Newer
Older
Discussion_Papers / Papers / 1993 / 93-04 / DP9304SM-original.TEX
\documentstyle[12pt,doublesp,a4_local]{article}

\setstretch{1.2}

\begin{document}

\pagestyle{plain}

\begin{center}

\vspace*{2.5cm}

{\Large Software Development, CASE Tools and 4GLs --}

\vspace{0.1cm}

{\Large A Survey of New Zealand Usage.}

\vspace{0.5cm}

{\large Part 1: 750 New Zealand Organisations}

\vspace{1.2cm}

Dr Stephen G. MacDonell\footnote{{\scriptsize Address correspondence 
to: Dr S.G. MacDonell, Research Fellow, Department of Information 
Science, University of Otago, P.O. Box 56, Dunedin, New Zealand. 
Fax: +64 3 479 8311  Email: stevemac@commerce.otago.ac.nz}}

\vspace{0.05cm}

Department of Information Science

\vspace{0.05cm}

University of Otago

\vspace{0.6cm}

November 1993

\vspace{2cm}

Abstract

\vspace{0.3cm}

\end{center}

\begin{quote}
This paper reports the results of a recent national survey which 
considered the use of CASE tools and 4GLs in commercial software 
development. Responses from just over 750 organisations show a high 
degree of product penetration, along with extensive use of package 
solutions. Use of 3GLs in general, and of COBOL in particular, is 
still relatively widespread, however. In terms of systems analysis 
and design techniques under a CASE/4GL environment, screen and 
report definition is the most preferred technique, although both 
data flow analysis and data modelling also feature strongly.
\end{quote}

\vspace*{2cm}

\pagebreak

\subsubsection*{1 Introduction}

In general terms fourth generation languages (4GLs) and computer 
aided software engineering (CASE) tools have been in existence for a 
decade or more. During this period they have received extensive 
attention from academia and industry as well-publicised solutions to 
the `software crisis', that is, the increasing backlog of new system 
requirements coupled with rising demands for maintenance. Early 
local industry perceptions of the benefits of 4GLs were typical of 
those reported worldwide---``Application development with fourth 
generation languages (4GL) has been shown to be at least 10 times 
faster than BASIC or COBOL.'' (as reported in {\em NZ Computer 
Interface\/}, April 1984). 

The extent to which these products have since been adopted for use 
in development, however, is unclear. As recently as 1987 McClure 
suggested that just 2\% of potential users were employing CASE 
tools. Furthermore, Pressman (1992) has remarked that even though 
CASE tools are more important than hardware for attaining greater 
levels of quality and productivity, they are still not widely used 
within software development. Locally, a 1987 survey found that less 
than 15\% of total users were employing a 4GL (Botica, 1987).

A number of reasons may have contributed to this apparently slow 
acceptance: one possibility would appear to have been the absence of 
clear definitions for the terms `4GL' and `CASE'. These and other 
related terms have been used very widely as marketing tools while 
the direct meaning of the words, in terms of product functionality, 
has remained vague (CIS, 1989). This is clearly reflected in the 
findings of a 1987 Australian survey: of 507 respondents, 19.6\% 
{\em weren't sure} whether they were using a 4GL or not, due to 
problems of definition (Pacific Reporter, 1987).

Another possible cause may have been the expense associated with the 
adoption of CASE/4GL tools---Huff (1992) has suggested that many 
organisations simply found the move to CASE too costly, resulting in 
the cancellation of many CASE-based projects. Maria (1991) has 
stated that CASE tools have been slow to gain acceptance because, in 
general, the tools have failed to address the requirements for 
maintenance of existing systems, in spite of the fact that 
maintenance has continually consumed significant amounts of 
developer effort, and consequently, development budgets. Conversely, 
other reports have suggested that the acceptance of development 
tools has been steadily increasing (e.g. see Chen and Norman (1992) 
or CIS (1989)). In general, however, no empirical data has been 
provided to confirm or refute these suggestions.

The main objective of the survey described here, then, was to 
ascertain whether this situation of low product adoption has changed 
over the past five years. One of its principle aims was therefore to 
determine the current levels of 4GL and CASE tool usage in this 
country. In a similar vein, the survey was also concerned with 
related changes in software development practices. A 1987 survey of 
280 organisations (resulting in 120 replies) has provided a baseline 
for comparison (McAulay, 1987).

\subsubsection*{2 The Survey}

Development and distribution of the survey was carried out in April 
and May of this year. A further four months were then allocated to 
the data collection phase of the project. The procedures used during 
this period are described in the next subsection, followed by the 
presentation and discussion of the survey results. \\

\noindent 2.1 Methodology \\

\noindent The survey instrument was developed with two conflicting 
objectives in mind. First, it needed to be comprehensive enough to 
ensure that the required data would be provided. Second, it needed 
to be relatively short so that potential respondents would not be 
discouraged from completing it. A set of (at most) eight sections 
were therefore included, considering issues relating to product use, 
development approaches, analysis and design methods and system types.

