Professional Documents
Culture Documents
ISO 9001 certification motivations can be classified into two main categories: (1)
internal motivations; and (2) external motivations. Internal motivations are related
with genuine organisational improvement goals (productivity, internal communication,
process performance), while external motivations are mainly related to promotional
and marketing issues (customer and market pressures, market share). Some companies
that become certified mostly upon the basis of external motivations define their main
goal as obtaining registration, and thus typically adopt a limited view over the
scope of quality management systems implementation and certification.
Based upon a detailed review of 100 ISO 9001 audit reports, we performed a detailed
statistical comparison between both types of motivations driving companies in their
certification efforts, explored their differences and similarities and derived a statistically
based classification model that was able to predict, for a particular organisation, what
kind of predominant motivation lead to its certification, from information that can be
retrieved from the contents of the corresponding ISO 9001 audit reports.
Keywords: ISO 9001; quality management; classification model; statistics
Introduction
According to the ISO 9001 related literature, a company becomes certified based mainly
upon either internal motivations and/or external motivations. Internal motivations are
present in those companies that are really committed to the continuous improvement of
their internal processes, and therefore aim to achieve effective organisational improvements. External motivations, on the other hand, are related mostly to promotional and
marketing issues, customers and market pressures and market share enlargement goals.
Even though all organisations present both kinds of motivations to some extent, only
one is usually the most predominant and determines the organisations decision to
become ISO 9001 certified. The implementation and certification of a quality management
system should be both an important organisational improvement tool an internal motivation, as well as a marketing and competitive advantage for certified companies an
external motivation (Sampaio, Saraiva, & Guimaraes Rodrigues, 2009). However, the
motivation for doing so is usually dominated by one or the other of both factors mentioned.
In this paper we will illustrate some key results derived from a detailed analysis of a
sample of 100 ISO 9001 audit reports that correspond to Portuguese certified companies.
This sample was randomly selected from a group of companies certified by the leading Portuguese certification body Associacao Portuguesa de Certificacao (APCER, 2006). The
Our main research question was: Can we predict a firms motivation in obtaining ISO
9001 certification form its audit report?. Thus, based upon the previous results, we
proposed, developed and tested a statistical classification model, aimed at predicting the
companies ISO 9001 certification main motivation from their audit report profiles. Such
a model has as its major inputs variables that can be identified and derived from an ISO
9001 audit report, and allowed us to identify and predict if a particular company faced
quality management system implementation and ISO 9001 certification as a real commitment internal motivations, or, on the other hand, if the company became certified mostly
because of promotional and marketing issues external motivations.
Quantitative results allowed us to evaluate the classification performance of our model,
both through training and testing sets of data, confirming its statistical significance and
validity. We used the Statistical Package for Social Sciences 15.0 (SPSS) to perform
the statistical analyses.
APCER is a private Portuguese organisation dedicated to the certification of management systems, services, products and people as a method of guaranteeing quality and
promoting the competitive advantage of organisations, whether public or private, national
or international.
In Portugal, APCER is the clear market leader. More than 4500 certificates of conformity have been issued since its foundation, including the certification of organisations in
Spain, Morocco, Mozambique, Angola, Brazil and China (Macao).
APCER is the only Portuguese certification entity representing the international
network IQNet The International Certification Network, which bestows immediate
international recognition on organisations certified by APCER.
Literature review
ISO 9001 certification motivations and benefits
ISO 9001 certification motivations can be classified according to one of two main
categories: internal and external motivations. Internal motivations are related to the goal
of achieving organisational improvement, while external motivations are mainly related
to promotional and marketing issues, customer pressures and market share gains (Brown,
van der Wiele, & Loughton, 1998; Bryde & Slocock, 1998; Buttle, 1997; Corbett, Luca,
& Pan, 2003; Douglas, Coleman, & Oddy, 2003; Escanciano, Fernandez, & Vasquez,
2001; Gonzalez, 2001; Gotzamani & Tsiotras, 2002; Gustafsson, Klefsjo, Berggren, &
Granfors-Wellemets, 2001; Jones, Arndt, & Kustin, 1997; Lee & Palmer, 1999; Lipovatz,
Stenos, & Vaka, 1999; Llopis & Tar, 2003; Magd & Curry, 2003; Mo & Chan, 1997;
Torre, Adenso-Diaz, & Gonzalez, 2001; Poksinska, Dahlgaard, & Antoni, 2002).
