reveal that even `successf ul’ quality initiatives have little or no im pact on perf ormance, due to psychometric properties associated with measuring the dependent variable of ® rm perf ormance. A nother line of thought is that ® rms may simply have a business strategy which is misaligned with its current competitive conditions. Wruck and Jensen ( 1994) document the case of Sterling Chemical which had a successf ul TQM program in place, yet experienced a decline in ® rm performance largely because of adverse industry and market f actors. Still another possibility is that ® rms may emphasise a quality strategy that is poorly matched with its competitive environm ent. The ® nal possibility is that an otherwise appropriate quality strategy may be deployed with an inappropriate set of quality tools and technologies .
The problem of ® nding empirical support f or the relationship between quality deployment and perf orm ance is a major driver of this research. This study examines
0020± 7543/99 $12. 00 Ñ 1999 Taylor & Francis Ltd.
R evision received May 1998. ² Department of Marketing and Supply Chain Management, Eli Broad Graduate
School of Management, Michigan State University, N370 North Business Complex, East Lansing, MI 48824, USA . Tel: + 1 (517) 353-6381; Fax: + 1 (517) 432-1112; e-mail: hand® el @ pilot.msu.edu.
³ Charles H. Lundquist College of Business, 1208 University of Oregon, Eugene, OR 97403-1208, USA .
§ Department of Management, Georgia Institute of Technology, 225 North A venue NW, A tlanta, GA 30084, USA. Tel.: + 1 (404) 894-4927.
* To whom correspondence should be addressed.
the im pact of quality tool selection/deploym ent on perf ormance by ® rst `grouping’ a number of tools used in a number of industries into classes ( using f actor analysis. ) We then investigate the relationship between these quality tool groups and a set of perf ormance indicators ( quality strategy, quality perf ormance and business perf orm- ance). These relationships are tested using data obtained through ® eld interviews in 14 ® rms and f rom a large-scale survey of 313 US ® rms in several industries. The study concludes with the strategic implications f or quality and technology managers f aced with the task of allocating resources, developing quality objectives, and evaluating the bene® ts that result f rom the deployment of quality tools and technologies.
2. Literature review
Three major them es can be discerned from the literature on quality tools. First, there have been many attem pts to group quality tools into comprehensive categories. In most cases these categorisatio ns are based on inductive case-based descriptions by practitioners and consultants ( Brocka and Brocka 1992, Greene 1993). Second, the literature emphasizes that the types of tool that ® rms employ have evolved over time ( Lee and Ebrahim pour 1985, Lascelles and Dale 1988, Wilkinson et al. 1995, Mann and Kehoe 1994). Firms are not using the same tools as they did a decade ago. Third, there has been an increasing emphasis on empirical studies geared towards under- standing the types of quality tools used by companies and the relative importance of the di� erent tools ( Modaress and A nsari 1989, Larson and Sinha 1995). Nevertheless, these studies do not yet reveal any signi® cant results regarding the types of tools to be used in di� erent business situations.
Brocka and Brocka ( 1992) proposed a spiral model which consists of the f our key principles of vision, empowerment, continuous evaluation, and customer orientation as a basis f or understandin g the role of quality tools in TQM implem entation. Examples of tools grouped into the eight categories are as f ollows:
� graphical tools ( cause± e� ect diagrams, quality function deploym ent) � company-wi de techniques ( benchmarki ng, quality circles, quality f unction
deployment ) � data analysis tools (control charts, design of experiments) � problem identi® cation tools (cause± e� ect diagrams, control charts) � decision-m aking tools ( auditing, nominal group technique) � modeling tools ( work ¯ ow analysis, quality f unction deployment) � preventive tools ( control charts, foolproo® ng) � creativity tools ( quality circles, brainstorm ing). In examining these categories, it is clear that some tools can be grouped in more
than one category. For example, quality f unction deployment has been categorized as a graphical tool, a company-wi de technique, a problem identi® cation tool, and a modeling tool.
In a somewhat more com prehensive typology, Greene ( 1993) described 98 quality tools and classi® ed them into the f ollowing groups: Group dynamics tools; Statistical tools; Managemen t tools; Implementation tools; Process tools; Know ledge tools; A dvanced statistics tools; Systems tools; Managemen t-by-events tools; Customer understandin g tools; Commitment tools; Innovation tools; Sof tware tools; and Social connectionism tools. A lthough comprehens ive in scope, a large num ber of the quality tools in this typology are not widely used by companies.
1404 R. Hand® eld et al.
Quality tools can also be classi® ed using an evolutionary approach. Traditionally, the use of quality programs and tools have emphasized reactive stances. For example, the use of tools such as Statistical Process Control and R eliability Engineering were employed with the objective of detecting defects. In later years, tools such as zero defects programs, and Total Quality Control ( TQC) were sought to prevent def ects and assure quality. Garvin ( 1987) was the ® rst to point out that a strategic quality initiative requires a deliberate and system-wide use of quality tools and techniques that enable ® rms to di� erentiate their products f rom those of their competitors.
Empirical studies addressing the use of quality tools can be grouped into three categories: Comparison s of practices across countries; Comparison s of practices within a country; and Com parisons of practices across f unctions in ® rms. A merican ® rms tend to use statistical tools such as control charts, acceptance sampling and histograms ( Lee and Ebrahimpou r 1985). However, Japanese ® rms emphasized tools such as Pareto charts and checksheets. In a later study, A merican ® rms that adopted Japanese techniques were f ound to employ tools such as quality planning, worker participation, and team work techniques ( Ebrahimpou r and Lee 1988).
A similar change in the emphasis on quality tools over time can be found in studies that examined practices within countries. In the UK ® rms traditionally emphasized inspection-or iented techniques and paid little attention to cross-f unc- tional techniques such as quality circles and value analysis ( Lascelles and Dale 1988). Over tim e, this pattern changed to an increasing emphasis on customer-driven tech- niques such as custom er satisf action surveys and quality improvement projects ( Wilkinson et al. 1995, Mann and Kehoe 1994). In Canadian ® rms, the current usage of tools such as TQM, benchmark ing and SPC appears to be motivated by strategic concerns ( Larson and Sinha 1995).
Some studies have pointed to di� erences in practices across f unctions within a company. Modaress and A nsari ( 1989) f ound that the manuf acturing f unction tended to use post f acto tools such as inspection, SPC and process capability. However, the design and engineering f unction used proactive tools such as design of experiments ( DOE) and Pareto charts.
To sum marize, there has been a global interest in the types of quality tools and techniques employed by ® rms to achieve a num ber of competitive objectives. The relative use of quality tools varies signi® cantly across countries, across industries and within ® rms. To date, no attem pt has been made to link the latent pattern of quality tool usage to speci® c quality or business perf ormance objectives. For instance, ® rms often have a broad set of objectives underlying their quality management strategy ( e.g. improving design manuf acturability versus process improvement versus inspection f or visual def ects). There is theref ore a need to examine the relationship between various quality strategies and the resulting pattern of quality tool deploym ent. A n analysis of the relationships among quality tools, quality perf ormance and business perf ormance is also required to understand when certain tools can be used successf ully. Finally, the possibility of industry-spe ci® c e� ects in the relationships among quality tools, quality strategies, quality perf orm ance and business perf ormance needs to be explored. This paper seeks to address these gaps in the quality management literature.
Q uality tool deploym ent patterns 1405
3. Q uality strategies and tools
3.1. Patterns of quality tool deploym ent A s can be seen in the discussion in section 2 the concepts associated with the
di� erent quality strategies are well known. However, the patterns of quality tool deployment are of ten unclear to quality managers. This was made clear in an inter- view with a quality manager at a large Fortune 500 computer manuf acturer who made the f ollowing observation: `We have had successes and f ailures in deploying a variety of quality tools. Sometimes they work, and sometimes they don’ t. It appears to be a hit-and-miss process. ’
A lthough it appears that the pattern of quality tool deployment is a ® rm-speci® c phenomenon , we are exploring the possibility that there are common groups of quality tools that tend to be used as a set across a large number of ® rms. Form ally this proposition is stated as f ollows.
Proposition 1: There exist a small ® nite number of underlying dimensions in the pattern of quality tool deployment across organizations.
Moreover, we posit that there exists a grouping of tool deployment patterns that vary based upon the objectives being sought. These strategies are next described.