The survey was distributed to two different samples. The main sample 
of more than 1000 sites was made up of a cross-section of New 
Zealand companies and organisations, details of which were obtained 
from a number of sources, including the New Zealand Business Who's 
Who, the NZ Post Business Directory and conference attendance lists. 
Site selection was performed in an attempt to ensure that 
organisations of varying size and function were surveyed. (The 
second sample was a much smaller set of software houses. The 
responses obtained from this sample will be reported at a later 
date.) The distribution of sites in the main sample included 
organisations from nineteen different functional classes. For 
comparative purposes, Table~\ref{propn} shows the percentage values 
for the number of sites in each class against the number of 
employees in the same class, as taken from the 1993 New Zealand 
Yearbook. Clearly this is only one classification method, and 
others, such as those based on contribution to GDP or on annual 
turnover, may in fact be more relevant. Accurately obtaining 
financial data of this kind over nineteen aggregated classes, 
however, is extremely difficult. Therefore the employee-based 
approach was adopted as a relatively crude but basically informative 
classification method, at least for the purposes of the survey.

\begin{table}[htb]
\begin{center}                                       
\begin{tabular}{||l|r|r|r||} \hline                      
\multicolumn{4}{||c||}{Site Type Classification (Percentages)} \\ \hline 
Site type  & Employees  & Survey sites  & Difference \\ \hline \hline
01 Primary Production                   &  2.5 &  2.3 & -- 0.2 \\ \hline
02 Oil, Gas, Minerals, Electricity      &  2.4 &  9.1 &  + 6.7 \\ \hline
03 Chemicals, Pharmaceuticals           &  1.0 &  2.6 &  + 1.6 \\ \hline
04 Manufacturing (Non-food)             & 12.0 & 18.9 &  + 6.9 \\ \hline
05 Agricultural/Horticultural Products  &  3.4 &  3.7 &  + 0.3 \\ \hline
06 Retail, Wholesale, Distribution      & 21.5 & 12.2 & -- 9.3 \\ \hline
07 Food and Beverage Manufacturing      &  2.6 &  4.6 &  + 2.0 \\ \hline
08 Hospitality, Tourism                 &  5.7 &  2.6 & -- 3.1 \\ \hline
09 Transportation, Storage              &  5.6 &  6.1 &  + 0.5 \\ \hline
10/11 Central and Local Government      &  7.8 &  8.2 &  + 0.4 \\ \hline
12 Health Care                          &  8.5 &  1.3 & -- 7.2 \\ \hline
13 Construction, Engineering            &  7.6 &  3.5 & -- 4.1 \\ \hline
14/20 Legal and Business Services       &  9.2 &  5.1 & -- 4.1 \\ \hline
16 Banking, Finance                     &  3.7 &  4.3 &  + 0.6 \\ \hline
17 Insurance                            &  1.3 &  5.1 &  + 3.8 \\ \hline
18 Communications, Media, Publishing    &  4.2 &  4.8 &  + 0.6 \\ \hline
19 Automotive Assembly and Sales        &  1.3 &  1.4 &  + 0.1 \\ \hline
\end{tabular}
\end{center}
\caption{Distribution of survey sites across industry classes}
\label{propn}
\end{table}

Some marked differences appear in Table~\ref{propn}, but in general 
these are not overly unusual. Classes 06, 12, 13 and 14/20, for 
example, show significantly lower percentages of sites when compared 
to the percentages of category employees. These classes, however, 
represent labour-intensive industries (Retail, Wholesale, 
Distribution, Health Care, Construction, Engineering, and Legal and 
Business Services), in which the proportion of software development 
units to total personnel could be expected to be small. In contrast, 
the proportions of sites for classes 02 and 04 are much greater than 
the corresponding percentages of employees. Given that this cannot 
be attributed to low survey response, this might be taken to suggest 
that the number of employees per functional unit (and consequently 
per software development unit) is relatively low in these industry 
types (Oil, Gas, Minerals, Electricity, and Manufacturing 
(Non-food)). Support for the inclusion of a large number of sites 
from the Manufacturing class is provided by a recent survey of the 
information technology departments of 2000 Australasian 
organisations (Philipson, 1993), in which sites from the 
manufacturing industry also made up the largest percentage of the 
sample (at 23.7\%). Thus the only potentially questionable 
difference in proportions occurs for class 02, which included all of 
the recently established electricity and gas distribution/retail 
authorities, hence the high percentage of sites in comparison to the 
number of employees. Over the entire sample, however, the 
distribution of site types is approximately proportional to the 
employee distribution. This would suggest that the responses of the 
selected sample should provide an appropriate pointer to those we 
could expect from the entire development industry in this country. 
(Results broken down according to industry type will be published in 
a subsequent paper.)

