You are on page 1of 15

The Journal of Systems and Software 66 (2003) 7–21

www.elsevier.com/locate/jss

A quantitative and qualitative analysis of factors affecting


software processes
Austen Rainer *, Tracy Hall
Department of Computer Science, University of Hertfordshire, Hatfield Campus, Hatfield, Hertfordshire AL10 9AB, UK
Received 10 March 2001; received in revised form 12 November 2001; accepted 2 March 2002

Abstract

Despite the growing body of research on software process improvement (SPI), there is still a great deal of variability in the
success of SPI programmes. In this paper, we explore 26 factors that potentially affect SPI. We also consider the research strategies
used to study these factors. We have used a multi-strategy approach for this study: first, by combining qualitative and quantitative
analysis within case studies; second, by comparing our case study results with the results of a previously conducted survey study.
Seven factors relevant to SPI (i.e. executive support, experienced staff, internal process ownership, metrics, procedures, reviews, and
training) were identified by the case studies and the survey study. Two factors (reward schemes and estimating tools) were found, by
both the case studies and the survey study, not to be relevant to SPI. Three additional factors (people, problems and change) were
identified by the case studies. The frequency with which people, problems and change are discussed by practitioners suggests that
these three factors may be pervasive in SPI, in a way that the other factors are not. These factors, however, require further in-
vestigation.
 2002 Elsevier Science Inc. All rights reserved.

1. Introduction disillusionment over the lack of resultsÕ, and another


26% report that Ônothing much has changedÕ. Cattaneo
A growing number of studies have examined the fac- et al. also point out similar results from the SPICE trials
tors that affect software development projects and soft- (SPICE, 1998, p. 159): 54% of respondents believe that
ware process improvement (SPI) programmes. These SPI is costing more than they had anticipated; 64% do
studies include surveys of practitionersÕ opinions (Kitson not believe that the assessment has had a major impact
and Masters, 1993; Brodman and Johnson, 1994; on the organisation; and only 28% disagree with the
Herbsleb et al., 1994; Goldenson and Herbsleb, 1995; statement that Ônothing much has changed since the
Hayes and Zubrow, 1995; Herbsleb and Goldenson, assessmentÕ.
1996; El Emam et al., 1999; Paulk et al., 2000; Rainer and A number of explanations for the mixed results of
Hall, 2002), case study reports of SPI in particular process assessments and SPI are available. For example,
companies (see Paulk (2000) for a bibliography of case the authors reporting the results of the SPICE trials
studies), and reviews of previous studies (e.g. Stelzer and suggest two explanations: first, that the expectations of
Mellis, 1998; Krasner, 1999). those involved in SPI may not have been managed
Despite the number of studies, there is still a great properly; second, that insufficient resources were made
deal of variability in the success of SPI programmes. As available. By contrast, Cattaneo et al. suggest a third
Cattaneo et al. (2001) point out, while Herbsleb and explanation: that most SPI efforts limit their attention to
Goldenson (1996) report quite positive results for the technical and engineering aspects of software de-
CMM-based SPI, 49% of their respondents had Ôa lot of velopment, and do not consider the relationships be-
tween these engineering factors and organisational and
market factors.
*
Corresponding author. Tel.: +44-1707-284-763; fax: +44-1707-284- We would like to consider two issues that relate to the
303. variable success of SPI: the factors that affect software
E-mail address: a.w.rainer@herts.ac.uk (A. Rainer). processes, and the research strategies that are used to

0164-1212/03/$ - see front matter  2002 Elsevier Science Inc. All rights reserved.
doi:10.1016/S0164-1212(02)00059-6
8 A. Rainer, T. Hall / The Journal of Systems and Software 66 (2003) 7–21

investigate these factors. We are interested in research 2. Background


strategies because they influence the way in which a
phenomenon is investigated, and therefore the kinds of For our survey study (previously reported in Rainer
findings that can be made, and the reliability of those and Hall (2002)), we conducted three sets of analysis (i.e.
findings (Yin, 1994). As an example, Herbsleb and a general analysis, a maturity-based analysis, and a
Goldenson (1996) appear to rely exclusively on a survey success-based analysis) against three data sets (compa-
strategy. By contrast, the SPICE trials appear to rely nies with appraised maturity, companies with self-rated
principally on a survey strategy, but supplement this maturity, and companies with no rating of maturity).
data with observation reports, problem reports, and We conducted this analysis to identify those factors that
Ôfield experimentsÕ. A multi-strategy approach, such as have a major impact, or no impact, on implementing
that used in the SPICE trials, can compensate for the SPI. Factors were identified as having a major impact
inherent weaknesses within any particular research (or no impact) on SPI if at least 50% of the respondents
strategy (Yin, 1994). in the sample(s) reported that the factor had a major
As part of our investigations of SPI, we have adopted impact (or no impact). Factors which the survey study
an approach similar to that of the SPICE trials. We have identified as having a major impact on SPI are presented
conducted both a survey study (Rainer and Hall, 2002) in Table 1. There was some evidence that two factors
and case studies of particular companies (Hall et al., (reward schemes and estimating tools) did not have an
2000b, 2001; Rainer et al., 2001). This allows us to ex- impact on SPI.
plore the relative benefits of survey studies and case As part of our survey investigation, we also reviewed
studies for SPI. three previous studies, each of which has investigated
In this exploratory paper, we quantitatively and process standards that are recognised internationally:
qualitatively analyse the words used by practitioners in
our case study group interviews to gain insights into the • Goldenson and HerbslebÕs (1995) survey of CMM-
factors that these practitioners think affect software appraised organisations in North America.
processes. We compare these analyses with the findings • El Emam et al.Õs (1999) survey of companies involved
from our survey study. This multi-strategy approach in the first phase of the SPICE trials.
should reveal insights into both software processes and • Stelzer and MellisÕ (1998) review of experience reports
into research strategies. and case studies of SPI and quality systems initia-
To focus our study, we investigated the following tives.
research questions:
Factors identified by these three studies as being as-
RQ1 What factors, as identified by our case studies, af- sociated with successful SPI programmes are presented
fect software processes? in Table 2. Goldenson and Herbsleb also identified a
RQ2 What factors, as identified by our case studies, do number of factors associated with unsuccessful SPI
not affect software processes? programmes: organisational politics, turf guarding,
RQ3 Are the factors identified from our survey study discouragement and cynicism from previous unsuccess-
consistent with the factors identified from our case ful improvement experience, belief that SPI gets in the
studies? way of real work, and the need for more guidance on
RQ4 Regardless of research strategy, how do the fac- how, not just what, to improve. As noted earlier, we
tors relate to each other? found in our survey study that reward schemes and es-
RQ5 Is a multi-strategy approach a potentially useful timating tools may not have an impact on SPI pro-
approach to investigating software development grammes.
and SPI? There are some similarities between the findings of
the four studies. In particular, all four studies recognise
The remainder of the paper is organised as follows. In the contribution of senior management and the in-
Section 2, we review previous work, including our sur- volvement of staff in the improvement effort (which we
vey study, to justify our multi-strategy approach. In have termed internal process ownership in our survey
Section 3, we discuss the design of this study. Section 4 study). Some of the factors identified by Stelzer and
presents the results of our quantitative analysis of 18 Mellis suggest ways in which other factors may be im-
ÔcoreÕ factors. Section 5 compares the results of our plemented. For example, encouraging communication
quantitative analysis with the results of the survey study. and collaboration, and recruiting and promoting change
Section 6 presents the results of our quantitative analysis agents and opinion leaders from the technical staff, may
of eight additional factors. Section 7 provides a quali- be ways to involve staff in SPI.
tative analysis of some of the major factors that emerged There are also a number of differences between the
from the quantitative analysis. Section 8 evaluates our findings, mainly due to the fact that different studies
multi-strategy approach. Section 9 presents conclusions. have investigated different factors. These differences
A. Rainer, T. Hall / The Journal of Systems and Software 66 (2003) 7–21 9