Internal benefits
B
B
B
B
B
B
B
B
B
B
B
B
productivity improvements;
product defect rate decreases;
quality awareness improvements;
definition of the personnel responsibilities
and obligations;
delivery times improvements;
internal organisation improvements;
nonconformities decreases;
customers complaints decreases;
internal communication improvements;
product quality improvement;
competitive advantage improvement;
personnel motivation.
(1)
According to Equation (1), E(Y|x) can take any value as x ranges from 21 to +1.
However, with dichotomous data the conditional mean must be greater than or equal to
p(x) =
exp(b0 + b1 x1 + ... + bi xi )
1 + exp(b0 + b1 x1 + ... + bi xi )
(2)
Performing the logit transformation in terms of p(x), the logistic regression model is
defined as:
p(x)
) = b0 + b1 x1 + ... + bi xi
(3)
g(x) = ln(
1 p(x)
The importance of the logit transformation (2) is that g(x) has many of the desirable
properties of a linear regression model. The logit is linear in its parameters, may be continuous and range form 21 to +1, depending on the range of x (Hosmer & Lameshow, 1989).
One of the most important differences between the linear regression model and the
logistic one is that in the linear regression model we assume that an observation of
the outcome variable may be expressed as y E(Y|x) + 1. The quantity 1 is called the
error and expresses an observations deviation from the conditional mean. The most
common assumption is that 1 follows a normal distribution with mean zero and some
variance that is constant across levels of the independent variable. It follows that the
conditional distribution of the outcome variable given x will be normal with a mean of
E(Y|x), and a variance that is constant. When the outcome variable is dichotomous, we
may express the value of the outcome variable given x as y p(x) + 1. Here the quantity
1 may assume one of two values (Hosmer & Lameshow, 1989):
.
.
Thus, 1 has a distribution with mean zero and variance equal to p(x)[1 2 p(x)]. The
conditional distribution of the outcome variable follows a binomial distribution with
probability given by the conditional mean, p(x).
According to Hosmer and Lameshow (1989), when the outcome variable is dichotomous it is important to point out that:
.
.
.
A more detailed overview of the NC, as they are related with the different clauses and
sections of the ISO 9001 standard, is presented below.
NC
Documentation requirements
Management review
Human resources
Product and service provision
Monitoring and measurement
64
24
34
64
66
Total
252
Number of
companies
NC
Average NC/
Company
2
2
7
8
22
17
49
51
11.00
8.50
7.00
6.38
12
3
10
5
5
12
7
2
9
4
1
1
7
1
2
1
72
17
53
24
23
55
32
9
39
17
3
3
20
2
2
0
6.00
5.67
5.30
4.80
4.60
4.58
4.57
4.50
4.33
4.25
3.00
3.00
2.86
2.00
1.00
0.00
3
16
12
29
NC (Average)
4.83
5.03
5.04
5.23
Figure 3. Histogram for the number of non conformities found in each company.
Classification model
Based on a detailed analysis of the 100 audit reports, we have developed a classification
model aimed at predicting the companies ISO 9001 certification motivations from their
audit report contents. Such a model has as its major inputs critical variables that can be
identified from an audit report and allow us to identify if a company faces ISO 9001
certification as a real commitment or, on the other hand, if it became certified mostly
because of promotional or marketing issues. Quantitative results will allow us to evaluate
the classification performance of this model, both through training and testing sets of data
(see Figure 4).
As was already stated, a company can become ISO 9001 certified based upon two main
motivation categories: internal motivations and external motivations. The aim of our
model is to predict the companys main motivation for ISO 9001 certification from a
set of input variables that can be found from audit reports.
Data gathering
During the first phase of our model development, we have collected different opinions
about each company, from people that we consider to be relevant concerning a fair
evaluation of the company main motivations for achieving certification (Figure 5). We
asked each of them to classify the corresponding companies according to their opinion
concerning the main ISO 9001 certification motivation.
We exhaustively analyzed all of the 100 audit reports and classified each one of the
companies according to their predominant certification motivation based on their audit
reports profiles. The classification protocol used to classify those companies was based on
the identification of common patterns (variables) that could be defined as inputs to classify
companies according to their ISO 9001 predominant motivation.