The survey was distributed by mail to Information Systems Managers 
over a period of two months. Each approach included a cover letter, 
explaining the motivation for the study, a copy of the survey and a 
pre-addressed reply-paid envelope. Every site that had not replied 
within six weeks of posting was contacted again in the same manner, 
but with a slightly different cover letter.

The response to the survey was extremely satisfying and quite 
unexpected. Of the original sample of 1068 sites 95 were found to be 
repeats---that is, more than one letter was sent to the same 
organisation, usually under a different name. A further 38 sites 
were found to be invalid---they were no longer trading, or declined 
to complete the survey. This left an actual sample of 935 sites. Of 
these, 753 usable replies were received. This represents a response 
rate of just over 80\%. It is to be hoped that such a large number 
of responses will help to ensure that the responses are 
representative of the development population. \\

\noindent 2.2 Results \\

\noindent The first response section of the survey included spaces 
for the respondent's name, title, address and so on. Of some 
interest here were the job titles of the respondents. The twelve 
most frequently used titles are shown in Table~\ref{jobs}.

\begin{table}[htb]
\begin{center}                                       
\begin{tabular}{||l|r||} \hline                      
\multicolumn{2}{||c||}{Respondent's Job Title} \\ \hline 
Title                              & Frequency  \\ \hline \hline
IS Manager                         & 109  \\ \hline
MIS Manager                        &  51  \\ \hline
(Unspecified)                      &  49  \\ \hline
Systems Manager                    &  32  \\ \hline
DP Manager                         &  30  \\ \hline
Company Secretary                  &  25  \\ \hline
Financial Controller               &  22  \\ \hline
General Manager                    &  21  \\ \hline
IT Manager                         &  18  \\ \hline
Consultant                         &  15  \\ \hline
Systems Development Manager        &  14  \\ \hline
Computer Manager                   &  13  \\ \hline
Computer Services Manager          &  12  \\ \hline
\end{tabular}
\end{center}
\caption{Frequency of respondent job titles}
\label{jobs}
\end{table}

The frequencies shown in Table~\ref{jobs} account for just over 400 
of the 753 replies. Responsibility for software development and use 
would appear to be well-defined, with jobs specific to computers or 
information systems making up nine of the top twelve job classes. 
Moreover, 449 of the responses were from managers, illustrating the 
relative importance that is now afforded to positions concerned with 
the control of software systems.

The first two software-related questions in the survey were 
concerned with system usage and overall system implementation 
strategies. The results of these questions are shown in 
Tables~\ref{useis},~\ref{strat}~and~\ref{stratpn}.
\begin{table}[htb]
\begin{center}                                       
\begin{tabular}{||l|r|r||} \hline                      
\multicolumn{3}{||c||}{Use of Information Systems} \\ \hline 
                          & Yes & No  \\ \hline \hline
Use information systems?  & 736 & 17  \\ \hline
\end{tabular}
\end{center}
\caption{Use of computer-based information systems}
\label{useis}
\end{table}

\begin{table}[htb]
\begin{center}                                       
\begin{tabular}{||l|r|r||} \hline                      
\multicolumn{3}{||c||}{System Implementation Strategies} \\ \hline 
                                      & Yes & No  \\ \hline \hline
Perform in-house development?         & 441 & 280 \\ \hline
Purchase packages off-the-shelf?      & 626 &  88 \\ \hline
Systems supplied by software house?   & 441 & 273 \\ \hline
Use other implementation approaches?  &  59 & 655 \\ \hline
\end{tabular}
\end{center}
\caption{Methods of system implementation}
\label{strat}
\end{table}

\begin{table}[htb]
\begin{center}                                       
\begin{tabular}{||l|r|r|r|r|r|r||} \hline                      
\multicolumn{7}{||c||}{System Implementation Strategy Proportions} \\ \hline 
                           & 1-20\% & 21-40\% & 41-60\% & 
61-80\% & 81-100\% & Average \\ \hline
                           & Cases: & Cases:  & Cases:  & 
Cases:  & Cases:   &          \\ \hline \hline
In-house development       & 176      & 60        & 50        & 
57        & 98         & 28\% \\ \hline
Packages off-the-shelf     & 222      & 99        & 81        & 
95        & 129        & 42\% \\ \hline
Software house             & 197      & 64        & 45        & 
61        & 74         & 26\% \\ \hline
Other                      & 14       & 8         & 6         & 
12        & 19         & 5\% \\ \hline
\end{tabular}
\end{center}
\caption{Proportions of systems implemented under various strategies}
\label{stratpn}
\end{table}

When compared to past reports, certain trends can be identified from 
the results presented in Tables~\ref{strat}~and~\ref{stratpn}. 
McAulay (1987) found that of 120 New Zealand respondents, the 
average proportions of systems implemented under the approaches 
considered in Tables~\ref{strat}~and~\ref{stratpn} were: 56\% of 
systems were developed in-house, 27\% were implemented from 
pre-packaged solutions, and 8\% were developed by software houses. 