Table 1
Factors having a major impact on SPI programmes (Rainer and Hall, 2002)
Factor Generala Maturityb Success
High Low Appraised Self-rated Not-rated
Training and mentoring Yes Yes Yes Yes Yes
Reviews Yes Yes Yes
Standards and procedures Yes Yes
Experienced staff Yes Yes
Inspections Yes Yes
Internal process ownership Yes Yes
Internal leadership Yes
Executive support Yes
Metrics Yes
Total 4 8 2 2 2 1
A blank cell indicates that the factor was not identified as having a major impact by the respective sample.
a
The general analysis makes no distinction between appraised, self-rated and not-rated data.
b
The high-maturity and low-maturity samples include appraised and self-rated data.

Table 2
Factors affecting SPI, as identified by other researchers (Rainer and Hall, 2002)
Factor Goldenson El Emam Stelzer
Senior management commitment Yes Yes Yes
Clear and relevant SPI goals Yes Yes Yes
Clear, compensated assignment of responsibility for SPI Yes Yes
Staff involvement Yes Yes Yes
SPI people highly/well respected Yes Yes
Staff time and resource. Yes Yes
Creating process action teams Yes
Change agents and opinion leaders Yes
Encouraging communication and collaboration Yes
Managing the SPI project Yes
Providing enhanced understanding Yes
Stabilising changed processes Yes
Tailoring improvement initiatives Yes
Unfreezing the organisation Yes
An ÔemptyÕ cell indicates that the factor was not studied by the respective researchers.

indicate that SPI researchers are still seeking to identify As already noted above, the body of case studies and
the key factors that affect SPI programmes. Dyb a survey studies on SPI is suggesting a large number of
(2000), for example, claims to have identified nearly 200 factors that might affect the success of SPI. It is possible
prescriptions for success, from an extensive review of to compare these studies (e.g. Stelzer and Mellis, 1998)
quality management, organisational learning and SPI in an effort to abstract more reliable and more widely
literature. Our investigations are a contribution to this applicable findings. One inevitable problem with such
search. comparisons, however, is that there is no coherent,
All four studies use a survey strategy, although in overall design to the studies being reviewed (e.g. because
different ways. Goldenson and Herbsleb, El Emam they were conducted by different researchers, at different
et al., and our study are all questionnaire surveys. By times, in different settings, with different objectives). A
contrast, Stelzer and Mellis survey previous literature. multi-strategy design (such as that of the SPICE trials)
In addition to these survey studies, there are many case allows a comparison of results from different research
studies of SPI. Table 3 summarises 39 case reports, or- strategies within a coherent, overall research design.
ganised according to the companies (cases) investigated.
(We consider these to be important publications because
the 11 organisations include all five of the organisations 3. Study design
that have been awarded the IEEE Computer Society
Award for Process Achievement. In addition, 12 of 14 3.1. An overview of the design
sites within these companies have been assessed at
CMM Level 3 or higher. We have examined these 39 We conducted group interviews with three types
cases in more depth in Rainer and Hall (2001)). of practitioners (i.e. developers, project managers and
10 A. Rainer, T. Hall / The Journal of Systems and Software 66 (2003) 7–21

Table 3
Publications relating to 11 key organisations in SPI
Organisation Publications
Advanced Information Services Ferguson et al. (1999)
Boeing Fowler (1997), Kness and Satake (1997), Wigle and Yamamura (1997), Yamamura and Wigle (1997),
Yamamura (1999) and Raman (2000)
Hughes Humphrey et al. (1991) and Willis et al. (1998)
Motorola Daskalantonakis (1992, 1994), Diaz and Sligo (1997) and Fitzgerald and OÕKane (1999)
NASA SEL McGarry et al. (1994), NASA (1994), SEMATECH (1995), NASA (1996) and Pajerski and Basili (1997)
Oklahoma City Air Logistics Butler (1995, 1997) and Butler and Lipke (2000)
Center
Raytheon Dion (1992, 1993), Haley et al. (1995), Haley (1996), Oshana (1998) and Raytheon Corporate Communi-
cations (1998)
Schlumberger Wohlwend and Rosenbaum (1993), Wohlwend and Rosenbaum (1994) and Schlumberger (2000)
Siemens Paulish and Carleton (1994), Gonauser et al. (1998) and Mehner et al. (1998)
Space Shuttle Software Project Billings et al. (1994), Krasner et al. (1994), Paulk et al. (1994) and Florac et al. (2000)
Telcordia Technologies Ferrera and Timko (2000) and Pitterman (2000)

senior managers) in four telecommunications compa- training and mentoring) were separated into four fac-
nies. This is part of a wider study, involving group in- tors, to allow for a more detailed examination. This gave
terviews with 13 companies and a survey of over 80 us 18 ÔcoreÕ factors to investigate.
companies. (See Rainer et al. (2001) for more detail.)
Each group interview was attended by between four and 3.2.2. Additional factors
six practitioners of the same type i.e. four to six devel- To more thoroughly understand the factors that im-
opers, or four to six project managers, or four to six pact SPI programmes, and in recognition that SPI
senior managers. We used a template of questions to programmes vary in their success, we also examined
direct the group discussion. Each interview was taped eight additional factors. These factors were chosen be-
and subsequently transcribed. In quantitatively analy- cause they relate to recurring issues in the software
sing the transcriptions, we chose to investigate 26 fac- process research community. SPI is directed at reducing
tors, and used sets of words to represent each factor. costs and schedule and improving quality, so we wanted
Our criterion for a factor relevant to SPI was that one or to see whether practitioners discussed these issues. SPI is
more of the factorÕs associated words should occur in also concerned with identifying and resolving problems
50% or more of the transcriptions. We compared those within the development process, and SPI responds to,
factors identified in the case studies with those factors requires, and also initiates, change. Change may come
identified from our survey study. We also qualitatively from the market, so we investigated markets. Finally,
analysed some of these factors to gain some general idea there are relationships between process and people e.g.
of how they might relate to each other. the creation, implementation and conformance to pro-
cesses depend on people.
3.2. Factors investigated
3.3. Sample
This study has investigated the presence of 26 factors
within software process and SPI. In referring to ÔfactorsÕ, We collected survey data from 84 companies, and
we are not making a distinction between causes and group interview data from 13 companies (these 13
effects, but are referring to the presence of 26 issues. For companies being a subset of the 84 companies). Focus-
example, two of the factors that we investigate are ex- ing on companies that develop a particular type of
perienced staff and quality. It may be that there is a software system allowed us to control for some of the
causal relationship between experience (perhaps tradi- organisational and commercial issues surrounding soft-
tionally known as a factor) and quality (an outcome). ware development. Our subsample still allowed a variety
We are not, however, investigating such relationships of maturity levels, and a reasonable Ôsample sizeÕ of
within this study. The results of this study may be useful interviews (15 in total). We chose to concentrate on
in suggesting a candidate set of factors within which to telecommunications companies because telecommuni-
subsequently search for cause–effect relationships. cations is a leading commercial market, resulting in the
requirement for, and development of, Ôleading edgeÕ
3.2.1. Core factors technology.
Our survey study examined the impact of 16 factors. Table 4 provides some characteristics of the four
Two of these factors (standards and procedures, and companies, taken from responses to the survey study.
A. Rainer, T. Hall / The Journal of Systems and Software 66 (2003) 7–21 11

