When it comes to best practice in IT projects, knowledge gain is taking a back seat to knowledge loss.

The "expensive bungles" that still occur with computer systems are an embarrassment to the computing profession and managers of IT installations. There is no simple answer to explain why these disasters still happen so frequently, but the way in which knowledge of good practice for IT projects is retained in the face of constant change appears one significant problem.

Recent expensive bungles include the recent high-profile failures at Royal Melbourne Institute of Technology (RMIT) and Sydney Water. The NSW Auditor-General found that $61 million of taxpayers' funds were likely to be written off after a Customer Billing and Information Service (CBIS) project at Sydney Water was abandoned.

Why are these disasters still occurring? Looking at these systems we see examples of knowledge loss in the information technology area, rather than knowledge gain. This is a serious and collective problem for industry, management, academia and research funding agencies. The rewards for using IT successfully are large. A 2001 report from the Australian Productivity Commission showed that Australia's productivity surge in the 1990s was in part due to the successful use of IT. Fewer bungles would mean even greater productivity gains.

An examination of one failed system is instructive. Problems with the Academic Management System at RMIT had serious financial implications for the university as a whole. The estimated cost for the system was $47 million, more than twice the original budget. The project involved packaged enterprise application software that required extensive customisation.

The Victorian Auditor-General detailed numerous problems with the project, including ineffective project management, poor definition of the project's scope and time frame, and a failure to check whether the system was working before it went "live".

Many of these problems showed a flagrant disregard for all knowledge of "good practice" of information systems development and project management as recognised in industry and universities. Why did this happen? Was there collective forgetting of all they had ever learned by those involved?

The bungles at Sydney Water and RMIT are not isolated incidents. A study by the Standish Group of 23,000 IT projects between 1994 and 1998 showed 26 per cent of the projects were successful, 46 per cent were challenged and 28 per cent failed. These statistics are disturbing, especially as in this study the RMIT case would have been classified as "challenged".

The prevalence of failures with IT systems can be seen as evidence of deeper underlying problems to do with the way knowledge about these systems is built up in industry and in research in universities.

Many new ideas for IT come from industry rather than research in universities - a case of "state of the practice" leading the "state of the art". Yet knowledge in industry tends not to be systemised or easily shared. Organisational knowledge is built up over time and often relies on the intellectual capital of long-serving employees.

Knowledge in industry can also be of the "fads and fashions" type, where new developments are associated with hype and unrealistic claims. Many of these fads contain valuable new ideas, but the over-statement of claims obscures the need to keep much of the older knowledge that we had before - such as principles of good project management - which do not change and are still needed no matter what new technology we have.

Universities also have problems with systemising knowledge of IT. The study of computing is very new, compared with other disciplines, and the knowledge areas have evolved very rapidly. The first commercial computer, the Univac, was produced in 1951. The first computer science departments appeared in the 1960s. This is a very recent history compared with other disciplines such as medicine that have been studied for thousands of years.

Studies associated with IT are carried on in many different departments under many different names. Researchers have vastly different backgrounds, including mathematics, science, engineering, philosophy, psychology, business, management, and sociology - meaning a lack of agreement on how research should be carried out.

An additional problem in universities is a type of intellectual snobbery. The study of the design of IT systems can be seen as lacking academic respectability, being too soft and "cookbooky", especially if it involves knowledge of the human beings using the systems. The tendency is to revert to research that looks more like traditional natural science or mathematics and is seen as "harder".

It is clear that much greater attention should be paid to a "design science" for IT systems, based on underlying theories of technology and human behaviour individually, socially and politically.

Managers, practitioners and university researchers need a more historical and holistic view of the complexity of IT systems that involve people and technology - meaning real, useful knowledge is accumulated and old knowledge is not lost in waves of fads and fancies.

The study of design science for IT is a continuing project of the e-Commerce group at the Australian National University. http://ecocomm.anu.edu.au/research/ecom.asp F

Shirley Gregor is professor of information systems at the Australian National University's Faculty of Economics and Commerce. This column is abridged from a seminar given at the National Institute of Engineering and Information Sciences on July 16, 2003