Based on further questions, McAulay stated that the dominance of 
in-house development was not surprising, but that changes were 
anticipated. More than 50\% of her respondents expected package use 
to increase relative to other methods, whereas only 36\% expected 
in-house development to increase. Costs of in-house development and 
a lack of time were cited as the main reasons for moving away from 
in-house development to packaged solutions. Of the 120 respondents, 
13\% felt that the proportion of package use would decrease, due 
mainly to the need for integrated systems and for systems that 
satisfied unusual user requirements. McAulay (1987) remarked, 
however, that the first of these two reasons was likely to become 
less prominent given the trend towards integration-oriented packages.

By 1989 it was reported that most New Zealand companies would look 
for a package solution to fulfil application requirements before 
they would consider in-house development (CIS, 1989). This trend 
appears to have continued, given the results of the current survey. 
When compared with McAulay's results (1987), it can be seen that 
in-house development has fallen from 56\% to 28\% on average, and 
that both package use and software house development have increased 
markedly, from 27\% and 8\% to 42\% and 26\% respectively. These
results confirm the expectations of McAulay's respondents with 
respect to changes in development strategies.

The remainder of the survey was concerned with in-house development 
so only those respondents that undertook their own development (a 
total of 456 sites) continued answering. The next question in the 
survey asked respondents who did not use CASE tools or 4GLs to 
specify the development languages that they did use. This generated 
valid responses from 187 sites. Just under half of this set of 
respondents (84) said they used one language, 56 respondents said 
that two languages were employed, 22 sites said they utilised three 
development languages, 13 cited use of four languages and 12 said 
that five or more languages were used for software development. The 
actual languages used are shown in Table~\ref{olang}.

\begin{table}[htb]
\begin{center}                                       
\begin{tabular}{||l|r||l|r||} \hline                      
\multicolumn{4}{||c||}{Development Languages Used} \\ \hline 
Language Used:   & Cases: & Language Used:   & Cases: \\ \hline \hline
COBOL            & 62     & SQL              & 7     \\ \hline
RPG              & 48     & FORTRAN          & 6     \\ \hline
DBASE            & 32     & LOTUS 123        & 6     \\ \hline
BASIC            & 21     & UNIX SCRIPTS     & 5     \\ \hline
C++              & 18     & PL/1             & 4     \\ \hline
C                & 16     & ASSEMBLER        & 3     \\ \hline
PICK TOOLS       & 14     & CL               & 3     \\ \hline
PARADOX          & 11     & DATAFLEX         & 3     \\ \hline
CLIPPER          & 10     & QUEO             & 3     \\ \hline
FOXPRO           & 10     & Twelve different languages      &  2 \\ \hline
PASCAL           &  9     & Forty-three different languages &  1 \\ \hline
MS ACCESS        &  8     &                  &       \\ \hline
(Unspecified)    &  8     & Distinct languages specified    & 76 \\ \hline
\end{tabular}
\end{center}
\caption{Development languages used by non-CASE/4GL users}
\label{olang}
\end{table}

Table~\ref{olang} illustrates that, at those sites not using CASE 
tools or 4GLs, COBOL continues to be the most widely used 
development language. Even considering that CASE tool/4GL users were 
excluded from this question, the results provide some support for 
Pressman's assertion (1992) that languages like COBOL and, to a 
lesser extent, FORTRAN, now nearly thirty years old, are still 
widely used. The results also illustrate a continuation of the trend 
reported by Botica (1987) for language use in this country. He cited 
a 1984 study in which 26\% of respondents said that COBOL was used 
and 23\% said that they used BASIC. This pattern was echoed in a 
1987 survey of New Zealand developers (unfortunately, no sample size 
was given): COBOL 26.6\%, BASIC 18.9\%, Others 13.8\%, RPG 13.1\%, 
Fortran 6. 4\%, Assembler 4. 8\%, PL/1 2.6\%. All of these languages 
feature strongly, in almost the same order of usage, in the 
responses to the current survey.

The final two sections of the survey were addressed only to those 
sites that employed CASE tools and/or 4GLs. It was originally 
intended that respondents would firstly specify the CASE tools 
they were using and then the 4GLs they were using. After an 
examination of both trade and academic literature, however, it was 
decided that the two product classes should in fact be left in one 
category, for the following reasons:
\begin{itemize}
\item As suggested in the introduction to this paper, the terms 
`CASE tool' and `4GL' have been defined in very vague terms. It was 
considered that this would make any objective and consistent 
categorisation of products unlikely.
\item When first introduced, many tools were oriented towards 
assisting specific development tasks---some were related only to 
front-end activities, that is, analysis and design tasks (Burkhard 
and Jenster, 1989; CIS, 1989). In contrast, the focus of other tools 
was on the back-end of development, to assist personnel in the 
coding and testing phases (Norman and Chen, 1992). In recent years, 
however, increased product integration has meant that, in many 
cases, diverse products have been pooled to create a single 
development environment. Moreover, several products now address the 
entire development life cycle. This could be taken as confirmation 
of the assertion that 4GLs would be subsumed by CASE (Kolodziej, 
1988; Stamps, 1989). It could equally suggest, however, that 4GL-type 
products have simply been provided with more effective front-end 
interfaces. Irrespective of these arguments, consistent 
classification of products given these circumstances was again 
thought to be unlikely.
\end{itemize}