As already stated, we have also collected the corresponding opinion from other people
who have specific knowledge about the given companies. For that purpose, we have interviewed each certified company Process Manager and two Coordinator Managers (from the
companies certification body), as well as the Audit Team Members. The Process Manager
is the person who is responsible for the management of the certification process at the certification body and is close to, and has a detailed knowledge of, each one of the sampled
companies. We contacted all APCERs Process Managers, who classified the companies
they were responsible for. The Coordinator Manager is the person who manages a team
of Process Managers, and does not have specific knowledge about a large set of companies. Generally, the Coordinator Manager is someone with wide experience in ISO
9001 audits. The last interviewed group were the audit teams that were involved in the
sampled companies last audits. In the Process and Coordinators Managers groups the
response rate was 100%. However, in the auditors group we have only reached a response
rate of 78%. Overall, this means that a total of 92 auditors contributed information
regarding the main motivations associated with this set of ISO 9001 certified companies.
After collecting all the responses, we assigned to each company its final classification
(observed value, y) corresponding to either Internal Motivations (y 1) or External
Motivations (y 2). In order to define each company classification, we computed the
average of the classifications attributed by the interviewed people (Research Team,
Process Manager; Coordinator Manager and Auditors) to each of the 100 sampled
companies.
After categorising all the companies that belong to our sample, we found that for 48
companies a consensual classification, of either 1 Internal Motivations or 2 External
Description
Did the company become certified based on customers, market pressure, or promotional
aspects?
How many non conformities were identified in the last ISO 9001 audit?
Did the company present non conformities related to the development of its quality
management system or that interfere with a continuous improvement philosophy?
Did the company present non conformities related to other past audits that were not yet
corrected?
Was ISO 9001 certification used by the company in order to get public funds or to achieve
public contracts?
Did the company have major non conformities in the last ISO 9001 audit?
Is top management involved and committed with the quality management system?
Is the quality management system implemented only to fulfil the minimum ISO 9001
requirements?
Was the quality management system implemented and certified in order to improve the
companys internal processes and internal organisation?
Was ISO 9001 certification imposed by the company headquarters?
What kind of relationship does the company have with the certification body Process
Manager?
Did the company present in the last ISO 9001 audit non conformities related to statutory
and regulatory requirements?
Do the human resources present knowledge and competences related with the quality
management system?
We would like to point out that x7, x9 and x11 are categorical variables, with three possible response levels. However, for Model 1, variables x10 and x12 were not analyzed,
because they have assumed a constant value for all the sampled companies.
According to Table 6, we can verify that for Model 1 only x1; x2; x3; x4 and x13 should
be considered in our classifier (p values smaller than 0,05). Concerning Model 2, the
variables to be kept for classification purposes are the following: x1; x2; x3; x4, x5; x7(2);
x8; x9(2); x13.
As a result of this initial univariate analysis, we identified which of the 13 variables
present a significant relationship with the independent variable. Our next step consisted
of the development of a multivariate logistic regression model, using the previously
selected variables. For that purpose, we use a Backward Stepwise procedure as a variable
selection technique, leading to the final models described in Tables 8 and 9.