3.2. Q uality strategies Historically, ® rms in the early stages of TQM implementation employ quality
tools in a random manner ( Hand® eld and Ghosh 1994). In contrast, ® rms in¯ uenced by Japanese quality control management recognized that quality excellence can be achieved by employing a series of interdepend ent strategies: designing f or quality; process control; process improvement and inspection ( Hand® eld 1989).
3.2.1. D esign quality This strategy involves recognizing that product design makes major contribu-
tions to the three major business outcomes of cost, quality and timeliness (Fleischer and Liker 1992). This recognition leads to the idea that quality is designed into the product at least as much as it is built in during manuf acture ( Boothroyd and Dewhurst 1987, Dean and Susman 1989, Hauser and Clausing 1988, Tushman 1979, Whitney 1988). This strategy is typically used in the engineering and product design phase.
3.2.2. Process control The strategy of process control is basically a precursor to the achievement of the
broader objective of process management (A nderson et al. 1994). The methodologi- cal aspects of process managem ent require the use of quality control tools, preven- tive maintenance and unif orm production workloads ( Mizuno 1988, Garvin 1984, 1983, Hayes 1981 ). Process control and measurement is required f or the stability and reliability of the manuf acturing process.
3.2.3. Process im provem ent The strategy of process improvement is the logical step af ter process control.
Examples of process improvement techniques recommended by experts include Taguchi methods, quality f unction deployment, and def ect prevention ( Taguchi 1979, A kao 1990, Sarazen 1990). A study conducted by the A merican Quality Foundation and Ernest & Y oung ( 1992) reported that only three managem ent prac-
1406 R. Hand® eld et al.
tices reportedly had a signi® cant impact on perf ormance, regardless of industry, country, or starting position. These three practices include process improvement methods, strategic plan deployment, and supplier certi® cation program s.
3.2.4. Inspection The strategy of inspecting incoming, in-process and ® nished goods f or def ects
completes the cycle. The inspection and process control strategy are precursor activ- ities in the evolution of quality control. These activities have been superseded by design improvement methods (Fortuna 1990). A dvanced companies producing piece parts typically employ design, whereas service-oriented ® rms utilize process control. Today, the intense competitive environment dictates that most ® rms cannot a� ord to ignore both product and process design improvements.
There also appears to be a new and evolving pattern to the deploym ent of quality strategies. For example, it has been emphasized that quality managers in leading edge companies have shif ted their attention f rom inspecting quality to designing quality into products and services ( Georgantzas et al. 1995). The f our strategies identi® ed here f orm a basis f or selection of quality tools. Tools such as Quality Function Deploym ent, quality circles, equipment calibration testing and design f or manuf acturability are more likely to be used by TQM proponents emphasizing design quality and process im provement as opposed to tools such as acceptance sampling and SPC which are pursued by inspection-or iented ® rms ( R adhakrishnan and Srinidhi 1994, Sower et al. 1993, Flynn et al. 1995). In con- trasting the strategies of high perf ormers with those of low perf ormers, the A merican Quality Foundation and Ernst & Y oung study (1992) found that low-perf orming ® rms tend to inspect-in quality, while high-perf orming ® rms tend to design-in qual- ity. However, the literature is not consistent with respect to the type of strategies ( design versus inspection) that ® rms should pursue. In an interesting contrast, R adhakrishnan and Srinidhi ( 1994) f ound that quality management decisions relat- ing to designing-in and inspecting-in are context-speci® c. They f ound that partial acceptance sam pling ( an inspection tool) was optim al in certain contexts. Moreover, the implementation of quality strategies such as inspection, design quality, process control and process improvement should be related to the types of tools deployed within the organization. This is stated in the f orm of the f ollowing proposition.
Proposition 2: The pattern of quality tool deployment is related to the type of quality strategy adopted by organizations .
A s noted in the literature review, f ew studies have examined the relationship between the pattern of quality tool deployment and quality perf ormance. We there- f ore explore this relationship through the f ollowing proposition.
Proposition 3: The pattern of quality tool deployment is related to quality perf orm- ance in organizations .
We also posit that the deployment of certain quality tools has the potential to a� ect overall ® rm perf ormance measures such as market share, return on assets and growth measures. This relationship is suggested in a recent study which f ound that conf ormance to speci® cations was signi® cantly related to three overall ® rm perf orm- ance measures: return on investment ( R OI) growth, sales growth, and return on sales ( R OS) grow th ( Forker et al. 1996). Our study posits that the relationship between
Q uality tool deploym ent patterns 1407
quality strategy and business perf ormance may be mediated by the types of quality tools used.
Proposition 4: The pattern of quality tool deployment is related to business perf orm- ance.
The f our stated propositions attempt to identif y overall patterns in the use of quality tools. The prolif eration of quality managem ent concepts and tools across a variety of industries is evident, given that previous winners of various quality awards ( including the Baldridge A ward and New Y ork’ s state quality award) span a variety of organizations, including a hotel, a rock supplier, the New Y ork State Police, a school system, and architects ( Godf rey 1993). R ecognizing that the pattern of tool deployment may vary across industries, the f our propositions stated above are tested both across multiple industries and within three subsamples: the automotive, elec- tronics and consumer products industries.
4. M ethodo logy
4.1. S urvey instrum ent A comprehensive list of quality tools and techniques was compiled using a two
stage technique. In the ® rst stage, a caref ul review of the literature was conducted to identif y an a priori set of tools and techniques used in di� erent settings and indus- tries. In particular, care was taken to include a wide variety of tools and techniques such as graphical tools, problem identi® cation tools, modeling tools, preventive tools and creativity tools ( Brocka and Brocka 1992). In the second stage, in-depth case studies of 14 North A merican and European manuf acturing organizations were used to selectively reduce the list of quality tools to include those that were commonly used. In selecting these 14 ® rms, care was taken to include a diverse sample of ® rms that were in various stages of TQM implementation. In most cases, the quality director or vice-president at the corporate o� ce was contacted and interviewed. A ll the interviews were carried out on-site using a structured interview protocol. The ® eld notes from these interviews were used as the basis f or choosing the quality tools f or the questions in this survey.
The two stage technique yielded a list of 38 quality tools and techniques (contact the authors f or a set of de® nitions f or these tools); 27 out of the 38 quality tools included in this research are f requently mentioned in the quality literature. The remaining 11 tools were being employed in a number of ® rms interviewed in stage one of the study.
For each of the 38 quality tools and techniques, respondents were asked to: rate the implementation status of the tool ( on a 4 point scale with 4 being `high use’ and 1 being `not used’ ); and rate the tool’ s impact on quality improvement ( on a 7 point scale with 7 being `high impact’ , and 1 being `low impact’ ). R espondents were also asked to allocate 100 points among f our quality strategies ( inspection, process con- trol, process improveme nt, and design). The allocation of points represented the relative emphasis of the ® rm’ s quality strategy.
4.2. S am ple The ® rms used in the sample were identi® ed through the A merican Society f or
Quality Control ( A SQC). A list of 3000 quality directors and vice-preside nts was obtained, and a sub-sample of 1469 manufactu ring ® rms in the automotive, chemi- cal, computer, construction , consumer products, def ense electronics, industrial prod-
1408 R. Hand® eld et al.
ucts, medical device, packaging, pharmaceu tical, paperboard, semiconduc tor and telecom munications industries were identi® ed. Two mailings with one f ollow-up reminder produced 351 surveys ( a 23% response rate). Of these surveys, 38 con- tained missing data f or the variables used in this study, and were removed from the sample, resulting in a ® nal sample size of 313 ® rms ( a 21% response rate). The ® rms were located in all 50 A merican states. The ® rms were f rom all of the f ollowing industry groups: autom otive, building materials, computer, consumer products, def ence and aerospace, electronics , f ood, chemicals, scienti® c, service industries and others. There is evidence to show that quality managem ent techniques can vary signi® cantly within the same organization across business units (Benson et al. 1991). Consequent ly, it was deemed appropriate to select the division level as the unit of analysis in this study.
5. Results
5.1. D escriptive statistics Prior to analysis, the sample was screened f or outliers and data errors. The
means, standard deviations, and correlations am ong the quality tool f actors, quality strategies, quality perf ormance and business perf ormance are shown in tables 1( a) ± ( d). The f actor analysis results which were used in developing the quality tool con- structs are discussed later.