Thus the remainder of the survey was completed by respondents whose 
organisation used a CASE tool and/or a 4GL---a set of 265 sites. 
This represents approximately 35\% of the original respondent total 
(753 replies) and 60\% of those who said they performed in-house 
development. It is difficult to compare these figures against those 
of previous surveys and reports as, in general, it is not clear 
whether the proportions cited in these reports referred to general 
organisations or only to those sites that performed in-house 
development. For illustrative purposes, however, some of the results 
are included here. King (1992) reported the results of two UK 
industry surveys, with both illustrating the failure of CASE to gain 
widespread industry acceptance. One survey of 250 UK companies found 
that 41\% had no plan to use CASE, 27\% were planning to use it, 
25\% were using it and 7\% had used and rejected it. Similarly, a 
1987 report from Australia revealed that, based on the responses of 
more than 500 sites, fewer than one in five Australian installations 
were using 4GLs, and only 4.6\% intended to install a 4GL in the 
following year (Pacific Reporter, 1987). In terms of usage trends in 
this country, McAulay (1987) found that 38\% of 120 New Zealand 
companies were using a 4GL of some sort. Botica (1987) reported a 
similar result---he found that fourth generation tools were being 
used by 31.3\% of sites, based on an independent survey of an 
unspecified number of installations. Another 5\% were considering 
using them, and 63.7\% had no intention of using them.

Returning to the results of the current survey, more than 150 of the 
265 CASE tool/4GL users said that they employed one CASE/4GL 
product, 57 sites cited use of two products, 32 respondents said 
that three products were used, seventeen organisations stated that 
they used four tools and five sites said that five or more CASE/4GL 
products were being utilised. The breakdown of usage over the 
various products is shown in Table~\ref{case4gl}.

\begin{table}[htb]
\begin{center}                                       
\begin{tabular}{||l|r||l|r||} \hline                      
\multicolumn{4}{||c||}{CASE Tools/4GLs Used} \\ \hline 
Product Used:     & Cases: & Product Used:     & Cases: \\ \hline 
\hline
POWERHOUSE        & 33     & SYSTEM ARCHITECT  & 5     \\ \hline
ORACLE            & 28     & AREV              & 4     \\ \hline
LINC              & 23     & ASSET             & 4     \\ \hline
ADW               & 21     & CUE-BIC           & 4     \\ \hline
IEW               & 17     & DEFT              & 4     \\ \hline
ORACLE CASE       & 17     & EXCELERATOR       & 4     \\ \hline
PARADOX           & 17     & IE:ADVANTAGE      & 4     \\ \hline
(Unspecified)     & 17     & SAS               & 4     \\ \hline
FOXPRO            & 16     & SUPERBASE         & 4     \\ \hline
INFORMIX          & 15     & ASAP              & 3     \\ \hline
INGRES            & 12     & CLARION           & 3     \\ \hline
MS ACCESS         & 10     & CQCS              & 3     \\ \hline
LDA               &  8     & EASYCASE          & 3     \\ \hline
PROGRESS          &  8     & GUPTA SQL         & 3     \\ \hline
SYNON             &  8     & IEF               & 3     \\ \hline
DATAFLEX          &  7     & NATURAL CONSTRUCT & 3     \\ \hline
NATURAL           &  7     & PICK TOOLS        & 3     \\ \hline
PACE              &  7     & RTMAS             & 3     \\ \hline
SPEED             &  7     & Twenty-two different CASE tools/4GLs & 
2 \\ \hline
PRIME INFORMATION &  6     & Forty-nine different CASE tools/4GLs & 
1 \\ \hline
ALL               &  5     &                   &       \\ \hline
LBMS              &  5     &                   &       \\ \hline
RALLY             &  5     & Distinct CASE tools/4GLs specified   & 
111 \\ \hline
\end{tabular}
\end{center}
\caption{CASE tools and 4GLs used for in-house software development}
\label{case4gl}
\end{table}

The results presented in Table~\ref{case4gl} illustrate the 
diversity of the CASE tool/4GL market, with more than 100 different 
development products cited as currently in use. The responses are 
dominated, however, by Oracle product users (45 sites), 
KnowledgeWare product users (ADW and IEW -- 38 sites) and PowerHouse 
users (33 sites).