Table 4
Characteristics of the four cases
Company CMM Level CMM assessment method Formal approach to SPI Successful SPI Number of group interviews
1 1 Self-estimate Yes Unknown 5
2 1 Self-estimate Yes Unknown 5
3 4 Formal appraisal Yes Yes 3
4 3 Formal appraisal Yes Marginally 2

(See Appendix D for further information on some of the Natural language analysis has been discussed in the
responses reported in Table 4.) All four companies are context of software development for a number of years.
ISO 9001 certified and have had SPI programmes for For example, almost two decades ago, Abbott (1983)
over 5 years. Companies 1 and 2 are sites within one provided advice on using nouns to identify data types.
larger company, which is over 50 years old, and the two Others having developed such ideas further (e.g. Chen,
companies each develop both hardware and software. 1983; Saeki et al., 1989; Carley, 1997; Misic and Russo,
By contrast, companies 3 and 4 are within separate 2000). Because the analysis of natural language is a
larger organisations, both of these organisations being relatively novel approach to investigating the software
between 10 and 20 years old. process, we have looked outside of the software engi-
While companies 1 and 2 claim to have formal SPI neering research literature to gather advice on this ap-
programmes, and use the CMM in full, they have not proach. In particular, we have drawn upon: BromleyÕs
formally assessed their maturity levels. This may be (1977) account of analysing ordinary language descrip-
because they are using the CMM as preparation for an tions of personality; HolstiÕs (1969) guide to content
assessment, or it may be that the companies do not think analysis as an approach to documentary research;
that an assessment is necessary. Another possibility is StraussÕs (1987) handbook for qualitative analysis for
that informal assessments suggest that the company has social science; and Miles and HubermanÕs (1994)
not matured to CMM Level 2, and so a formal assess- sourcebook of qualitative data analysis.
ment would not be considered worthwhile. It is not
possible to tell from the data whether companies 3 and 4 3.4.1. Units of analyses
have been using the CMM for the duration of their SPI We have used two main units of analyses: the group
programmes, or have only recently started to use the interview (transcript) and the company. The group in-
CMM. Their maturity levels (4 and 3 respectively) terviews provide a finer-grained unit of analysis, and we
would suggest that they have been using the CMM for have aggregated interviews to establish a company unit
some time (cf. Hayes and Zubrow, 1995). of analysis. For the comparison of the results from the
Company 4 has been formally assessed at CMM survey study with these results, we have used a third unit
Level 3 and yet their SPI programme is only marginally of analysis: the study.
successful (according to the respondentÕs answer to a
particular survey question). Perhaps company 4Õs ob- 3.4.2. The interviews
jectives are to attain Level 4 or Level 5 and so Level 3 is The researchersÕ questions for the group interviews
judged as ÔonlyÕ marginally successful. The variability of asked about various aspects of the software develop-
SPI success in our four telecommunications companies ment process and SPI. These questions were open-ended
ÔmapsÕ well to the variability of SPI, as reported by and intended to give practitioners freedom to discuss
Herbsleb and Goldenson, and the SPICE trials. issues within the context of a particular aspect of soft-
Responses from the questionnaire also indicate that ware development. (See Appendix B for more informa-
company 4 makes full use of an alternative model to the tion.) Given that these were group interviews, efforts
CMM. Companies 1, 2 and 3 make partial use of an were made to compensate for undesirable group dy-
alternative model to the CMM. namics. For example, in some interviews, particular in-
dividuals attempted to dominate the group discussions.
3.4. Analysis The researchers compensated for this by encouraging
other practitioners to contribute.
As we have conducted group interviews with practi-
tioners, we are particularly interested in the analysis of 3.4.3. Factors and words
natural language, While there are many case studies of We assume that a practitioner is more likely to discuss
the software process, and SPI, few provide detailed in- those issues that they consider relevant, rather than those
formation on the techniques used to analyse the data issues that they consider not relevant. To simplify our
collected. analysis, we also assume that a set of words can be used
12 A. Rainer, T. Hall / The Journal of Systems and Software 66 (2003) 7–21

to represent a factor, so that if a word relating to a factor 3.4.5. Threats to the study
was used by at least one member of a group then the The main threat to this study is the problem of using
group was considered to have discussed that factor. a set of words to represent a factor. This problem occurs
(Word frequency then provides an indication of how because, for example, the same word can ÔhaveÕ different
important a factor is to a group of practitioners.) meanings in different contexts, and different words can
Clearly, this second assumption is disputable because, ÔhaveÕ the same meaning in the same or different con-
for example, a word can mean different things in different texts. The lack of context when quantitatively analysing
contexts. Given that one of our units of analyses is the words introduces some further problems:
group interview, provided at least one occurrence of a
word in a transcript is within a suitable context, then the • There is little indication of the direction of impact of
practitioners in that group interview have discussed the a factor e.g. whether a factor has a major positive im-
respective factor that the word represents. Furthermore, pact or a minor negative impact.
the questions asked by the research should help to nar- • A word may be used in a context that is irrelevant to
row the possible contexts, increasing the likelihood that this study. This can bias the results.
the words occur in an appropriate context and thus in- • The occurrence of a word may be conditional on the
creasing the likelihood that the words represent the fac- previous occurrence of a word. For example, one re-
tors. We discuss this issue in more depth in Section 3.4.5. spondent may use a word to refer to one of our fac-
We began with 161 words that were used to identify tors and this may ÔtriggerÕ other people to discuss
the 26 factors in our study (see Appendix A for more that factor and to use that word. We alluded to this
information). In assessing the preliminary word fre- earlier, where a dominant individual may discourage
quencies we noted that many words were either not used the discussion of some factors, or encourage the dis-
or very rarely used. These words did not really con- cussion of other factors.
tribute to the analysis. In response to this, and to • The frequency of occurrence of a word may actually
sharpen our analysis, we chose to focus on those words be an indication of the complexity of the factor,
that were used in at least 50% of the transcripts. We rather than an indication of the relevance of that fac-
found as a result of this refinement that: tor to SPI.