The next step comprised an evaluation and validation of these models, through the use
of a receiver operating characteristic (ROC) curve methodology and the Pyramid Population graphic, in order to evaluate the models performance. Concerning the Pyramid
SE (b )
Wald
Exp(b )
p value
2.959
0.532
2.923
2.457
21.961
21.111
219.123
222.845
23.059
221.673
242.406
2
222.072
242.406
2
22.269
0.887
0.170
1.140
0.887
40192.970
17300.440
28420.702
28420.702
12118.637
12118.637
14634.645
2
23205.412
46410.839
2
0.786
11.124
9.853
6.578
7.666
0.000
0.000
0.000
0.000
0.000
0.000
0.000
2
0.000
0.000
2
8.331
19.286
1.703
18.600
11.667
3.0E + 09
1.0E + 09
0.000
0.000
1.0E + 10
0.000
0.000
2
0.000
0.000
2
0.103
0.001
0.002
0.010
0.006
1.000
0.999
0.999
0.999
0.998
0.999
0.998
2
0.999
0.999
2
0.004
SE (b )
Wald
Exp(b )
p value
1.504
0.374
2.134
1.771
2.207
21.104
20.944
22.632
2.193
21.014
23.997
21.524
221.560
242.406
21.497
21.360
0.457
0.100
0.809
0.622
1.103
13761.628
1.185
1.114
0.682
0.726
0.876
28420.722
200096.496
44937.111
40192.970
0.476
10.845
13.981
6.954
8.109
4.003
0.000
0.635
5.575
10.327
1.952
20.840
0.000
0.000
0.000
0.000
8.144
4.500
1.454
8.448
5.875
9.091
1.0E + 09
0.389
0.072
8.960
0.363
0.018
2.0E + 09
0.000
0.000
2.0E + 09
0.257
0.001
0.000
0.008
0.004
0.045
0.999
0.426
0.018
0.001
0.162
0.000
0.999
0.999
0.999
1.000
0.004
SE (b)
Wald
Exp(b)
p value
Constant
x1
x2
x3
x4
28.049
4.635
0.780
5.763
3.437
2.913
1.876
0.353
2.657
1.570
7.634
6.106
4.888
4.706
4.794
0.000
102.900
2.182
318.428
31.090
0.006
0.013
0.027
0.030
0.029
SE (b)
Wald
Exp(b)
p value
Constant
x1
x2
x3
x4
x5
x9(1)
x9(2)
23.053
1.437
0.312
2.222
2.140
2.445
0.390
21.862
1.437
0.719
0.145
1.266
0.862
1.433
1.008
1.118
4.513
3.989
4.610
3.079
6.163
2.913
0.150
2.773
0.047
4.206
1.366
9.223
8.498
11.532
1.477
0.155
0.034
0.046
0.032
0.079
0.013
0.088
0.699
0.096
Population graphics, we used a cut value of 0.5 for the predicted probability (y). Therefore,
all companies that presented a predicted probability smaller than 0.5 were classified as
having Internal Motivations, while those companies that presented a predicted probability
higher than 0.5 were classified as having External Motivations. The software predicted
probability of 1 corresponds to an observed value of 2 (External Motivations), and the predicted probability of 0 corresponds to an observed value of 1 (Internal Motivations).
As illustrated in Figure 6, the results obtained with Model 1 are better than those for
Model 2, as should be expected, since companies used to develop Model 1 were those
companies with an exact observed value. These results are also reinforced with the
model ROC curves. As we can verify from Figure 7, Model 1 presents an area under
Figure 6. Pyramid Population for estimated values in both Model 1 and Model 2.
Model 2a
Model 2b
Model 2c
x1
x2
x3
x4
x5
x9(1)
x9(2)
0.017
0.002
0.009
0.003
0.034
N/A
N/A
0.007
0.056
0.034
N/A
0.569
0.002
0.002
0.003
0.017
0.006
N/A
N/A
N/A
ROC
0.897
0.912
0.882
the curve of 0.978 and Model 2 presents an area of 0.926, with only 2% of the sampled
companies not being correctly classified under Model 1.
We have also analyzed other possible alternative solutions for Model 2, with the aim of
finding a model which enables us to classify companies and predict their ISO 9001 motivations by only making use of information that is readily available and can be directly
extracted from audit reports. For that purpose, we have tested the following models,
based upon Model 2, but with different sets of independent variables being considered
Model 2 alternatives, as shown in Table 10.
In Model 2a we have excluded x9 from the set of predictors, leading to an area under
the ROC curve of 0.897. Concerning Model 2b, we have excluded x5, and found that x2 and
x9(1) were not significant at a 0.05 significance level. Regarding Model 2c, we have
excluded x5 and x9 from the underlying model. The ROC curve values, for the two
previous models, are, respectively, 0.912 and 0.882.
0.05145
0.08010
0.06554
0.09595
0.02862
0.01407
0.04447
Model 2
0.92607
0
20.01455
0.01585
Model 2a
0.89744
1
0.03040
Model 2b
0.91200
5
79.6%
41.6%
Model 2a
28.0%
11.6%
81.6%
45.6%
39.6%
Model 2b
32.8%
Model 2c
0,88160
75.2%
Model 2
0,0%
18.8%
0.0%
83.6%
52.4%
35.6%
42.0%
Model 2c
The last result is related to the application of a bootstrap test, in order to indentify
statistical significant differences between the ROC curves. According to Table 13, we
can verify that only for the case of Model 1 were we able to find significant statistical
differences, when compared with the remaining models.