The next question in the survey asked respondents to indicate 
whether they used their CASE tool and/or 4GL products during more 
than one phase of development. The responses to this question are 
shown in Table~\ref{phase}.

\begin{table}[htb]
\begin{center}                                       
\begin{tabular}{||l|r|r|r||} \hline                      
\multicolumn{4}{||c||}{CASE Tool/4GL Use During Phases} \\ \hline 
                    & Yes & No  & Blank \\ \hline \hline
Feasibility study?  &  63 & 180 & 19    \\ \hline
Analysis?           & 134 & 115 & 13    \\ \hline
Design?             & 192 &  63 &  7    \\ \hline
Coding?             & 220 &  33 &  9    \\ \hline
Unit testing?       & 175 &  77 & 10    \\ \hline
Implementation?     & 176 &  80 &  6    \\ \hline
System testing?     & 165 &  82 & 15    \\ \hline
\end{tabular}
\end{center}
\caption{Development phases during which CASE tools/4GLs are used}
\label{phase}
\end{table}

It can be seen from Table~\ref{phase} that the tools are used most 
frequently for assistance with coding---more than 80\% of the CASE 
tool/4GL users stated that one or more products were utilised in 
this phase of development. Five of the remaining six phases also 
experienced strong CASE tool/4GL use, with between 50\% and 72\% of 
sites using the tools at those times. Only the feasibility study 
showed an overall absence of tool use, with just 23\% of product 
users stating that CASE tool/4GL support was employed for the 
relevant tasks. Although this could be interpreted as a reflection 
of a lack of product support for the feasibility study, it may 
simply be evidence of the assertion that support is generally 
unnecessary at this time, as the importance of this stage is 
relatively low (CIS, 1989). 

The final product-related question in the survey was concerned with 
development organisations' analysis and design methods. The 
popularity of structured analysis and design techniques is widely 
acknowledged, as is the general expectation that CASE support for 
these techniques leads to increased product quality (CIS, 1989; Tse 
and Pong, 1989; Vessey {\em et al\/}., 1992). Moreover, 4GLs are 
notably strong in their support for screen and report definition, 
often as part of a prototyping methodology (Keuffel, 1991; Jones, 
1988). Given this type of support, widespread use of these methods 
was expected. The first part of the question therefore addressed the 
methods that were supported by the tools, and the second part 
considered actual use of the methods (irrespective of tool support). 
The associated responses are shown in 
Tables~\ref{andes1}~and~\ref{andes2}.

\begin{table}[htb]
\begin{center}                                       
\begin{tabular}{||l|r|r|r||} \hline                      
\multicolumn{4}{||c||}{Analysis and Design Methods With Tool Support} \\ \hline 
                                & Yes & No  & Blank \\ \hline \hline
Data models?                    & 127 & 103 & 26    \\ \hline
Data flow diagrams?             & 103 & 124 & 29    \\ \hline
Database operation definition?  & 115 &  99 & 42    \\ \hline
Functional decomposition?       & 114 & 102 & 40    \\ \hline
Screen/Report definition?       & 226 &  20 & 10    \\ \hline
Other?                          &  26 & 230 &  0    \\ \hline
\end{tabular}
\end{center}
\caption{Analysis and design methods supported by the CASE 
tools/4GLs used}
\label{andes1}
\end{table}

\begin{table}[htb]
\begin{center}                                       
\begin{tabular}{||l|r|r|r||} \hline                      
\multicolumn{4}{||c||}{Analysis and Design Methods Used} \\ \hline 
                                & Yes & No  & Blank \\ \hline \hline
Data models?                    & 132 &  76 & 48    \\ \hline
Data flow diagrams?             & 106 & 102 & 48    \\ \hline
Database operation definition?  & 100 &  93 & 63    \\ \hline
Functional decomposition?       & 105 &  97 & 54    \\ \hline
Screen/Report definition?       & 207 &  33 & 16    \\ \hline
Other?                          &  26 & 230 &  0    \\ \hline
\end{tabular}
\end{center}
\caption{Analysis and design methods used in development}
\label{andes2}
\end{table}

McAulay (1987) found that 47\% of the 120 respondents to her survey 
did not use any structured analysis technique---data flow methods 
were employed by 25\% of the sites, while data analysis techniques 
were in use in just 21\% of the responding companies. The results 
presented in Tables~\ref{andes1}~and~\ref{andes2} show that the use 
of data analysis procedures has increased markedly in the 
intervening six years. Even if it is assumed that none of the 
non-CASE/4GL sites use data modelling, the ratio of developers using 
data analysis has still increased by 9\% to 30\%. In contrast, the 
use of data flow diagrams (DFDs) appears to have remained stable at 
around 25\% (under the worst case scenario). These results would 
appear to provide industry support for the assertions that both data 
models and data flow diagrams are used extensively in development 
(Tate {\em et al\/}., 1992; Gray {\em et al\/}., 1991). Clearly the 
most widely employed method, however, is screen and report 
definition. Nearly 80\% of the CASE tool/4GL users said that this 
technique was used as part of their development process. 