Q uality tool deploym ent patterns 1409
Std. Minimum Maximum Sample Variable Mean Dev. value value size
1 Designing quality² 25.858 15.565 0 80 313 2 Inspection² 22.645 18.121 0 100 313 3 Proces control² 27.359 13.234 0 80 313 4 Process improvement² 24.138 12.663 0 100 313 5 HR factor ³ 4.324 1.385 1 7 155 6 Measurement Factor ³ 4.253 1.295 1 7 172 7 Design Factor ³ 3.527 1.488 1 7 126 8 Discipline Factor ³ 4.944 1.284 1 7 232 9 Customer rejects§ 0.060 0.428 1 + 1 255
10 Defect rates§ 0.079 0.352 1 + 1 225 11 Rework rates§ 0.070 0.335 1 + 1 191 12 Scrap rates§ 0.047 0.310 1 + 1 232 13 Market share³ 5.155 1.403 1 7 310 14 ROA ³ 4.813 1.321 1 7 305 15 MS growth ³ 4.860 1.434 1 7 307 16 Sales growth ³ 4.938 1.455 1 7 307 17 ROA growth ³ 4.683 1.327 1 7 303 18 Production costs³ 4.482 1.159 1 7 305 19 Customer service³ 5.360 1.090 2 7 308 20 Product quality ³ 5.594 1.009 2 7 308 21 Competitive position ³ 5.271 1.165 1 7 310 22 Customer relationship ³ 5.023 1.147 1 7 310
For items marked ² scale is from 0 to 100. For items marked ³ scale is from 1 to 7. For items marked § scale is from 1 to + 1.
Table 1( a). Descriptive statistics of quality tool factors, quality strategy and performance items.
In table 1(a) the descriptive statistics ( mean, standard deviations, minimum and maximum values) f or the items comprising the quality tool f actors, quality strategies and performance variables are presented. Table 1( b) shows the correlations of the f our quality strategies ( design, inspection, process control and process improvement) with the f our quality tool f actors (Hum an R esources, Measurement, Design and Discipline). A s can be seen f rom this table, all quality strategies are correlated with one another at a p < 0.05 signi® cance level. In table 1( c) the correlations between quality tool f actors and indicators of quality perf ormance are presented. In table 1( d) the correlations between quality tool factors and indicators of business perf ormance are presented. The implications of these statistics f or the results will be discussed f ollowing the presentation of the proposition testing.
5.2. Measurem ent The items f orming the quality tool factor scales were subject to a process of item
puri® cation by testing f or unidim ensionality within quality tool f actors using factor analyses, and by testing f or internal consistency using Cronbach’ s alpha (Cronbach 1951). A ll scales within the quality tool f actors were f ound to be unidimension al. A s can be seen f rom table 2, the scales f or the quality tool f actors were internally consistent and reliable, with the Cronbach’ s alphas ranging f rom 0.79 to 0.84. Consistent with the technique of item puri® cation suggested by Churchill ( 1979), items within a scale were eliminated if their corrected item-total correlation (® rst column in table 2) was less than 0. 45. A high score for an item’ s corrected item-total correlation indicates that all items within a domain of a concept have an equal
1410 R. Hand® eld et al.
1 2 3 4 5 6 7
1 Designing quality Ð 2 Inspection 0.455² Ð 3 Process control 0.336² 0.505 ² Ð 4 Process improvement 0.259² 0.516 ² 0.246 ² Ð 5 HR factor 0.246 ² 0.377 ² 0.136 0.159 ³ Ð 6 Measurement f actor 0.101 0.368 ² 0.236 ² 0.247² 0.567² Ð 7 Design f actor 0.233 ² 0.322 ² 0.151 0.078 0.631² 0.653 ² Ð 8 Discipline factor 0.148 0.274 ² 0.178 ² 0.087 0.693² 0.636 ² 0.633 ²
n 107. ² Signi® cant at 0.05. ³ Signi® cant at 0.10.
Table 1( b). Correlations f or quality tool f actors and quality strategy items.
1 2 3 4 5 6 7
1 Customer rejects Ð 2 Def ect rates 0.605 ² Ð 3 R ework rates 0.396 ² 0. 548² Ð 4 Scrap rates 0.476 ² 0. 582² 0.539² Ð 5 HR factor 0.086 0.169 0.004 0.071 Ð 6 Measurement f actor 0.140 0.042 0.079 0.089 0.631² Ð 7 Design f actor 0.149 0.054 0.057 0.045 0.727² 0.651 ² Ð 8 Discipline factor 0.279 ² 0.216 ³ 0.081 0. 041 0.705² 0.670 ² 0.735 ²
n 69. ² Signi® cant at 0.05. ³ Signi® cant at 0.10.
Table 1( c). Correlations f or quality tool f actors and quality performance items.
Q uality tool deploym ent patterns 1411
1 2
3 4
5 6
7 8
9 1 0
1 1
1 2
1 3
1 M
a rk
et sh
ar e
Ð 2
R O
A 0.
31 1 ²
Ð 3
M S
gr o
w th
0. 24
6 ²
0 .4
4 9²
Ð 4
S a le
s g ro
w th
0. 21
4 ²
0 .3
9 0²
0 .7
8 1²
Ð 5
R O
A gr
o w
th 0.
12 2
0 .6
6 8²
0 .5
7 6²
0. 6 4 3 ²
Ð 6
P ro
d u
ct io
n co
st s
0. 02
3 0 .0
44 0 .0
6 8
0. 1 0 3
0. 1 0 4
Ð 7
C u
st o
m er
se rv
ic e
0. 00
9 0 .2
3 9²
0 .2
6 2²
0. 2 8 1 ²
0. 1 7 3³
0. 0 5 2
Ð 8
P ro
d u
ct q
u a li
ty 0.
17 8
³ 0 .1
7 1³
0 .2
1 8²
0. 2 6 0 ²
0. 1 0 6
0. 1 5 2
0 .5
12 ²
Ð 9
C o
m p
et it
iv e
p o
si ti
o n
0. 53
3 ²
0 .3
6 2²
0 .4
7 2²
0. 4 6 7 ²
0. 3 0 9²
0. 0 2 0
0 .1
94 ³
0 .4
37 ²
Ð 1 0
C u
st o
m er
re la
ti o
n sh
ip 0.
09 7
0 .3
4 7²
0 .2
7 1²
0. 3 1 7 ²
0. 3 8 4²
0 . 05
4 0 .3
53 ²
0 .2
98 ²
0 .1
76 ³
Ð 1 1
H R
fa ct
o r
0. 11
2 0 .1
4 1
0 .2
9 7²
0. 2 7 1 ²
0. 2 4 8²
0 .1
2 9
0 .1
37 0 .1
91 ³
0 .1
06 0 .3
8 7 ²
Ð 1 2
M ea
su re
m en
t fa
ct o
r 0.
24 3 ²
0 .0
3 5
0 .2
0 5²
0. 1 5 6
0. 0 7 1
0 .0
3 8
0 .1
28 0 .2
51 ²
0 .1
88 ³
0 .2
4 8 ²
0. 5 6 7²
Ð 1 3
D es
ig n
fa ct
o r
0. 00
4 0 .0
6 5
0 .2
4 1²
0. 2 1 7 ²
0. 2 1 0²
0 .0
3 7
0 .2
10 ²
0 .2
44 ²
0 .1
19 0 .2
7 9 ²
0. 6 4 6²
0 .6
4 4²
Ð 1 4
D is
ci p
li n
e fa
ct o
r 0.
04 2
0 .1
8 8
0 .2
5 4²
0. 1 6 9
0. 1 9 9
0 .0
8 5
0 .2
23 ²
0 .2
14 ²
0 .1
29 0 .2
1 9
0. 6 9 0²
0 .6
3 1²
0 .6
36
n 10
4. ²
S ig
n i®
ca n
t a t
0. 0 5 .
³ S
ig n
i® ca
n t
a t
0 .1
0 .
T a b
le 1 (d
). C
o rr
el at
io n
s fo
r q
u a li
ty to
o l
fa ct
o rs
an d
b u
si n
es s
p er
fo rm
an ce
it em
s.
amount of common core, and hence responses to these similar items should have high intercorrela tions. The third column in table 2 indicates that f or all the items included in the ® nal instrument, the non-inclusion of each item results in a reduction of internal consistency ( as can be seen f rom the reduction in alpha values). Overall, the measurement analyses indicated that the scales were reliable. A f ter the item puri® cation process, an examination of the item s within the f our quality tool factors revealed that the items had high content validity. For example, the ® ve items under the hum an resources f actor ( HR ) are item s indicating the deliberate deployment of methods that empow er and recognize employees in quality im provement e� orts.