• 32 of the original 161 words accounted for almost We have taken the following steps to compensate for
80% of the citations. these problems.
• Only eight of the 18 core factors were cited in 50% or
more of the transcriptions. • We used sets of words to represent a factor. The
• Seven of the eight additional factors were cited in problem still remains, however, that the sets of words
50% or more of the transcriptions. do not completely and comprehensively ÔcaptureÕ the
factor. Consequently, we conducted some prelimi-
The choice of a 50% ÔthresholdÕ is consistent with our nary and informal analysis to check whether the sets
previous study, in which factors were only identified as of words do represent the factors. As noted earlier (in
having a major impact or no impact on SPI if 50% or Section 3.4.3) we also refined the words that we used
more of the survey respondents considered that the in the analysis.
factor had either a major impact or no impact. • We chose to focus on the frequency of occurrence of
a factor across transcripts, rather than the frequency
3.4.4. Qualitative analysis of the transcripts of occurrence of a factor within each transcript. Pro-
In order to gain a more thorough understanding of vided that one or more words relating to a factor oc-
the factors identified by the quantitative analysis, we curred in a transcript, we considered that the
used NVivo (QSR International Pty Ltd., 1999) to ex- practitioners in that group interview had discussed
tract each and every occurrences of a word, together the factor. With this criterion, we only need one word
with the surrounding paragraph, from every transcript. to genuinely refer to the factor.
For pragmatic reasons we could not investigate all fac- Word frequencies within a transcript may provide
tors. Therefore, we focused on the factor people, and our some indication of relevance, but this is together with
investigations of this factor revealed connections be- some indication of complexity. We cautiously present
tween this factor and a number of the other 25 factors. word frequencies, and cautiously interpret them in
We identified these connections by noting that words our discussion of the additional factors, in Section 8.
relating to different factors were occurring within the • We compared the results of the analysis of words
same paragraphs. The qualitative analysis allowed us to with results from the survey study.
take account of context and allowed us to ensure that
co-occurrence of words did suggest a meaningful con- Another threat to this study may be the lack of op-
nection between factors. erational definitions of the factors. This paper is at-
A. Rainer, T. Hall / The Journal of Systems and Software 66 (2003) 7–21 13

tempting to identify, in a general way, issues that prac- any of the practitioner groups. See the Appendix C for
titioners consider relevant to SPI. Having identified information on word frequencies.) The differences be-
general issues, the next step might be to Ôsharpen upÕ our tween the case studies and the survey study suggest that
understanding of these issues. This is where operational different strategies reveal different factors and that a
definitions would be developed. In the qualitative anal- multi-strategy approach is necessary for a more reliable
ysis for this paper (Section 7) we have started to clarify understanding of factors affecting SPI.
our understanding of the people factor. Further work on A more detailed examination of the citations by
this factor and other factors like change and problems practitioner groups reveals that the factor metrics is not
are clearly required. Therefore, we see operational defi- discussed by any of the practitioner groups in company
nitions as being a future outcome of this research rather 2. Yet metrics is discussed by at least one instance of
than a present outcome. each type of practitioner group in companies 1 and 3,
and by project managers in company 4. Of the eight
factors, metrics is the only factor that might distinguish
4. Quantitative analysis of the core factors the lower maturity companies (companies 1 and 2) from
the higher maturity companies (companies 3 and 4). It is
This section reports our investigations of the first and also the only factor that might distinguish those com-
second research questions: panies that can make a judgement on the success of their
SPI programmes from those that cannot make a
RQ1 What factors, as identified by our case studies, af- judgement. (Company 3 reports a successful SPI pro-
fect software processes? gramme, whilst company 4 reports a marginally suc-
RQ2 What factors, as identified by our case studies, do cessful SPI programme. Both companies 1 and 2 report
not affect software processes? that they did not know about the success of the pro-
gramme.) It may be, however, that companies 1 and 2
4.1. Factors identified by the four companies cannot judge the success of their SPI programmes be-
cause they are at a low level of maturity. This is a point
Table 5 summarises the eight factors that were cited recognised by, for example, Catteneo et al.:
by the majority of types of practitioner groups, in each
company. Of the eight factors, seven are cited by prac- In immature companies, it is difficult and some-
titioners in all four companies, and two (executive sup- times impossible to base an improvement action
port and experienced staff) are cited by all 15 on quantitative evaluation. (Cattaneo et al., 2001,
practitioner groups. p. 17)
The fact that only eight of the 18 core factors are cited
by the majority of practitioners is surprising, given some We have further explored practitionersÕ opinions of
of the findings of our previous study. For example, in our metrics in (Hall et al., 2000a, 2001).
previous study we identified training and mentoring (a
combined factor in the previous study) as the most
commonly cited factor having an impact on SPI. Table 5 5. Comparison with the survey study
indicates, however, that mentoring is not discussed by
the majority of practitioner groups. (Indeed, the word This section reports our investigations of the third
frequencies indicate that this factor was not discussed by research question:

Table 5
Factors identified by each of the four companies
Factor Company Total number Normalised Word freq.
1 2 3 4 of transcripts total

Executive support Yes Yes Yes Yes 15 11 176


Experienced staff Yes Yes Yes Yes 15 11 176
Reviews Yes Yes Yes Yes 14 10 100
Internal process ownership Yes Yes Yes Yes 13 10 39
Training Yes Yes Yes Yes 10 9 30
Procedures Yes Yes Yes Yes 10 8 64
Stringent control Yes Yes Yes Yes 9 8 12
Metrics Yes No Yes Yes 8 7 34
The normalised total takes account of differences in the actual number of practitioner groups in each company. For example, there were actually three
developer groups in company 1, but only one developer group in company 4. If one or more of the developer groups in company 1 cited the factor
executive support, then these citations are counted just once in the normalised total. The normalised total is, therefore, a count of citations of each
factor per type of practitioner group, per company. The normalised total provides a more reliable indicator of practitionersÕ opinions on each factor.
14 A. Rainer, T. Hall / The Journal of Systems and Software 66 (2003) 7–21

RQ3 Are the factors identified from our survey study One factor (stringent control) was identified as rele-
consistent with the factors identified from our case vant in the case studies. There was no clear indication, in
studies? the survey study, as to whether stringent control had a
major impact or no impact on SPI programmes. It may
Table 6 compares the factors identified in the survey be that stringent control has a minor impact. The minor
with the factors identified in the case studies. As the impact of a factor was not, however, investigated in the
table indicates, there are a number of similarities be- survey study.
tween the two studies. Seven factors are identified by
both studies as having some relevance to SPI. Two
factors (reward schemes and estimating tools) are con- 6. Quantitative analysis of additional factors
sidered, in both studies, to have no relevance to SPI.
(For reward schemes, Goldenson and Herbsleb (1995) Because over 50% of our core factors were not cited
drew a similar conclusion from their study.) in at least 50% of the transcriptions, and in an effort to
Four factors (inspections, internal leadership, stan- more thoroughly understand the factors that impact SPI
dards and mentoring) were identified in the survey, but programmes, we also examined the presence of eight
were not identified in the case studies. The most sur- additional factors. To be consistent with our analysis of
prising of these is mentoring, because this emerged (to- the core factors we also refined the word list for these
gether with training) as the most cited factor in the additional factors. Table 7 summarises the additional
survey study (cf. Table 1). factors that were cited by the majority of types of
practitioner groups, in each company.
While the factors quality and schedule are cited by all
companies, and a majority of texts and practitioner
Table 6 groups, the frequencies for these two factors suggest that
Comparison of factors from the survey study with factors from the they are not discussed in much detail. This may be be-
case study
cause the discussions focused on the implementation of
Factor Survey study Case studies SPI rather than the outcomes (see the interview ques-
Executive support Yes Yes tions in Appendix B).
Experienced staff Yes Yes
We recognised, in Section 3, that there are difficulties
Internal process ownership Yes Yes
Metrics Yes Yes interpreting the word frequencies from our analysis.
Procedures Yes Yes Accepting these difficulties, we are interested in the fre-
Reviews Yes Yes quency with which the additional factors are cited. The
Training Yes Yes factors change, people, problems and SPI are cited
Inspections Yes No much more frequently than the other additional fac-
Internal leadership Yes No tors. Indeed, these four factors are cited more frequently
Standards Yes No than any of the ÔcoreÕ factors, although the core fac-
Mentoring Yes No
tors executive support, experienced staff and reviews
Stringent controla Uncertain Yes have frequencies of the same order of magnitude (see
Reward schemes No No Table 5).
Estimating tools No No It is unsurprising that the factor SPI is mentioned so
a
For stringent control, it was not clear whether the factor had a frequently, given that the focus of the interview ques-
major impact or no impact. See Rainer and Hall (2002) for more in- tions was on SPI. For the other three factors, their high
formation. frequency of citation suggests that these factors may be