The last pair of questions in the survey were rather more 
exploratory in nature, in that they addressed the number and type of 
systems produced using the organisations' CASE tools and 4GLs over 
the last three years. The results of these questions are shown in 
Tables~\ref{num}~and~\ref{type}.

\begin{table}[htb]
\begin{center}                                       
\begin{tabular}{||l|r||l|r||} \hline                      
\multicolumn{4}{||c||}{Number of New Systems Completed} \\ \hline 
Number of systems:   & Cases: & Number of systems:   & Cases: \\ \hline \hline
None                 & 15     & Eight                &  4     \\ \hline
One                  & 33     & Nine                 &  2     \\ \hline
Two                  & 37     & Ten                  & 15     \\ \hline
Three                & 41     & 11-15                & 12     \\ \hline
Four                 & 32     & 16-20                &  2     \\ \hline
Five                 & 23     & 21-40                &  0     \\ \hline
Six                  & 20     & 41-60                &  3     \\ \hline
Seven                &  6     &                      &        \\ \hline
\end{tabular}
\end{center}
\caption{Number of new systems completed with CASE tools/4GLs in 
last 3 years}
\label{num}
\end{table}

\begin{table}[htb]
\begin{center}                                       
\begin{tabular}{||l|r||} \hline                      
\multicolumn{2}{||c||}{Types of Systems Developed} \\ \hline 
Type:                              & Cases:  \\ \hline \hline
Interactive transaction processing & 112  \\ \hline
Interactive reporting              &  88  \\ \hline
Real time                          &  71  \\ \hline
Batch reporting                    &  33  \\ \hline
Interactive                        &  23  \\ \hline
All types                          &  19  \\ \hline
Batch                              &  10  \\ \hline
Batch transaction processing       &   9  \\ \hline
\end{tabular}
\end{center}
\caption{Types of systems developed using CASE tools and 4GLs}
\label{type}
\end{table}

\subsubsection*{3 Conclusions}

If it is reasonable to conclude that 750 responses constitute an 
adequate sample then the results just presented provide a relevant 
and useful insight into software development in this country. These 
findings have implications not only for those directly involved in 
software development but also for those who teach it.

In terms of system implementation methods, the last five years have 
seen a significant increase in the amount of package use when 
compared to other methods. Not only is the number of sites using 
packages for system implementation very high, at 85\% of the 
respondents, but packages are now also used to implement the 
greatest proportion of systems. There may be several reasons for 
this situation---probably the greatest contributors are the 
increasing availability of high-quality packages and the high costs 
of in-house development. These findings also suggest that, along 
with traditional training in software development, a reasonably 
large component of package-based education might be equally 
appropriate in preparing students for the information systems 
workplace.

A large number of sites undertaking in-house development continue to 
use tools and languages developed prior to the CASE tool/4GL `era', 
with 76 such languages being specified by 187 respondents. The 
strong use of COBOL and RPG may indicate that organisations are 
still making extensive use of some older applications and that the 
long-term investment in these systems continues to create a 
cost-based reluctance to migrate crucial and/or highly complicated 
systems to a new environment. Moreover, given that junior 
development staff are often assigned to maintenance tasks it would 
appear that teaching at least some COBOL remains a relatively 
sensible education strategy.

Just over 60\% of the survey respondents who said they performed 
in-house development also stated that they used at least one CASE 
tool or 4GL, a figure much higher than expected given the slow 
initial acceptance of these products. This result indicates a 
significant shift towards the use of these products when compared to 
the situation just six years ago. Given that the proportion of 
in-house development has also decreased in this time, it could be 
concluded that these advanced tools have been purchased for the 
development of strategic systems, while the concurrent development 
of less important operational systems has continued either under a 
3GL-type environment or through the use of package solutions.

The breakdown of CASE tool/4GL products in use highlights the 
diversity of the development tool market, with more than 100 
products mentioned. The listing also illustrates the problem of 
product classification (as discussed earlier in the paper)---of the 
41 CASE tool/4GL products in use at more than two sites, 17 of these 
also appeared in the non-CASE tool/4GL product list.

Of the development phases considered in the survey, tool support was 
used most frequently to assist coding tasks. This is almost 
certainly a result of the large number of programming aid tools 
available and in use relative to the number of front-end or full 
life cycle tools. The solid support for and use of various analysis 
and design methods provided no real surprises. A few respondents 
stated that they used a technique that was not supported, and {\em 
vice versa}, but in general those methods with tool support were 
widely utilised. These results also provide reasonably strong 
evidence of the continued use of structured analysis and design 
techniques, with data analysis and modelling being the most popular 
of these approaches. This represents something of a turnaround to 
the situation six years ago when data flow techniques were more 
widely used. In terms of development personnel education and 
training, these results suggest that equal attention should be given 
towards data and process analysis techniques, along with a 
consideration of screen and report prototyping methods.