5.3. T esting of propositions A combination of f actor analysis, sim ple regression analysis, stepw ise regression
analysis, and subsample testing was used to test the f our propositions identi® ed in earlier paragraphs.
5.3.1. Proposition 1 The results of testing Proposition 1 are shown in table 3. A n exploratory factor
analysis using principal components with varimax rotation was conducted on the set of 38 quality tools, identi® ed earlier in the section on developing the survey instru- ment. Only f actors that accounted for variances greater than one ( i.e. , eigen- values > 1) were extracted. Four f actors were extracted that accounted f or 63. 3%
1412 R. Hand® eld et al.
Corrected Cronbach’ s item-total Cronbach’ s alpha if item Sample
Factor, with items correlation alpha is deleted size
HR factor 0.8282 155 Workers perform ® nal inspection 0.6990 0.7658 Workers perform in-process inspection 0.6954 0.7712 Workers responsible for defect does 0.6607 0.7770
rework Employee suggestion program 0.5516 0.8078 Quality circles 0.5152 0.8227
Measurement f actor 0.8429 172 Histograms 0.6897 0.8006 Pareto analysis 0.6614 0.8058 Process capability studies 0.7447 0.7808 R egression 0.4928 0.8499 Statisical process control 0.6639 0.8047
Design f actor 0.8054 126 Quality function deployment 0.6980 0.7347 Zero defects program 0.5849 0.7684 Design of experiments 0.6294 0.7542 Failure mode and e� ects analysis 0.4529 0.8073 Design f or manuf acturability 0.5910 0.7664
Discipline factor 0.7352 232 Continuous improvement programs 0.5857 0.6252 Total quality management program 0.6006 0.5984 Preventive maintenance 0.5005 0.7177
Table 2. R eliabilities of items f or the four quality tool factors.
of the total variation in the observed variables. Table 3 shows the total and the cumulative variance f or each extracted f actor. To interpret the f actors only items which had `strong’ f actor loadings (greater than 0.5 in absolute value, shown in bold) were included ( Norusis 1990). The resulting f our f actors may be interpreted, respect- ively, as Human R esource Tools, Measurem ent Tools, Design Tools, and Discipline Tools. Note that the names provided to the f our latent factors are by no means unique. How ever, regardless of the nomenclatur e used, the substantive import of the pattern of tools is unif orm. It may also be noted that the item `employee involvement in quality planning’ did not load uniquely on one of the f actors. A s can be seen f rom table 3, this tool loaded high on both HR f actor and Discipline Factor. Similarly, the item `pro® t sharing with employees’ loaded high on both HR factor and Design Factor. Since cross loading can be a problem, especially in summated scales, we excluded both these items in arriving at the four quality tool f actors.
Proposition 1 posits that a small number of underlying dimensions in the pattern of quality tool deployment exists across a large number of organizations . The results f rom table 3 indicate that there exist f our fundament al types of quality tool deploy- ment: Human R esource Tools; Measurement Tools; Design Tools; and Discipline Tools.
The set of items under Human R esource Tools ( inspection and rework and involvement in the f orm of employee suggestions and participation in team s such
Q uality tool deploym ent patterns 1413
Factor 1 Factor2 Factor 3 Factor 4 Variables ( HR ) ( Measurement) (Design) (Discipline)
Workers perform ® nal inspection 0.795 0.152 0.128 0.192 Workers perform in-process inspection 0.731 0.292 0.049 0.298 Workers responsible f or defect does rework 0.671 0.152 0.313 0.294 Employee suggestion program 0.468 0.114 0.112 0.524 Quality circles 0.506 0.172 0.383 0.342 Histograms 0.199 0.842 0.144 0.089 Pareto analysis 0.136 0.803 0.089 0.269 Process capability studies 0.102 0.691 0.381 0.311 R egression 0.307 0.626 0.390 - 0.038 Statistical process control 0.100 0.651 0.175 0.446 Quality function deployment 0.139 0.187 0.788 0.209 Zero defects program 0.431 0.141 0.685 0.017 Design of experiments 0.177 0.419 0.621 0.271 Failure mode and e� ects analysis - 0.137 0.219 0.599 0.287 Design f or manufacturability 0.277 0.246 0.540 0.393 Continuous improvement programs 0.196 0.255 0.151 0.748 Total quality management program 0.277 0.178 0.331 0.687 Preventive maintenance 0.241 0.382 0.284 0.513 Customer satisfaction surveys 0.305 0.228 0.447 0.300 Supplier certi® cation/quali® cation 0.376 0.276 0.345 0.310 Pro® t sharing with employees 0.471 0.023 0.572 0.088 Employee involvement in quality planning 0.453 0.163 0.259 0.664
Eigenvalue 9.938 1.329 1.685 0.989 Percentage of variance explained 45.172 6.039 7.660 4.494 Cumulative proportion of total variance 45.172 51.211 58.871 63.365
explained
Table 3. R otated f actor loadings for the four quality tool f actors.
as quality circles) suggest that a common theme, empowerment through higher responsibility, underlies the use of these tools. The second grouping (M easurement Tools) constitute a set of techniques that are aimed at monitoring the results of processes. Statistical process control, process capability studies, histograms, pareto analysis and regression collectively represent visual and analytical devices that can be used f or measuring quality-relate d results. The third grouping ( Design Tools) are composed of techniques for enhancing quality issues in the design process. Quality f unction deployment ( QFD), Zero Def ects programs, design of experim ents ( DOE), f ailure mode and e� ects analysis ( FMEA ) and design f or manuf acturability ( DFM), all f acilitate one com mon objective: proactive avoidance of def ects (both in-house and ® eld) and waste.
A ll the above-ment ioned tools require cross-f unctional coordination with a clear customer f ocus. The primary impact of the use of these tools is on the design of products and processes. The ® nal grouping of quality tools (Discipline Tools) con- sists of continuous improvement programs, total quality managem ent programs ( TQM) and preventive maintenance. These tools are called Discipline Tools because the philosophy underlying these tools is related to f act-based decision-mak ing and applies in a variety of contexts. For example, TQM and continuous improvement principles can be applied successf ully in manuf acturing and service ® rms, pro® t and not-f or-pro® t ® rms, and indeed in ® rms across di� erent cultures. Even at a micro level, these techniques can be applied to manuf acturing problem s, procurement problems or engineering problems. The discipline set of tools typically guide, de® ne and shape the evolving `quality culture’ within organizations.
In testing Propositions 2± 4, a combination of sim ple regression analysis and stepwise regression analysis was used. Before discussing the results f or these propo- sitions, we point out some general issues that have relevance f or interpreting the regression results. For the regression analyses, the independent variables were the f our quality tool factors that were f ound as a result of testing Proposition 1. The dependent variables in each case related to the proposition that was being tested. For example, in testing Proposition 2, the dependent variable was the quality strategy type, i.e. inspection, design, process control, and process improvement. Similarly, in testing Proposition 3, the dependent variable was quality perf ormance which was operationalize d as customer rejects, def ect rates, rework rates, and scrap rates. The scores f or each of the f our quality factors were arrived by ® rst sum ming the responses f or each of the underlying items and then dividing the sum by the number of items included in each f actor.
The possibility of multicollinear ity am ong the f our independent variables ( quality tool f actors) exists. Consequen tly, while interpreting results we relied on stepwise regression and the F test, both of which are less a� ected by multicollinear ity.