Table 7
Additional factors cited by the four companies
Factor Company Normalised Total number Word freq.
1 2 3 4 total of transcripts

Change Yes Yes Yes Yes 11 15 214


People Yes Yes Yes Yes 11 15 415
Problems Yes Yes Yes Yes 11 15 321
S. process improvement Yes Yes Yes Yes 11 15 256
Quality Yes Yes Yes Yes 10 13 98
Market Yes Yes Yes Yes 10 12 31
Schedule Yes Yes Yes Yes 9 11 30
Costa No No No No
a
The factor cost was not retained in the refined word list. It is shown here for completeness. See Table 5 for an explanation of normalised totals.
A. Rainer, T. Hall / The Journal of Systems and Software 66 (2003) 7–21 15

particularly interesting and relevant to SPI. It may be tionships between factors, and use these ÔsketchesÕ to
that these factors are much more important to success- indicate the kinds of issues that need to be addressed in
ful SPI than the ÔcoreÕ factors, or that these three factors further research.
are more abstract versions of some of the ÔcoreÕ factors. One implication of the close relationships between the
The importance of people has been repeatedly raised factors is for survey research. Many of the studies, in-
by researchers and practitioners within software engi- cluding our own, of factors affecting SPI present a list of
neering and SPI (e.g. Kellner et al., 1991; Cugola and factors, perhaps ranking the importance of these factors.
Ghezzi, 1998; Baddoo et al., 2000; van Solingen et al., These lists imply (although not necessarily intentionally)
2000). Less attention has been directed explicitly at that these factors can be treated independently of each
change, but such disciplines as SPI, management learn- other. Our qualitative analysis suggests that attention
ing and organisational development are fundamentally should be directed at how such factors relate to each
about change. Similarly, there appears to be little re- other.
search that takes a broad perspective on the range of
actual problems faced by practitioners. Most research 7.1. Experienced staff, people, training, reviews and
appears to begin with a particular problem, and then inspections
develops the investigation from that problem. An obvi-
ous implication of these discussions is that addressing There appears to be relationships between the factors
people, problems and change may lead to more effective experienced staff, people, training, reviews and inspections.
SPI. This implication requires further investigation. Clearly, one reason that people are so important is for
One implication of the frequency of the factor people their expertise and experience. This is recognised by the
is that although researchers and practitioners recognise need for training, as one method of increasing, or im-
the importance of people to software engineering, and to proving, expertise. One developer, in company 2, stated:
SPI, they may not recognise just how important people
are. As de Bono states: We can whinge now that we have lost all our exper-
tise and that people only want experienced staff but
Most of our thinking is not to do with puzzles or itÕs our fault that the less experienced staff did not
games. Most thinking is to do with other people. gain more experience when we had the opportunity
(De Bono, 1994, p. 77) to train them.

This suggests that people pervade many of the other This may also be an example of the need for men-
factors in SPI. This relates to our second interpretation: toring, a factor which was not explicitly mentioned in
that people, change and problems may be more abstract any group interview.
factors than the ÔcoreÕ factors. For example, executive The issue of experienced staff also relates to a po-
support and experienced staff clearly relate to people. tentially new, unexplored factor: recruitment policy. A
Training relates to change. Reviews and inspections are senior manager from company 1, stated:
intended to find certain types of problems. This inter-
pretation requires further investigation. Another problem is that historically we mainly re-
cruit people straight from university, people with
very little experience. We get very few people who
7. Qualitative analysis of the factors have experience from elsewhere.

In this section, we investigate the fourth research Reviews or inspections also provide an opportunity
question: for sharing expertise:

RQ4 Regardless of research strategy, how do the fac- I think [that] the important thing about a review
tors relate to each other? inspection process is that a senior, experienced
designer can make a contribution into many differ-
During the qualitative analysis of the transcriptions, ent pieces of development but in a managed and
two general insights about the factors emerged. First, controlled way. . . (Project manager)
some factors appear to be closely linked with each other.
Second, factors in addition to the 26 that we have in- This suggests an alternative perspective: one contri-
vestigated (but that relate to some of these 26 fac- bution of training, reviews, inspections, standards,
tors) emerge from the analysis. The nature of our procedures, and mentoring is that they all share exper-
analysis, and of the data, make specific or comprehen- tise. This also appears to be one of the functions of
sive descriptions of these links, and of the new factors, processes and standards i.e. to ÔcaptureÕ, store and dis-
difficult. We offer two ÔsketchesÕ of some of the rela- seminate knowledge.
16 A. Rainer, T. Hall / The Journal of Systems and Software 66 (2003) 7–21

7.2. People, procedures, change and ownership bining qualitative and quantitative analysis within our
case studies; second, by comparing our results with our
Comments in the transcriptions suggest that under- previous survey study results. We found that:
standing the importance of the factor people is, at least
partly, about understanding: • Many of the factors identified in the survey study were
also identified in the case studies. Also, both types of
• A personÕs ability to do work. This relates to experi- studies found these factors to have similar kinds of rel-
ence and expertise, and to shared expertise. evance. The similarity of results is encouraging be-
• A personÕs ability to change, which involves a recog- cause they come from two different types of study.
nition and willingness to accept that their knowledge • There were a small number of factors where the two
and work habits are no longer effective, but also a types of study produced opposing results. In one
recognition and willingness to accept that the pro- instance, mentoring, the differences were dramatic.
posed changes will be effective. The survey study found training and mentoring, as a
combined factor, to be the factor most obviously rel-
For example, a developer from company 1 stated: evant to SPI. (This does not mean that training and
mentoring was the factor with the greatest impact.)
In our local process we have our own review proce- Within the case studies, however, training was rec-
dure, which we developed about 4 to 5 years ago ognised by the majority of practitioner groups,
because we weren’t happy with what was there at whereas mentoring was not recognised by any of
a corporate level. (emphases added) the practitioner groups. This absence of mentoring
contradicts other research (Billings et al., 1994; Kras-
In this example, the developer indicates that they ner et al., 1994; Paulk et al., 1994). It may be that, due
were prepared to change (to develop a review procedure) to the problems of using individual words to repre-
but that they were not prepared to accept the corporate sent a factor, certain factors were not properly repre-
process. This example further suggests that part of the sented by a set of individual words.
problem of corporate standards is that they do not A related issue is that an alternative case study design
(maybe they cannot) take account of all local situations. (e.g. one that used a different set of words to repre-
This affects a personÕs willingness to change when they sent a factor) may also ÔdetectÕ different factors. The
do not believe that corporate standards will be effective different results of the two studies suggests potential
in their own situations. Ownership of procedures may design problems for both the survey study and the
occur when people recognise that the procedure is useful case studies. For the survey study, there was a design
for them in their local situation. The provision of a problem with investigating a combined factor. For
corporate-wide procedure, without evidence of how this the case studies there is the problem of using words to
procedure will help in the local situation, may be one represent factors. (This was discussed in Section
reason people are reluctant to change. 3.4.5.)
• Although the two types of study are identifying sim-
ilar factors, the case studies may have identified more
8. Evaluation of our multi-strategy approach fundamental factors i.e. people, change and prob-
lems. While a survey study could be designed to in-
8.1. An evaluation of our study design vestigate these factors, the nature of survey studies
limits the ways in which these factors could be stud-
In this section we investigate the final research ied. The very high frequencies with which people,
question: change and problems are cited suggests that these fac-
tors are complex and that they relate to many issues
RQ5 Is our approach a potentially useful approach to within software development. The high frequencies
investigating software processes? further suggest that practitioners do not realise how
often they refer to these issues.
There is a great deal of variability in the success of • The case studies reveal potentially relevant factors
SPI programmes, and there are a number of explana- and relationships between factors in ways that the sur-
tions for this variability. We examined the factors that vey studies could not.
affect software processes, and examined the research
strategies used to investigate these factors. We examined
the research strategies because the use of any particular 9. Conclusions
type of research strategy will influence the kinds of data
that can be collected from practitioners. We have used a Despite the growing body of research on SPI, there is
multi-strategy approach for this study: first, by com- still a great deal of variability in the success of SPI
A. Rainer, T. Hall / The Journal of Systems and Software 66 (2003) 7–21 17