According to the results obtained, we can conclude that Model 1 has the best performance, in order to predict why companies become ISO 9001 certified, if based mostly on
internal motivations or on external motivations. However, one must remember that
Model 1 was developed based on companies that reached a precise observed value, i.e.
companies for which the evaluation teams did share a complete consensus about their
dominant ISO 9001 motivation.
Next, we further evaluated the model performance over two different data sets. One set
comprises the training data, which correspond to the companies with a precise observed
0.000
Model 2
No Sig.
No Sig.
No Sig.
0.000
0.973
Model 2a
No Sig.
No Sig.
0.000
0.875
0.999
Model 2b
No Sig.
0.000
0.124
0.999
0.991
Model 2c
ROC Curve
% of wrong classifications
% of wrong classifications
With x1
Without x1
0.978
12.50
26.20
0.940
16.67
31.00
categorisation value (n 48). For those companies, we have performed an internal model
validation, using cross validation. We have also tested the model performance over the
group of companies with a non exact consensual observed dependent value (n 42).
We have performed such additional validations under two different situations. First, we
have validated the model using all the variables that composed it (x1; x2; x3; x4), and
second using only variables that can be found directly and easily extracted from audit
reports (x2; x3; x4).
According to Table 14, we can verify that the model performance is better with x1, but
there is not a significant difference between the two described situations. As expected, the
percentage of wrong classifications was smaller with our gold standard data, when
compared with the results obtained with the application of the model over companies
without a consensual dependent variable definition.
Therefore, the final mathematical expressions for the models developed are as follows:
p(x) =
(4)
(5)
p(x) =
with (4) representing the model with x1 and (5) the model without x1.
Conclusions
According to the literature, ISO 9000 certification motivations can be classified into two
main categories: internal motivations and external motivations. Internal motivations are
related to genuine organisational improvement goals (productivity, internal communication, internal processes performance improvement), while external ones are mainly
related to promotional and marketing issues (customer and market pressures, market
share improvement). Some companies that become certified based mostly upon external
motivations defined their main goal as obtaining registration, and thus are of a very
limited nature in terms of quality management systems implementation and certification.
From the same sets of data, together with opinions collected from auditors and other
experts, we were able to derive statistically sound and valid classifiers, which allow us
to predict if a given company follows mostly internal or external motivations from information contained in the corresponding audit reports.
Acknowledgements
The authors acknowledge financial support provided by Fundacao para a Ciencia e a Tecnologia (FCT) through research grant (BD/16032/2004).
References
Associacao Portuguesa de Certificacao (APCER). (2006). Retrieved from http://www.apcer.pt
Bhuiyan, N., & Alam, N. (2004). ISO 9001: 2000 implementation the North American experience.
International Journal of Quality & Reliability Management, 53(1), 1017.
Braga, A.C, Costa, L.A., & Oliveira, P.N. (2003). Alternativa para comparacao de duas curvas ROC
na avaliacao de sistemas diagnostico. Proceedings of the Annual Congress of the Portuguese
Society of Statistics, Faro, Portugal.
Braga, A.C., Costa, L.A., & Oliveira, P.N. (2004). Metodologia nao parametrica para a comparacao
global e parcial de curvas ROC. Proceedings of the Annual Congress of the Portuguese
Society of Statistics, Evora, Portugal.
Brown, A., van der Wiele, T., & Loughton, K. (1998). Smaller enterprises experiences with ISO
9000. International Journal of Quality & Reliability Management, 15(3), 273285.
Bryde, D., & Slocock, B. (1998). Quality management systems certification: A survey. International
Journal of Quality & Reliability Management, 15(5), 467480.
Buttle, F. (1997). ISO 9000: Marketing motivations and benefits. International Journal of Quality &
Reliability Management, 14(9), 936947.
Casadesus, M., Gimenez, G., & Heras, I. (2001). Benefits of ISO 9000 implementation in Spanish
industry. European Business Review, 13(6), 327335.
Casadesus, M., Heras, I., & Arana, G. (2004). Costes y beneficios de la implantacion de la normative
de calidad ISO 9000. Evolucion temporal. Proceedings of the XIV Congreso Nacional
ACEDE, Murcia, Spain.
Copyright of Total Quality Management & Business Excellence is the property of Routledge and its content
may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express
written permission. However, users may print, download, or email articles for individual use.