5.3.2. Proposition 2 In Proposition 2 we sought to understand the impact that the pattern of quality
tool deployment had on the type of quality strategy being pursued. The results of the simple regression and step-wise regressions are shown in tables 4( a) and ( b), respect- ively. In table 4( a), the model R 2 , F-value, the two-tail p-value f or the signi® cance of beta ( or the regression itself ), and the estimates f or standardized (beta) and unstan- dardized slopes are shown. The simple regression results show that the e� ect of quality tool f actors on design quality and on inspection are statistically signi® cant based on the F value and the associated p-values. The design tool f actor is signi® -
1414 R. Hand® eld et al.
Q uality tool deploym ent patterns 1415
Dependent variable, Regression estimates with independent Model
variables R 2 F -value p -value§ Standardized¶ Unstandardized
Design quality 0.083 2.318 0.062 ³ HR f actor 0.106 0.229 2.558 Measurement factor 0.356 - 0.126 - 1.391 Design factor 0.137 0.211 2.243 Discipline f actor 0.661 - 0.065 - 0.695
Inspection 0.183 5.699 0.000 ² HR f actor 0.038 ² - 0.278 - 3.705 Measurement factor 0.062 ³ - 0.242 - 3.196 Design factor 0.669 - 0.057 - 0.723 Discipline f actor 0.437 0.108 1.392
Process control 0.058 1.560 0.191 HR f actor 0.881 - 0.021 - 0.184 Measurement factor 0.114 0.220 1.862 Design factor 0.884 - 0.021 - 0.170 Discipline f actor 0.657 0.066 0.548
Process improvement 0.088 2.471 0.049 ² HR f actor 0.246 0.164 1.264 Measurement factor 0.010 ² 0.356 2.726 Design factor 0.247 - 0.163 - 1.203 Discipline f actor 0.313 - 0.149 - 1.109
² Signi® cant at 0.05. ³ Signi® cant at 0.10. § The p-value in the rows for dependent variables indicates the signi® cance level of the F-statisic, and the p-value in the rows for independent variables indicate the signi® cance of the individual estimates. ¶ Signi® cant beta coe� cients are indicated in bold.
Table 4( a). Simple regression results of quality strategies versus quality tool factors.
Dependent variable, Variables Entered with independent Model
variables R 2 F -value p -value§ Standardized¶ Unstandardized
Design quality 0.061 6.792 0.010 ² HR f actor 0.010 ² 0.246 2.747
Inspection 0.177 11.195 0.000 ² HR f actor 0.024 ² - 0.247 - 3.290 Measured f actor 0.037 ² - 0.228 - 3.011
Process control 0.056 6.203 0.014 ² Measurement factor 0.014 ² 0.236 2.001
Process improvement 0.061 6.851 0.010 ² Measurement factor 0.010 ² 0.247 1.896
Table 4( b). Step-wise regression results of quality strategies versus quality tool f actors. ( Key as f or table 4( a).)
cantly related to the design quality strategy at p < 0.05. The HR tool f actor and measurement tool f actor are signi® cantly related to the inspection strategy, although the negative sign associated with the beta coe� cients is unexpected. Further analysis of the sample revealed that a vast majority of ® rms did not emphasize an inspection strategy ( only 6% of the ® rms gave more than 50 points out of a total of 100 points to inspection strategy). Consequent ly, the results of this regression may not be representative of ® rms emphasizing an inspection strategy.
A nother exploratory technique was carried out to investigate the negative signs associated with the HR and measurem ent tool f actors. Firm s were split into two groups of `high’ and l̀ow’ based on their inspection strategy responses. In order to improve com parability between the two groups, the overall sample was divided into a `high inspection’ group and a `low inspection’ group based on a cuto� score of 25 points out of 100 points. A s the research intent was to verif y whether the negative signs associated with the HR and measurement tool f actors was an artefact of the di� erential emphasis placed on inspection strategies, a simple group separation based on f requency distribution was conducted. The high group consisted of 122 ® rms and the low group consisted of 136 ® rms. Separate regressions were run for the two groups. The regression f or the low group did not have any independent variables with signi® cant beta estimates. For the high group, only the HR f actor had a sig- ni® cant beta estimate, with the negative sign still present. A n additional investigation regressed the individual variables within the HR tool f actor on the inspection strat- egy variable. Only the item of `employee suggestion programs’ was signi® cant with a p-value of 0.0884. One possible explanation may be that ® rms emphasizing inspec- tion relied less on human intervention of empow erment strategies and more on a structured and programmed inspection processes.
The measurement tool f actor was signi® cantly related to both process control and process improvement strategies. This ® nding con® rms the utility of tools such as SPC and process capability studies in process control and process improvement initiatives. The results of step-wise regression ( see table 4( b)) are, in general, consis- tent with the simple regression results. However, the stepwise regression operates on more stringent conditions: the most signi® cant independent variable enters the equa- tion by taking into account correlations among the independent variables. In such a setting one can be reasonably sure that if more than one independent variable enters the equation, then each additional variable has new increm ental explanatory power in the equation. For the stepwise regression of quality tool f actors on quality strat- egy, the model R
2 ranged f rom 5% to 17% . The design tool f actor was still signi® -
cantly related to design quality strategy. HR and measurement tool f actors were again negatively related to inspection. The f act that the measurement tool f actor was signi® cantly related to both process control and process improvemen t strategies, even in the stepwise regression, lends f urther validity to this result.
In summary, it appears that some support exists for Proposition 2: an e� ective design quality strategy requires the use of design tools such as Design of Experiments, and Design f or Manuf acturability. The case of ® rms using an inspec- tion strategy is still inconclusive, in large part because our sample did not represent ® rms that uniquely deployed inspection strategy. However, process improvement and process control strategies were clearly being addressed through measurement tools such as histograms, pareto analysis, process capability studies, regression and SPC.
1416 R. Hand® eld et al.
5.3.3. Proposition 3 In Proposition 3 we were interested in understandin g the impact that the four
quality tool f actors have on quality perf ormance. Four objective indicators of qual- ity perf orm ance were used: customer rejects, def ect rates, rework rates and scrap rates. The results of the simple regression and step-wise regressions are shown in tables 5( a) and (b), respectively. The simple regression results show that the e� ect of quality tool f actors on def ect rates was statistically signi® cant based on the F value and the associated p-value. The discipline tool f actor was signi® cantly related to
Q uality tool deploym ent patterns 1417
Dependent variable, Regression estimates with independent Model
variables R 2 F -value p -value§ Standardized¶ Unstandardized
Customer rejects 0.067 1.668 0.164 HR f actor 0.173 0.201 0.060 Measurement factor 0.635 - 0.067 - 0.020 Design factor 0.771 - 0.044 - 0.013 Discipline f actor 0.068 ³ - 0.280 - 0.082
Def ect rates 0.087 1.989 0.104 HR f actor 0.137 - 0.235 - 0.058 Measurement factor 0.299 0.159 0.038 Design factor 0.265 0.174 0.040 Discipline f actor 0.106 - 0.261 - 0.061
R ework rates 0.017 0.320 0.864 HR f actor 0.816 0.042 0.008 Measurement factor 0.356 0.151 0.028 Design factor 0.705 - 0.071 - 0.013 Discipline f actor 0.471 - 0.138 - 0.027
Scrap rates 0.022 0.507 0.731 HR f actor 0.421 0.128 0.027 Measurement factor 0.952 - 0.009 - 0.002 Design factor 0.717 0.056 0.011 Discipline f actor 0.183 - 0.219 - 0.043
Table 5( a). Simple regression results of quality performance versus quality tool f ators. ( Key as f or table 4( a).)
Dependent variable, Variables entered with independent Model
variables R 2 F -value p -value§ Standardized¶ Unstandardized
Customer rejects 0.047 4.719 0.032 ² Discipline f actor 0.032 ² - 0.216 - 0.064
Def ect rates No variable entered
R ework rates No variable entered
Scrap rates No variable entered
Table 5( b). Step-wise regression results of quality performance versus quality tool f actors. ( Key as for table 4( a).)
customer rejects at p < 0.10. The HR tool f actor was signi® cantly related to the def ect rate at p < 0.10. Both of these regression models estimated positive beta coe� cients, suggesting that the use of discipline tools reduces customer rejects and the use of HR tools reduces product def ect rates.
The results of the step-wise regression are shown in table 5( b). Only the stepwise regression of quality tool factors on custom er rejects was statistically signi® cant at p < 0.05, with a model R 2 of 6% . The discipline tool f actor was negatively and signi® cantly related to customer rejects as predicted. This result con® rms the pro- f essed purpose of discipline related tools such as continuous im provement programs and TQM in improving custom er satisf action. Given the low value of R
2 , the results
must be interpreted with caution. Speci® cally, predictive relationships between qual- ity tools and quality performance cannot be established on the basis of these results.
In summary, there is some support f or an indirect relationship between quality tool deployment and quality perf ormance. In certain instances the relationship is strong enough to have a direct impact. For example, our study f ound a strong relationship between the use of discipline tools and reduction in custom er rejects. On a moderate level, the role of HR tools on reduction of def ect rates was also f ound. In particular, it appears that empowerment and involvement related HR tools can help improve the rate of success of defect reduction program s.