Table 8
Factors and words used in the analysis
Factor Original words Revised words Factor Original words Revised words
Automated tools Automation Problems Difficult Difficult
Automate Difficulty
Automated Issue Issue
Automatic Issues Issues
Cost Cost Obstacle
Costs Obstacles
Change Change Change Problem Problem
Changed Changed Problems Problems
Changes Changes Procedures Framework
Changing Changing Guidance
Reorganisation Guide
Estimating tools Estimate Guideline
Estimates Guidelines
Estimating Procedure Procedure
Overestimate Procedures Procedures
Overestimated Project post-mortem Mortem
Overestimates Mortems
Underestimate Quality Defect
Underestimated Defects
Underestimates Quality Quality
Executive support Executive Reviews Review Review
Executives Reviewed
ExecutiveÕs Reviewing
Management Management Reviews Reviews
ManagementÕs Reward schemes Incentive
Manager Manager Incentives
Managers Managers Reward
ManagerÕs Rewarded
Experienced staff Experience Experience Rewarding
Experienced Rewards
Experiences Risk assessment Risk
Expert Risked
Expertise Risks
Experts Schedule Date
Knew Deadline
Know Know Deadlines
Knowing Deliver Deliver
Knowledge Delivered
Skill Delivering
Skilled Delivers
Skills Delivery Delivery
External consultants Consultant Finish
Consultancies Finished
Consultancy Finishes
Consultants Milestone
Inspections Inspections Milestones
Fagan Schedule
Fagen SPI Improve Improve
Inspect Improved Improved
Inspected Improvement Improvement
Inspecting Improves
Inspection Improving
Inspects PI PI
Internal leadership Leader SPI
Leaders Standards Standards
Leadership Standard
Internal process ownership Ownership Standardise
Own Own Standardised
Owns Standardised
Market Industries Standarize
Industry Training Educate
IndustryÕs Educated
Market Market Educates
(continued on next page)
18 A. Rainer, T. Hall / The Journal of Systems and Software 66 (2003) 7–21

Table 8 (continued)
Factor Original words Revised words Factor Original words Revised words
Markets Educational
Market’s Taught
Mentoring Mentor Teach
Mentored Train
Mentoring Training Training
Mentors Tutorial
Metrics Measure Tutorials
Measured Stringent control Control Control
Measurement Controlled
Measures Controlling
Measuring Controls
Metric Stringent
Metrics Metrics
People Culture
Cultures
Human
Humans
People People
Peoples
People’s
Person Person

programmes. We have considered two issues that relate this paper, and to Nathan and Prof. David Wilson for
to this variability of success: the factors that affect their help in collecting the data. We also thank the re-
software processes, and the research strategies that are viewers for their helpful comments.
used to investigate these factors.
In terms of the factors, we found that the relevance of
Appendix A. Factors and words
several factors to SPI was repeated across a survey study
and a multiple case study. These factors are: executive
Table 8 presents the original list of 161 words and the
support, experienced staff, internal process ownership,
revised list of 32 words, as used in the word frequency
metrics, procedures, reviews and training. We also
analysis of the group interview transcripts. No distinc-
found two factors, reward schemes and estimating tools,
tions were made between upper-case and lower-case
which the survey study and multiple case study found
spellings.
was not relevant to SPI.
The case studies also investigated additional factors
not investigated by the survey study. Three of these Appendix B. Interview questions
additional factors appear to be pervasive in SPI. They
are: people, change and problems. The possible rela- We used the following questions to guide the group
tionships between these factors can be expressed with interviews.
two conjectures: people change in response to problems;
and, problems occur because people do not change. What do you think SPI is?
In terms of the research strategies, the comparison of What do you think the problems and issues in soft-
results from the survey study and case study suggest that ware development at company X are?
a multi-strategy approach is necessary for a more com- How will SPI address these issues?
prehensive and reliable understanding of SPI. What are the strengths of your current software de-
velopment process?
What are the weaknesses of your software develop-
Acknowledgements ment process?
What do you think the aims and objectives of SPI at
We are sincerely grateful to all the companies and company X should be?
practitioners for their participation in our Managing What do you think managementÕs aims and objec-
Practitioner Impact on Process and Product (PPP) pro- tives of SPI at company X are?
ject. The PPP project is funded by the UKÕs Engineering What are the potential obstacles to SPI at company
and Physical Science Research Council, under grant X?
number EPSRC GR/L91962. We are grateful to Nathan What are the potential motivators for SPI at com-
Baddoo for his helpful comments on an earlier draft of pany X?
A. Rainer, T. Hall / The Journal of Systems and Software 66 (2003) 7–21 19

Table 9
Factors identified in the original and revised word lists
Factor Original word list Revised word list
Number of transcripts Freq. of citation Number of transcripts Freq. of citation
Executive support 15 179 15 176
Experienced staff 15 235 15 176
Reviews 14 113 14 100
Internal process ownership 13 46 13 39
Metrics 13 60 8 34
Training 12 40 10 30
Procedures 11 75 10 64
Stringent control 11 17 9 12
Inspections 10 67 0 0
Estimating tools 7 24 0 0
Internal leadership 7 17 0 0
Standards 7 27 0 0
Risk assessment 4 10 0 0
Reward schemes 3 5 0 0
Project post-mortem 2 3 0 0
Automated tools 1 1 0 0
External consultants 1 1 0 0
Mentoring 0 0 0 0
Total 920 631
% citations retained N/A 69
People 15 471 15 415
Problems 15 336 15 321
S/w process improvement 15 276 15 256
Change 15 218 15 214
Schedule 12 108 11 30
Quality 13 106 13 98
Market 12 48 12 31
Cost 9 25 0 0
Total 1588 1365
% citations retained N/A 86

Appendix C. Factors identified by the majority of prac- 2. How successful has your improvement programme
titioners been so far?