In summary it can be concluded that the adoption of CASE tools and 
4GLs has increased significantly over the last five years, so much 
so that these products now appear to be an integral component of the 
suite of tools used by organisations involved in software 
development. Given the tremendously high degree of response to the 
survey we intend to make it a regular event, so that trends can be 
identified quickly and accurately. Such an approach should enable 
all those involved---vendors, developers, managers and 
educators---with effective and up-to-date indications of industry 
product requirements.

\subsubsection*{Acknowledgements}

\noindent The author is grateful for the support provided by the 
Information Science department at the University of Otago.

\subsubsection*{References}

\noindent Botica, A., ``Number Eight Style is Strength'', {\em 
Computerworld New Zealand\/}, 27 July 1987, pp. 21-31. \\

\noindent Burk\-hard, D.L. and Jen\-ster, P.V., ``Ap\-pli\-ca\-tions 
of Com\-pu\-ter-Ai\-ded Soft\-ware En\-gi\-nee\-ring Tools: Sur\-vey 
of Cur\-rent and Pro\-spec\-tive Us\-ers'', {\em ACM SIG\-BDP Da\-ta 
Base\/}, Fall 1989, pp. 28-37.\\

\noindent Chen, M. and Norman, R.J., ``A Framework for Integrated 
CASE'', {\em IEEE Software\/}, March 1992, pp. 18-22.\\

\noindent CIS, {\em CASE Project - Report on Computer Aided Software 
Engineering in New Zealand\/}. Case Study Report, Center for 
Information Science, University of Auckland, May 1989.\\

\noindent Gray, R.H.M., Carey, B.N., McGlynn, N.A. and Pengelly, A.
D., ``Design Metrics for Database Systems'', {\em BT Technology 
Journal\/} 9 (4), October 1991, pp. 69-79.\\

\noindent Huff, C.C., ``Elements of a Realistic CASE Tool Adoption 
Budget'', {\em Communications of the ACM\/} 35 (4), April 1992, pp. 
45-54. \\

\noindent Jones, R., ``A Quantum Leap in Languages: Major Gains in 
Using 4GLs'', {\em Computerworld New Zealand\/}, 31 October 1988, 
pp. 18-19.\\

\noindent Keuffel, W., ``Trans\-for\-ma\-tion Stra\-te\-gies'', {\em 
Com\-pu\-ter Lan\-guage (USA)\/}, Oc\-to\-ber 1991, pp. 35-42.\\

\noindent King, S.F., ``Making CASE Work''. Unpublished Research 
Proposal, Advanced Technology Centre, University of Warwick, 
Coventry, April 1992.\\

\noindent Kolodziej, S., ``Fighting the 4GL Identity Crisis'', {\em 
Computerworld New Zealand\/}, 14 November 1988, pp. 32-34. \\

\noindent Maria, A., ``CASE Technology: Today's Reality'', {\em 
Journal of Systems Management\/}, February 1991, pp. 18, 26.\\

\noindent McAulay, K., ``Information Systems Development and the 
Changing Role of MIS in the Organisation'', {\em Proceedings, 1st 
New Zealand MIS Management Conference\/}, Wellington/Auckland, March 
1987, pp. S1.1-S1.16. \\

\noindent Norman, R.J. and Chen, M., ``Working Together to Integrate 
CASE (Guest Editors' Introduction)'', {\em IEEE Software\/}, March 
1992, pp. 13-16.\\

\noindent Pacific Reporter, ``4GL Sales Hype Baf\-fles Aus\-sie Users'', 
{\em Computer\-world New Zealand\/}, 1 June 1987, p. 6. \\

\noindent Philipson, G. (ed.), {\em MIS 2000\/}. Strategic Publishing 
Group, Sydney, 1993. \\

\noindent Pressman, R.S., {\em Software Engineering: A 
Practitioner's Approach\/}. McGraw-Hill (3rd ed.), New York, 1992. \\

\noindent Stamps, D., ``CASE vs. 4GLs'', {\em Datamation\/}, 15 
August 1989, pp. 29-32.\\

\noindent Tate, G., Verner, J. and Jeffery, R., ``CASE: A Testbed 
for Modeling, Measurement and Management'', {\em Communications of 
the ACM\/} 35 (4), April 1992, pp. 65-72.\\

\noindent Tse, T.H. and Pong, L., ``Towards a Formal Foundation for 
DeMarco Data Flow Diagrams'', {\em The Computer Journal\/} 32 (1), 
1989, pp. 1-11.\\

\noindent Vessey, I., Jarvenpaa, S.L. and Tractinsky, N., 
``Evaluation of Vendor Products: CASE Tools as Methodology 
Companions'', {\em Communications of the ACM\/} 35 (4), April 1992, 
pp. 90-105.

\end{document}