5.3.4. Proposition 4 Proposition 4 posits a relationship between the f our quality tool factors and
business perf ormance. Ten indicators of business perf ormance were used: market share, R OA , market share growth, sales growth, R OA growth, production costs, customer service, product quality, com petitive position and custom er relationship. One dependent variable, Customer R elationship, was a composite index of several items, including determining f uture customer expectations , handling custom er com- plaints, custom er responsivene ss and customer f eedback. The unidimension ality of this composite index was veri® ed. The results of the sim ple regression and step-wise regressions are shown in tables 6( a) and (b), respectively. The simple regression results show that of the ten regressions of quality tool factors on business perf orm- ance, ® ve regression equations were statistically signi® cant. Two of these regressions were statistically signi® cant at a p < 0. 05 signi® cance level indicating strong relation- ships between tool deploym ent and business perf ormance. The model R
2 ranged
f rom 3% to 20% . The HR tool f actor was signi® cantly related to grow th measures ( market share and sales) and to the customer relationship variable. The measurement tool f actor was strongly related to market share ( b 0.4750). The design tool factor was also signi® cantly related to market share, but the b -coe� cient had a negative sign. It is possible that the perceptual judgment (Likert scale of 1 to 7) through which this indicator was measured could have contributed to this anomaly.
The results of the step-wise regression are shown in table 6( b). Seven of the ten regression models were statistically signi® cant at p < 0.05. Six of these seven regres- sion equations were statistically signi® cant at p < 0. 01 indicating strong relation- ships. The HR tool factor was signi® cantly related to three growth measures ( market share, sales, and R OA ) and to custom er relationships . The Discipline tool f actor was signi® cantly related to custom er service. This ® nding is similar to the earlier ® nding of a signi® cant relationship between the discipline tool f actor and the quality per- f ormance indicator of custom er rejects. Discipline tools appear to impact customer driven measures of quality and business perf ormance. Measurem ent tools were sig-
1418 R. Hand® eld et al.
Q uality tool deploym ent patterns 1419
Dependent variable, Regression estimates with independent Model
variables R 2 F -value p -value§ Standardized¶ Unstandardized
Market share 0.117 3.369 0.012 ² HR f actor 0.306 0.142 0.136 Measurement factor 0.001 ² 0.448 0.427 Design factor 0.048 ² - 0.276 - 0.253 Discipline f actor 0.252 - 0.166 - 0.154
R OA 0.050 1.314 0.270 HR f actor 0.611 0.074 0.065 Measurement factor 0.354 - 0.129 - 0.112 Design factor 0.688 - 0.058 - 0.048 Discipline f actor 0.093 ³ 0.254 0.214
MS growth 0.095 2.645 0.038 ² HR f actor 0.136 0.209 0.196 Measurement factor 0.968 0.005 0.005 Design factor 0.665 0.061 0.054 Discipline f actor 0.643 0.068 0.061
Sales growth 0.079 2.160 0.079 ³ HR f actor 0.067 ³ 0.260 0.255 Measurement factor 0.906 - 0.016 - 0.016 Design factor 0.473 0.101 0.094 Discipline f actor 0.670 - 0.063 - 0.059
R OA growth 0.085 2.319 0.062 ³ HR f actor 0.187 0.188 0.174 Measurement factor 0.158 - 0.193 - 0.177 Design factor 0.228 0.171 0.151 Discipline f actor 0.572 0.083 0.074
Production costs 0.021 0.533 0.711 HR f actor 0.249 0.172 0.138 Measurement factor 0.816 - 0.033 - 0.026 Design factor 0.611 - 0.076 - 0.058 Discipline f actor 0.822 0.035 0.027
Customer service 0.058 1.573 0.187 HR f actor 0.748 - 0.046 - 0.032 Measurement factor 0.602 - 0.072 - 0.050 Design factor 0.386 0.124 0.082 Discipline f actor 0.136 0.224 0.150
Product quality 0.070 1.920 0.113 HR f actor 0.860 0.025 0.016 Measurement factor 0.320 0.137 0.089 Design factor 0.505 0.095 0.060 Discipline f actor 0.749 0.047 0.030
Competitive position 0.035 0.926 0.452 HR f actor 0.905 0.017 0.012 Measurement factor 0.198 0.181 0.129 Design factor 0.807 - 0.035 - 0.024 Discipline f actor 0.858 0.027 0.019
Customer relationship 0.165 5.051 0.001 ² HR f actor 0.001 ² 0.443 0.317 Measurement factor 0.654 0.058 0.041 Design factor 0.817 0.031 0.021 Discipline f actor 0.304 - 0.145 - 0.100
Table 6( a). Simple regression results of business performance versus quality tool factors. ( Key as for table 4( a).)
ni® cantly related to market share and product quality. The importance of meas- urement tools such as SPC and process capability f or achieving conf ormance quality has long been established ( Benson et al. 1991, A dam 1994). This ® nding corroborate s earlier evidence in the literature. The Design f actor was signi® cantly related to market share, albeit, with a negative sign. This ® nding is counterintui tive and a more objective measure of market share may have yielded di� erent results.
In sum mary, it appears that certain quality tool f actors have a strong impact on business perf orm ance measures, as posited in Proposition 4. Growth measures of business perf orm ance appear to be impacted more by HR tools. Customer-driven measures of business perf orm ance appear to be a� ected by discipline tools. Business perf ormance measures such as product quality, which are largely dependent on internal processes, appear to be impacted by measurement tools.
5.4. Post hoc analysis A lthough these relationships are important, the issue of industry e� ects in the
pattern of relationships among quality tool f actors, quality strategy, quality per- f ormance, and business perf ormance is required to f urther strengthen the validity of these ® ndings. The sample of 313 ® rms comprised a wide variety of industries such as def ence and aerospace, medical products, consumer products, computers, auto- motive, electronics, petroleum, f ood and building materials. In order to test the propositions within an industry, a su� ciently large industry sample size was desir-
1420 R. Hand® eld et al.
Dependent variable, Variables entered with independent Model
variables R 2 F -value p -value§ Standardized¶ Unstandardized
Market share 0.102 5.922 0. 004² Measurement factor 0. 001² 0.422 0.402 Design f actor 0. 027² - 0.275 - 0.252
R OA No variables entered
MS growth 0.088 10.050 0. 002² HR f actor 0. 002² 0.297 0.278
Sales growth 0.073 8.225 0. 005² HR f actor 0. 005² 0.271 0.266
R OA growth 0.060 6.533 0. 012² HR f actor 0. 012² 0.244 0.226
Production costs No variables entered
Customer service 0.051 5.590 0. 020² Discipline f actor 0. 020² 0.225 0.151
Product quality 0.059 6.595 0. 012² Measurement factor 0. 012² 0.243 0.158
Competitive position No variables entered
Customer relationship 0.156 19.454 0. 000² HR f actor 0. 000² 0.395 ² 0.283
Table 6( b). Step-wise regression results of business performance versus quality tool factors. ( Key as f or table 4( a).)
able to assess overall patterns. The f ollowing industry groupings contained the lar- gest number of ® rms within each industry: automotive ( 56 ® rms); computer ( 35 ® rms); electronics ( 17 ® rms); f ood ( 17 ® rms) and consumer products ( 12 ® rms). To achieve a large industry sam ple size, we grouped similar industries together to the extent possible. This grouping led to three industry subsamples: automotive industry ( 56 ® rms), electronics and computer industry ( 52 ® rms), and consumer products and f ood industry ( 29 ® rms).
5.4.1. Autom otive industry The regression results f or the automobile industry are summarized in table 7. For
the sake of brevity, only stepwise regression results will be highlighted in this dis- cussion. HR tools were signi® cantly related to design quality strategy, growth meas- ures ( market share and sales), and customer relationship. The anom alous ® nding of a negative relationship between inspection strategy and HR tools persisted in the autom obile sam ple. Discipline tools were signi® cantly related to the quality perf orm- ance indicator of defect rates and the business perf orm ance indicators of market share and custom er service. Measurement tools were signi® cantly related to com- petitive position.
