Why big IT projects go wrong

by John Naughton

The Observer

In 1975, a computer scientist named Fred Brooks published one of the seminal texts in the literature of computing. It had the intriguing title of “The Mythical Man-Month” and it consisted simply of a set of essays on the art of managing large software projects. Between its covers is distilled more wisdom about computing than is contained in any other volume, which is why it has never been out of print. And every government minister, civil servant and chief executive thinking about embarking on a large IT project should be obliged to read it — and answer a multiple-choice quiz afterwards.

Why? Because Brooks was the guy who led the team that in the 1960s created the operating system for IBM’s 360 range of mainframe computers. This was probably the largest non-military software project ever mounted, and it was of vital strategic importance to IBM, which then completely dominated the computer business. It also turned out to be vastly more complex than anyone — including Brooks — anticipated, and it rapidly metamorphosed into a kind of death march.

The project fell further and further behind schedule. But because IBM was a rich company and OS/360 was so important, it was able to throw more and more resources (i.e., programmers) at the task. But as it did so, the problems got worse, not better. At which point Brooks had his epiphany: He realized that every time he added a programmer to the team the project fell further behind.

In the end, however, the job was done. The death march ended, OS/360 was delivered and IBM went on to make a lot of money from it. Brooks, for his part, resigned from the company, became professor of computer science at the University of North Carolina in Chapel Hill and then sat down to write the book that made him famous. His aim was to distil into a set of elegant essays everything he had learned from the OS/360 experience. The striking title came from his epiphany — the realization that man-months are a hopeless metric for assessing the size of a complex software project.

How come? Basically because a big software project involves two kinds of work: the actual writing of computer code and coordinating the work of maybe hundreds of programmers working on different parts of the overall system. Coordination represents an essential but unproductive overhead — and the more programmers you have, the bigger that overhead becomes. Hence Brooks’s law: adding manpower to a late software project makes it later.

Over the years, fragments of Brooks’s wisdom have percolated into the consciousness of ministers, civil servants and chief executives. But only fragments. In Britain we are wearily familiar with the long, dreary catalogue of botched or outlandishly expensive government IT projects. This is not just a public sector problem, however. Research conducted by two Oxford academics and published in the Harvard Business Review suggests that the private sector has almost as much difficulty managing big software projects, and that some such projects have even endangered the survival of the companies that embarked upon them.

A case in point was the jeans manufacturer Levi Strauss. In 2003 it was a global company, with operations in more than 110 countries but with an IT system that was an ancient, “Balkanized” mix of incompatible country-specific systems. So its bosses decided to migrate to a single SAP system and hired a team of fancy consultants (from Deloitte) to lead the effort. “The risks seemed small,” wrote the researchers. “The proposed budget was less than $5 million.” But very quickly things fell apart. One major customer, Walmart, required that the system interface with its supply chain management system, creating additional work. During the switchover to the new system, Levi Strauss was unable to fulfil orders and had to close its three U.S. distribution centres for a week. In 2008, the company took a $192.5 million charge against earnings to compensate for the botched project — and fired its chief information officer.

The Oxford researchers examined more than 1,400 big IT projects — comparing their budgets and estimated performance benefits with the actual costs and results. The average project cost $167 million and the largest a whopping $33 billion. The researchers’ sample drew heavily on United States-based projects but found little difference between them and European projects. Likewise, they found little difference between private companies and public agencies. One in six had a cost overrun of 200 percent.

The message is clear: If you run a big company or a government department and are contemplating a big IT product, ask yourself this question: Can your company or your ministerial career survive if the project goes over budget by 40 percent or more, or if only 25-50 percent of the projected benefits are realized? If the answer is “No” go back to square one. And read Fred Brooks’s lovely book.

  • Starviking

    From my time working in the IT Industry, a long time ago, there were two important facts that related to any IT Project:

    Project goals are: delivery on time, delivery to budget, and delivery to specifications. Choose any two.

    Any changes after the project has been defined and work started has a massive effect on delivery time, budget, and spec.

    Customers seem to think that just because the product is etherial, composed of lines of computer code and electrons, that changes can be made at any time. This is not the case: IT projects can be more complex than machinery, and no-one but a fool would think that re-casting the design of a piece of machinery during production would have little or no cost.