Table 9 summarises two sets of word frequencies for


Response Company
the 26 ÔcoreÕ and additional factors. The first set is based
on the original 161 words. The second set is based on the Not successful
refined 32 words. Marginally successful 3
Successful 4
Very successful
Appendix D. Survey questions and responses on SPI Not able to assess 1, 2
success Do not know

Some of the data in Table 4 was taken from responses


to the following two survey questions: References
1. In relation to your companyÕs attempts to improve
its approach to software development, how formal is Abbott, R.J., 1983. Program design by informal English descriptions.
your companyÕs current approach Communications of the ACM 26 (11), 882–894.
Baddoo, N., Hall, T., Wilson, D.N., 2000. Implementing a people
focused SPI programme. In: 11th European Software Control and
Metrics Conference and The Third SCOPE Conference on Soft-
Response Company ware Product Quality, Munich, 18–20 April.
Formal (documented process) Billings, C., Clifton, J., Kolkhorst, B., Lee, E., Wingert, W.B., 1994.
Journey to a mature software process. IBM Systems Journal 33 (1),
Informal (undocumented process) 1, 2, 3, 4
46–61.
20 A. Rainer, T. Hall / The Journal of Systems and Software 66 (2003) 7–21

Brodman, J.G., Johnson, D.L., 1994. What small businesses and small Haley, T.J., Ireland, B., Wojtaszek, E., Nash, D., Dion, R., 1995.
organizations say about the CMM. In: 16th International Confer- Raytheon electronic systems experience in software process im-
ence on Software Engineering (ICSE16), Sorrento, Italy, 16–21 provement. Software Engineering Institute, Carnegie Mellon Uni-
May, pp. 331–340. versity, Pittsburgh, PA, CMU/SEI-95-TR-017.
Bromley, D.B., 1977. Personality description in ordinary language. Hall, T., Baddoo, N., Wilson, D.N., 2001. Measurement in software
John Wiley & Sons, London. process improvement programmes: An empirical study. In: Dumke,
Butler, K.L., 1995. The economic benefits of software process R., Abran, A. (Eds.), New approaches to software measurement
improvement. CrossTalk 8 (7), 1–5. Proceedings of IWSM 2000. Springer-Verlag, pp. 73–83.
Butler, K.L., 1997. Process lessons learned while reaching level 4. Hall, T., Baddoo, N., Wilson, D.N., Rainer, A.W., 2000a. Optimising
CrossTalk 10 (5), 1–6. software measurement programmes using practitioner input. In:
Butler, K.L., Lipke, W., 2000. Software process achievement at Tinker Australian Conference on Software Measurement, Sydney, 1st–3rd
Air Force Base. Software Engineering Institute, Carnegie Mellon November.
University, Oklahoma, Pittsburgh, CMU/SEI-2000-TR-014. Hall, T., Wilson, D.N., Baddoo, N., 2000b. Towards implementing
Carley, K.M., 1997. Extracting team mental models through tex- successful software inspections. In: International Conference on
tual analysis. Journal of Organizational Behavior 18 (S1), 533–558. Software Methods and Tools (IEEE Computer Society), Wollon-
Cattaneo, F., Fuggetta, A., Sciuto, D., 2001. Pursuing coherence in going, Australia, 6–10 November.
software process assessment and improvement. Software Process–– Hayes, W., Zubrow, D., 1995. Moving on up: Data and experience
Improvement and Practice 6 (1), 3–22. doing CMM-based process improvement. Software Engineering
Chen, P.P.-S., 1983. English sentence structure and entity-relationship Institute, CMU/SEI-95-TR-008.
diagrams. Information Sciences, 29. Herbsleb, J., Goldenson, D., 1996. A systematic survey of CMM
Cugola, G., Ghezzi, C., 1998. Software processes: A retrospective and experience and results. In: 18th International Conference on
a path to the future. Software Process––Improvement and Practice Software Engineering (ICSE-18), Berlin, Germany, 25–26
4 (3), 101–123. March.
Daskalantonakis, M.K., 1992. A practical view of software measure- Herbsleb, J., Carleton, A.D., Rozum, J., Siegel, J., Zubrow, D., 1994.
ment and implementation experiences within Motorola. IEEE Benefits of CMM-based software process improvement: Initial
Transactions on Software Engineering 18 (11), 998–1010. results. Software Engineering Institute, Carnegie Mellon Univer-
Daskalantonakis, M.K., 1994. Achieving higher SEI levels. IEEE sity, Pittsburgh, CMS/SEI-94-TR-013.
Software 11 (4), 17–24. Holsti, O.R., 1969. Content analysis for the social sciences and
De Bono, E., 1994. De BonoÕs thinking course. BBC Worldwide Ltd., humanities. Addison-Wesley, London.
London. Humphrey, W.S., Snyder, T.R., Willis, R.R., 1991. Software pro-
Diaz, M., Sligo, J., 1997. How software process improvement helped cess improvement at Hughes Aircraft. IEEE Software 8 (4), 11–
motorola. IEEE Software 14 (5), 75–81. 23.
Dion, R., 1992. Elements of a process-improvement program. IEEE Kellner, M.I., Curtis, B., DeMarco, T., Kishida, K., Schlumberger,
Software 9 (4), 83–85. M., Tully, C., 1991. Panel: Non-technological issues in software
Dion, R., 1993. Process improvement and the corporate balance sheet. engineering. In: 13th International Conference on Software Engi-
IEEE Software 10 (4), 28–35. neering (ICSE-13), Austin, Texas, 13–17 May, pp. 144–
Dyb a, T., 2000. An instrument for measuring the key factors of success 155.
in software process improvement. Empirical Software Engineering Kitson, D.H., Masters, S.M., 1993. An analysis of SEI software
5 (4), 357–390. process assessment results: 1987–1991. In: 15th International
El Emam, K., Fusaro, P., Smith, B., 1999. Success factors and barriers Conference on Software Engineering, Baltimore, Maryland.
for software process improvement. In: Tully, C., Messnarz, R. Kness, S.P., Satake, M.S., 1997. A level 5 organization looks at the
(Eds.), Better Software Practice for Business Benefit: Principles and personal software process. CrossTalk 10 (10).
Experience. IEEE Computer Society, pp. 355–371. Krasner, H., 1999. The payoff for software process improvement:
Ferguson, P., Leman, G., Perini, P., Renner, S., Seshagiri, G., 1999. What it is and how to get it. In: El Emam, K., Madhavji, N.H.
Software process improvement works! Software Engineering Insti- (Eds.), Elements of software process assessment & improvement.
tute, Carnegie Mellon University, Pittsburgh, PA, CMU/SEI-99- IEEE Computer Society, Washington.
TR-027. Krasner, H., Pyles, J., Wohlwend, H., 1994. A case history of the space
Ferrera, L., Timko, C., 2000. The Telcordia Technologies road to shuttle onboard systems project. SEMATECH Inc., Austin, TX,
quality. Cutter IT Journal 13 (2), 28–34. 94092551A-TR.
Fitzgerald, B., OÕKane, T., 1999. A longitudinal study of software McGarry, F., Pajerski, R.S., Page, G., Waligora, S., Basili, V.R.,
process improvement. IEEE Software 16 (3), 37–45. Zelkowitz, M., 1994. Software process improvement in the NASA
Florac, W.A., Carleton, A.D., Barnard, J.R., 2000. Statistical process software engineering laboratory. Software Engineering Insti-
control: Analyzing a space shuttle onboard software process. IEEE tute, Carnegie Mellon University, Pittsburgh, PA, CMU/SEI-94-
Software 17 (4), 97–106. TR-22.
Fowler, K.M., 1997. SEI CMM level 5: A practitionerÕs perpsective. Mehner, T., Messer, T., Paul, P., Paulisch, F., Schless, P., V€ olker,
CrossTalk 10 (9). A., 1998. Siemens process assessment and improvement ap-
Goldenson, D.R., Herbsleb, J.D., 1995. After the appraisal: A proaches: Experiences and benefits. In: The Twenty-Second Annual
systematic survey of process improvement, its benefits, and factors International Computer Software and Applications Conference
that influence success. Software Engineering Institute, Carnegie (Compsac Õ98). IEEE Comput. Soc, Los Alamitos, CA, USA.
Mellon University, CMU/SEI-95-TR-009 (alternatively ESC-TR- Miles, M.B., Huberman, A.M., 1994. Qualitative data analysis. SAGE
95-009). Publications, London.
Gonauser, M., Paulisch, F., V€ olker, A., 1998. SiemensÕ experience with Misic, M.M., Russo, N.L., 2000. Reading between the lines: An
people/process/technology: Lesson learned. 3rd Annual European examination of systems analysis and design texts. Journal of
Software Engineering Process Group (ESEPGÕ98) Conference, Systems and Software 50 (1), 65–73.
London, June 1998, C307a 1-36. NASA, 1994. An overview of the software engineering laboratory.
Haley, T.J., 1996. Software process improvement at raytheon. IEEE NASA Goddard Space Flight Center, Greenbelt, Maryland, SEL-
Software 13 (6), 33–41. 94-005.
A. Rainer, T. Hall / The Journal of Systems and Software 66 (2003) 7–21 21