5.4.2. Electronics and com puter industry The regression results f or the electronics and computer industry are summ arized
in table 7. Measurement tools were signi® cantly related to design quality strategy and to R OA . However, the R OA -measurement tool relationship was negative. Design tools were signi® cantly related to product quality. The de® nition of quality f or ® rms in the electronics industry appears to go beyond conf orm ance to include design quality consideration s. Discipline tools were signi® cantly related to R OA and market share growth. HR tools were signi® cantly related to two growth measures ( sales and R OA ) and to improved customer relationships .
5.4.3. Consum er products and f ood industry The regression results f or the consumer products and f ood industry are summ ar-
ized in table 7. Discipline tools were signi® cantly related to design quality strategy and the business perf ormance indicator of production costs. Firms in the consumer and f ood industries f ace intense competition and typically operate on low pro® t margins. The dual pressures of stream lining operating costs and di� erentiation lead to a f ew survivors in this industry who f ace intense rivalry f rom each other. Our study points to the use of discipline tools as a mechanism to deal with the dual pressures of di� erentiation and low costs. HR tools were signi® cantly and negatively related to market share growth, which is contrary to intuition. Measurement tools were emphasized by ® rms pursuing process improvement strategies.
In summ ary, the post hoc analyses f ound support f or di� erent patterns of quality tool deployment in di� erent industries. Moreover, the e� ects of quality tool groups on performance also varied across industries. It is interesting to note that several ® rm perf ormance measures appeared to be signi® cantly a� ected by quality tool deploy- ment. In general, the impact on growth measures appear to be more consistent across the three subsamples , perhaps because quality tool deployment has a proactive impact on perf ormance.
Q uality tool deploym ent patterns 1421
1422 R. Hand® eld et al.
Quality Quality Business Industry A nalysis strategy performance performance
A ll HR tools inspection MS growth (- 0. 247, 0.24) (0.297, 0. 002) design sales growth (0.246, 0.010) (0.271, 0. 005)
R OA growth (0.244, 0. 012) customer relations (0.395, 0. 000)
measurement inspection market share tools (- 0. 228, - 0.037 ) (0.422, 0. 001)
process control product quality (0.236, 0.014) (0.243, 0. 012) process improvement (0.247, 0.010)
design tools market share (- 0.275, 0.027 )
discipline tools customer rejects customer service (- 0.216, 0.032) (0.225, 0. 020)
Food HR tools MS growth (- 0.721, 0.0123 )
measurement process improvement tools (0.654, 0.021) discipline design production cost tools (- 0. 751, 0.005 ) (0.551, 0. 079)
Electronics HR tools sales growth R OA growth (0.665, 0. 002) customer relations (0.564, 0. 012)
measurement design tools (0.500, 0.031) design product quality tools (0.511, 0. 025)
R OA (- 0.627, 0.040 )
discipline R OA tools (0.914, 0. 005)
MS growth (0.536, 0. 018)
A uto HR tools design MS growth (0.626, 0.001) (0.481, 0. 013) inspection sales growth (- 0. 711, 0.000 ) (0.466, 0. 016)
customer relations (0.558, 0. 003)
measurement competitive position tools (0.454, 0. 020) discipline defect market share tools (- 0.503, 0.024) (0.585, 0. 002)
customer service (0.497, 0. 010)
Table 7. Summary of step-wise regression analyses within industries. The independent vari- ables for the stepwise regresson analyses are HR tools, measuement tools, design tools and discipline tools. The dependent variales for which the four independent variables had statistically signi® cant beta estimates are reported in the table. The items in the bracket are standardized beta estimates and p-values, respectively.
6. Discussion
The overall deployment pattern of quality tools appears to involve f our prim ary dimensions. They are: Discipline Tools; Human R esource Tools; Measurement Tools; and Design Tools. These f our dimensions were empirically derived by employing factor analysis on a large sample of diverse ® rms. Clearly, each class of tools is deployed for di� erent reasons. The HR tools serve to empower and involve employees in quality improvement initiatives. Measurem ent tools serve the role of assessing the results of quality improvem ent initiatives. Design tools are deployed to re-engineer processes, and they typically involve multiple f unctions simultaneous ly. Discipline tools are deployed to overhaul an existing quality culture or to instill a new culture. Two themes stand out while considering the f our groups of tools together. First, these tools have long-term implications and the results f rom deploy- ing these tools may not be immediate. Second, there is a f air amount of consistency in the overall pattern of deployment across a large set of ® rms. A t the ® rm level, however, the deployment of quality tools appears to be more complex. Based on our accumulated learning f rom this project, we can speculate that the com mitment to deploy a particular set of tools appears to be made on the basis of the f ollowing consideration s: customer f ocus; desired quality culture; and strategic competitive advantage. On a practical front, issues such as ease-of -use of the tool, inherent complexity, and in som e cases sheer trial-and error learning may dominate the selection and/or disbanding of quality tools.
Measurem ent tools appear to be equally e� ective f or both process control and process improvement strategies. Business perf ormance measures such as product quality, which are largely dependent on internal processes, appear to be a� ected by measurem ent tools. Deploying discipline tools, especially in contexts in which there is a high incidence of custom er rejects and def ect rates, appears to be e� ective. A lso, discipline tools were related to customer-driven measures of quality and business perf ormance. On a moderate level, HR tools could a� ect the reduction of def ect rates. In particular, it appears that empower- ment and involvement related HR tools impact def ect reduction programs. Moreover, growth measures of business performance appear to be a� ected more by HR tools.
In summary, this research suggests that f or re-engineering internal processes, an e� ective strategy is to deploy measurement tools such as SPC, and process capabil- ity. Firms plagued with customer com plaints, def ects, etc., will need the deployment of discipline or companywid e tools such as continuous im provement programs, and total quality managem ent ( TQM). The use of HR tools seems to serve the purpose of a support or enabling f unction as opposed to a stand-alone initiative. Finally, design tools appear to be more conducive to contexts in which the de® nition of quality stresses design quality as opposed to conf ormance quality as in the case of ® rms in the electronics industry. Such ® rms have such short product lif e cycles that the time required to eliminate a bug can seriously harm market share.
Some limitations of this research are worth noting. This paper addresses the joint impact of speci® c tools on a `static’ basis. Hence, the evolution and learning associ- ated with quality tool usage could not be captured. This would require the use of multiple longitudinal case studies. A lso, the impact of quality tools on perf ormance is subject to the conf ounding of extraneous e� ects that could not be controlled in this study. Field studies lack the rigour of control f or extraneous e� ects as is possible in laboratory experiments. This continues to be a pragmatic challenge, as the use of lab
Q uality tool deploym ent patterns 1423
experiments f or studying quality tool usage is prohibitively expensive, especially if one is using objective data f rom companies.
7. Conclu sions
A s we noted in the introduction , the incidence of f ailures of quality initiatives is alarmingly high across a large cross-section of industries. This research attempts to address this anomaly through an investigation of the e� ect of quality tool deploy- ment patterns on perf orm ance. The study contributes to our knowledge of quality managem ent in several ways.
First, the study provides a convenient method of grouping the di� erent quality tools that ® rms use in quality improveme nt e� orts. This study identi® ed f our groups: human resource tools, measurem ent tools, discipline tools, and design tools, to be common across a wide variety of industries. Quality managers can use these group- ings to perf orm an internal assessment of tool deployment.
Second, the study provides support f or the notion that quality deploym ent is a strategic initiative that can contribute signi® cantly to achieving higher perf orm ance. Our study f ound that quality tool groups signi® cantly a� ected several dimensions of perf ormance. These dimensions related to quality perf ormance ( def ects, scrap rates and rework) and also to overall ® rm perf ormance (m arket share, com petitive posi- tion, growth). Quality tool deployment appears to be a prerogative of both quality managers and strategic planning executives.
Finally, the study has shown that patterns of quality tool deploym ent and their impact on perf ormance vary across industries. In particular, this variance can be attributed to the di� erent competitive conditions that ® rms are likely to f ace in di� erent industries. Findings relating to the three sub-samples examined are indica- tive ( though not conclusive ) of possible within-indust ry e� ects.
Several research directions can be o� ered based on our study. A n initial attempt at investigating industry-spe ci® c patterns of quality tool deploym ent and their impact on quality perf ormance was carried out. How ever, we recognize that industry speci® c patterns could perhaps be studied using a larger sample of ® rms f rom one industry. Future research should also study industries other than those considered in this project. Patterns that emerge across di� erent studies within an industry could be more powerf ul evidence f or theory building. Issues relating to causation can be explored using techniques such as path analysis, and structural equation modeling ( SEM).
A ckno wledgme nt
This research was f unded by a grant f rom the center f or International Business Education R esearch at Michigan State University.
R eferences
Ada m, E. E., Jr, 1994, A lternative quality improvement practices and organization perform- ance. Journal of O perations Managem ent, 12, 27± 44.
Aka o, Y., 1990, Q uality Function Deploym ent ( Cambridge, MA : Productivity Press). American Quality Foundation and Ernest & Young, 1992, T he International Quality
Study Best Practices Report: an Analysis of Managem ent Practices that Im pact Performance (Montvale, NJ: National Association of A ccountants).
Anderson, J. C., Rungtusanatham, M. and Schroeder, R. G., 1994, A theory of quality management underlying the Deming management method. Academ y of Managem ent Review, 19, 472± 509.
1424 R. Hand® eld et al.
Boothroyd, G. and Dewhurst, P., 1987, Production D esign for Assembly (Wake® eld, R I: Boothroyd Dewhurst Inc.).
Benson, P. G., Saraph, J. V. and Schroeder, R. G., 1991, The e� ects of organizational context on quality management: an empirical examination. Managem ent Science, 37, 1107± 1124.
Brocka, B. and Brocka, S. M., 1992, Q uality Managem ent: Im plem enting the Best Ideas of the Masters (Burr R idge, IL: R ichard D. Irwin).
Buz z ell, R.D. and Wiersema, F. D., 1981, Successful share-building strategies. S trategic Managem ent Journal, 2, 27± 42.
Churchill, G. A., 1979, A paradigm for developing better measures of marketing constructs. Journal of Marketing Research, 16, 64± 73.
Craig, C. S. and Douglas, S. P., 1982, Strategic factors associated with market and ® nan- cial performance. Q uarterly Review of Econom ics and Business, Summer, pp. 101± 111.
Cronbach, L. J., 1951, Coe� cient alpha and the internal structure of tests. Psychom etrika, 16, 297± 334.
Dale, B. G. and Shaw, P., 1990, Failure mode and e� ects analysis in the UK motor industry: a state-of -the-art study. Q uality and Reliability Engineering International , 4, 179± 188.
Dean, J. W. and Susman, G. I., 1989, Organizing f or manuf acturing design. Harvard Business Review, 67, 28± 36.
Ebrahimpour, M. and Lee, S. M., 1988, Quality management practices of A merican and Japanese electronic ® rms in the United States. Production and Inventory Managem ent Journal, 29, 28± 31.
Fleischer, M. and Liker, J. K., 1992, The hidden professionals: product designers and their impact on design quality. IEEE T ransactions on Engineering Managem ent, 39, 254± 264.
Forker, L. B., Vickery, S. K. and Droge, C. L. M., 1996, The contribution of quality to business performance. International Journal of Operations and Production Managem ent.
Fortuna, R. M., 1990, Quality of design. In T otal Q uality: an Executive Guide f or the 1990’ s (Homewood, IL: Business One Irwin).
Flynn, B. B., Schroeder, R. and Sakakibara, S., 1995, Determinants of quality perform- ance in high- and low-quality plants. Q uality Managem ent Journal, Winter, 8± 25.
Garvin, D. A., 1983, Quality on the line. Harvard Business Review, 61, 64± 75; 1984, Japanese quality management. Colum bia Journal of W orld Business, 19, 3± 12; 1987, Competing on the eight dimensions of quality. Harvard Business Review, 65, 101± 109.
Georgantz as, N. C., Nicholas, C. and Hessel, M. P., 1995, The intermediate structure of designs f or quality. International Journal of Quality and Reliability Managem ent, 12, 97± 108.
Godfrey, B. A., 1993, Ten areas for f uture research in total quality management. Q uality Managem ent Journal, Oct., 47± 70.
Greene, R. T., 1993, Global quality: A S ynthesis O f T he W orld’ s Best Managem ent Methods (Milwaukee, WI.: A SQC Quality Press).
Handfield, R. B., 1989, Quality management in Japan versus the United States: an overview. Production and Inventory Managem ent Journal, 30, 79± 84; 1995, Re-Engineering For Tim e Based Com petition: Benchmarks And Best Practices For Production (Wesport, CT, Quorum Books).
Handfield, R. B. and Ghosh, S., 1994, Creating a quality culture through organizational change: a case analysis. Journal of International Marketing, 2, 7± 36.
Hauser, J. R. and Clausing, D., 1988, The house of quality. Harvard Business Review, 66, 63± 73.
Hayes, R., 1981, Why Japanese factories work. Harvard Business Review, 59, 56± 66. Huge, E. C., 1990, Quality of conformance to design. In T otal Q uality: an Executive Guide for
the 1990’ s ( Homewood, IL: Business One Irwin). Kaplan, R. S. and Norton, D. P., 1992, The balanced scorecard: measures that drive-
performance. Harvard Business Review, 70, 71± 79. Larson, P. D. and Sinha, A., 1995, The TQM impact: a study of quality managers’ percep-
tions. Q uality Managem ent Journal, Spring, 53± 66. Lascelles, D. M. and Dale, B. G., 1988, A study of the quality management methods
employed by U.K. automotive suppliers. Q uality and Reliability Engineering International , 4, 301± 309.
Q uality tool deploym ent patterns 1425
Lee, S. M. and Ebrahimpour, M., 1985. A n analysis of Japanese quality control systems: implications f or A merican manufacturing ® rms. SA M Advanced Managem ent Journal, Spring, 24± 31.
Mann, R. and Kehoe, D., 1994, A n evaluation of the e� ects of quality improvement activities on business performance. International Journal of Q uality and Reliability Managem ent, 11, 29± 44.
McCloskey, L. A. and Collett, D. N., 1993, T Q M: a Basic T ext (Metheun, MA : COA L/ OPC).
Miller, W. H. and Sanders, T. J., 1989, Design for inherent manuf acturability of electronic products. International S em iconductor Manuf acturing Science Sym posium , IEEE, pp. 85± 91.
Miz uno, S., 1988, Managem ent For Quality Im provem ent: T he Seven Q C T ools (Cambridge, MA : Productivity Press).
Modaress, B. and Ansari, A., 1989, Quality control techniques in US ® rms: a survey. Production and Inventory Managem ent Journal, 30, 58± 62.
Montgomery, D. C., 1991, Introduction to Statistical Q uality Control ( New York, NY : John Wiley).
Norusis, M. J., 1990, SPSS Base System User’s Guide. Chicago, SPSS. Pennings, J. M., 1984, Productivity: some old and new issues. In A rthur P. Brief (Ed.)
Productivity Research in the Behavioral and S ocial S ciences (New Y ork: Praeger), pp. 127± 140.
Phillips, L. W., Chang, D. R. and Buz z ell, R. D., 1983, Product quality, cost position, and business performance: a test of some key hypotheses. Journal of Marketing, 37, 26± 43.
Rodriguez , J. R., 1990, Total productive maintenance. In T otal Q uality: An Executive Guide for the 1990’ s ( Homewood, IL: Business One Irwin).
Radhakrishnan, S. and Srinidhi, B., 1994, Should quality be designed in or inspected in? a cost-of -quality framework. Q uality Managem ent Journal, Fall, 72± 85.
Saraz en, J.S., 1990, The tools of quality. Part II: Cause-and-e� ect diagrams. Q uality Progress, July, 59± 62.
Sower, V., Motwani, J. and Savoie, M., 1993, Are acceptance sampling and SPC comple- mentary or incompatible? Q uality Progress, Sept., 85± 90.
Taguchi, G., 1979. Introduction to O� -line Quality Control ( Tokyo: Japanese Standards Association).
Tushman, M. L., 1979, Managing communication networks in R &D laboratories. Sloan Managem ent Review, 20, 37± 49.
Whitney, D. A., 1988, Manufacturing by design. Harvard Business Review, 66, 63± 73. Wilkinson, A., Redman, T. and Snape, E., 1995. New patterns of quality management in
the United Kingdom. Q uality Managem ent Journal, Winter, 37± 51. Wru ck, K. H. and Jensen, M. C., 1994. Science, speci® c knowledge, and total quality man-
agement. Journal of Accounting and Econom ics, 18, 247± 287.
1426 Q uality tool deploym ent patterns