NASA, 1996. Software process improvement handbook (revision 1), SEMATECH, 1995. A case history of process improvements at the
NASA Goddard Space Flight Center, Greenbelt, Maryland, NASA software engineering laboratory, Sematech, Austin, TX,
NASA-GB-001-95. 94122662A-TR.
Oshana, R.S., 1998. Tailoring Cleanroom for industrial use. IEEE SPICE, 1998. SPICE Phase 2 trials interim report version 1. 17th June
Software 15 (6), 46–55. 1998. http://www.iese.fhg.de/SPICE.
Pajerski, R.S., Basili, V.R., 1997. The SEL adapts to meet changing Stelzer, D., Mellis, W., 1998. Success factors of organizational change
times. In: 22nd Software Engineering Workshop, NASA/Goddard in software process improvement. Software Process––Improvement
SEL, Greenbelt, Maryland, December 1997. and Practice 4 (4), 227–250.
Paulish, D.J., Carleton, A.D., 1994. Case studies of software-process- Strauss, A.L., 1987. Qualitative analysis for social scientists. Cam-
improvement measurement. Computer 27 (9), 50–57. bridge University Press, Cambridge.
Paulk, M.C., 2000. Software CMM case study bibliography. Software van Solingen, R., Berghout, E., Kusters, R., Trienekens, J., 2000.
Engineering Institute. Available from <http://www.sei.cmu.edu/ From process improvement to people improvement: Enabling
cmm/docs/roi.html>. learning in software development. Information and Software
Paulk, M.C., Weber, C.V., Curtis, B., Chrissis, M.B., 1994. A high- Technology 42 (14), 965–971.
maturity example: Space shuttle onboard software. In: Paulk, Wigle, G.B., Yamamura, G., 1997. Practice of an SEI CMM level 5
M.C., Weber, C.V., Curtis, B., Chrissis, M.B. (Eds.), The capability SEPG. CrossTalk 10 (11). Available from <http://www.stsc.hill.
maturity model: Guidelines for improving the software process. af.mil/CrossTalk/1997/nov/cmm5sepg.asp>.
Addison-Wesley, Harlow, England. Willis, R.R., Rova, R.M., Scott, M.D., Johnson, M.I., Ryskowski,
Paulk, M.C., Goldenson, D., White, D.M., 2000. The 1999 survey of J.F., Moon, J.A., Shumate, K.C., Winfield, T.O., 1998. Hughes
high maturity organizations. Software Engineering Institute, Car- AircraftÕs widespread deployment of a continuously improving
negie Mellon University, CMU/SEI-2000-SR-002. software process. Software Engineering Institute, Carnegie Mellon
Pitterman, B., 2000. Telcordia technologies: The journey to high University, Pittsburgh, PA, CMU/SEI-98-TR-006.
maturity. IEEE Software 17 (4), 89–96. Wohlwend, H., Rosenbaum, S., 1993. Software improvements in an
QSR International Pty Ltd., 1999. QSR NVivo. http://www.qsr.co- international company. In: 15th International Conference on
m.au/. Software Engineering (ICSE), Baltimore, Maryland, pp. 212–220.
Rainer, A.W., Hall, T., 2001. An analysis of some Ôcore studiesÕ of Wohlwend, H., Rosenbaum, S., 1994. SchlumbergerÕs software
software process improvement. Software Process––Improvement improvement program. IEEE Transactions on Software Engineer-
and Practice 6 (4), 169–187. ing 20 (11), 833–839.
Rainer, A.W., Hall, T., 2002. Key success factors for implementing Yamamura, G., 1999. Head to head: Process improvement satisfies
software process improvement: A maturity-based analysis. Journal employees. IEEE Software 16 (5), 83–85.
of Systems and Software. Yamamura, G., Wigle, G.B., 1997. SEI CMM level 5: For the right
Rainer, A.W., Hall, T., Baddoo, N., Wilson, D.N., 2001. An overview reasons. CrossTalk 10 (8). Available from <http://www.stsc.hil-
of the practitioners, processes and products project. In: 6th Annual l.af.mil/crosstalk/1997/aug/seicmm5.asp>.
Conference of the UK Academy of Information Systems (UK- Yin, R.K., 1994. Case Study Research: Design and Methods. SAGE
AIS2001), 18–20 April, University of Portsmouth, UK. Publications, London.
Raman, S., 2000. It is software process, stupid. IEEE AES Systems
Magazine (June).
Raytheon Corporate Communications, 1998. RaytheonÕs command Austen Rainer is currently Research Fellow in the Department of
Computer Science at the University of Hertfordshire. He received a
and control division awarded software engineering instituteÕs level B.Sc. (Hon) in Applied Psychology and Computing from Bourne-
five rating. Press release, 7th January, Raytheon, Arlington, mouth University and, subsequently, a Ph.D. in Software Engineering
Virginia. Management in collaboration with IBM Hursley Park. He has previ-
Saeki, M., Horai, H., Enomoto, H., 1989. Software development ously worked in the software industry.
process from natural language specification. In: 11th International
Conference on Software Engineering, Pittsburgh, PA, 15–18 May. Tracy Hall is currently Principal Lecturer in the Department of
IEEE Computer Society, pp. 64–73. Computer Science at the University of Hertfordshire. Previously she
Schlumberger, 2000. Schlumberger joins the worldÕs software elite–– was Head of the Centre for Systems and Software Engineering at
South Bank University. She has published widely in the area of soft-
award confirms quality of processes and people. Press release, 27th ware quality and measurement. Following B.A. and M.Sc. degrees she
November, Schlumberger. was awarded her Ph.D. from City University.

You might also like