Issue: Vol. 13, No. 5, September 1996 | PDF

Classic Mistakes

In this column, instead of examining the software industry’s best practices I want to step through the looking glass. I want to examine its worst practices– specifically, those that have been used so often, by so many people, to produce such predictably bad results that they should be labeled “classic mistakes.”

Most of the classic mistakes have seductive appeal, which is part of the reason they’ve been made often enough to be considered classics. Do you need to rescue a project that’s behind schedule? Add more people. Do you want an earlier delivery date? Just set a more aggressive schedule. Should you keep a key contributor who’s aggravating the rest of the team? Yes, the project is too important to let him go.

Here are my nominations for software’s top 12 classic mistakes.

Undermined motivation. Study after study has found that motivation probably has a larger impact on productivity and quality than any other factor (Barry Boehm, Prentice-Hall, Software Engineering Economics, 1981). Considering that, you would expect a full-fledged motivation program to occupy a position of central importance on every software project. But that’s not the case. Motivation is a soft factor; it’s difficult to quantify, and it often takes a back seat to other factors that might be less important but are easier to measure. Every organization knows that motivation is important, but only a few organizations do anything about it. Many common management practices are penny-wise and pound-foolish, trading huge losses in motivation and morale for minor methodology improvements or dubious budget savings.

Uncontrolled problem employees. Failure to deal with rogue programmers has been a well-understood mistake at least since Gerald Weinberg published Psychology of Computer Programming in 1971 (Van Nostrand Reinhold). But a study by Carl E. Larson and Frank M. J. LaFasto found that failure to deal with a problem employee is still the most common complaint that team members have about their leaders (Teamwork: What Must Go Right; What Can Go Wrong, Sage, 1989). Larson and LaFasto’s study was not specifically about software, but I think software teams are just as susceptible to this problem as other kinds of teams. At best, failure to deal with problem employees undermines the morale and motivation of the rest of the team. At worst, it increases turnover among the good developers and damages product quality and productivity.

Noisy, crowded offices. Need to save money? A common economy is to cram developers into low-budget office space. Most developers rate their working conditions as unsatisfactory and report that they are neither sufficiently quiet nor sufficiently private. Workers who occupy noisy, crowded work bays or cubicles tend to perform significantly worse than workers who occupy quiet, private offices (Tom DeMarco and Timothy Lister, Peopleware, Dorset House, 1987).

Abandoning planning under pressure. Projects make plans and then routinely abandon them (without replanning) when they run into schedule trouble (Watts Humphrey, Managing the Software Process, Addison Wesley, 1989). Without a coherent plan, projects tend to fall into a chaotic code-and-fix mode, which is probably the least effective development approach for all but the smallest projects.

Shortchanging upstream activities. Projects that are in a hurry try to cut out nonessential activities, and since requirements analysis, architecture, and design don’t directly produce code, they are easy targets for the schedule ax. On one disaster project that I took over, I asked to see the design. The team lead told me, “We didn’t have time to do a design.” Also known as “jumping into coding,” the results of this classic mistake are all too predictable. Time is wasted implementing hacks, which later have to be thrown out and redeveloped with more care. Projects that skimp on upstream activities typically have to do the same work downstream at anywhere from 10 to 100 times the cost of doing it earlier (Barry W. Boehm and Philip N. Papaccio, “Understanding and Controlling Software Costs,” IEEE Transactions on Software Engineering, October 1988). “If you can’t find time to do the job right in the first place,” the old chestnut goes, “how will you find time to do it again later?”

Shortchanging quality assurance to improve development speed. Projects that are in a hurry often cut corners by eliminating reviews, test planning, and all but the most perfunctory testing. This is a particularly unfortunate decision. Short-cutting a day of QA activity early in the project is likely to add 3 to 10 days of unnecessary activity downstream (Capers Jones, Assessment and Control of Software Risks, Yourdon Press, 1994).

Lack of feature-creep control. The average project experiences about a 25-percent change in requirements from “requirements complete” to first release (Jones 1994). That 25-percent change in requirements produces at least a 25-percent addition to the software schedule–probably much more because of the multiplicatively higher costs associated with doing work downstream instead of upstream. Many projects lack formal change-control processes that could help them limit changes to those that are absolutely necessary

Silver-bullet syndrome. Silver-bullet syndrome occurs whenever managers or developers expect any single new tool or methodology to solve all its productivity problems. Silver-bullet tools and methodologies damage projects in two ways. First, the new tools or methodologies virtually never deliver improvements as dramatic as promised. Project-wide productivity improvements of more than 25 percent from first use of a new tool or methodology are virtually unheard of. Second, belief in silver bullets leads to serialization of improvements that could be made in parallel. Because managers or developers put all their faith into a single silver bullet, they try promising new tools and methods one at a time rather two or more at a time, which slows the adoption of potentially beneficial new tools and methods other than the silver bullet. The bottom line is that organizations that succumb to silver-bullet syndrome tend not to improve their productivity at all; indeed, they often go backwards (Jones 1994).

Wasting time in the “fuzzy front end.” The “fuzzy front end” is the time before the project starts, the time normally spent in the approval and budgeting process. It’s easier, cheaper, and less risky to shave a few weeks or months off the fuzzy front end than it is to compress a development schedule by the same amount. But it’s not uncommon for a project to spend months or years in the fuzzy front end and then to burst out of the starting gates with an aggressive, often unattainable schedule.

Insufficient user input. A Standish Group survey in 1994 (“Charting the Seas of Information Technology”) found that the number one reason that IT projects succeed is because of end-user involvement. Projects without early end-user involvement increase the risk of misunderstanding the project’s requirements and are especially vulnerable to time-consuming requirements creep.

Overly aggressive schedules. The 1994 Standish Group survey found that the average IT project took about 220 percent of its planned schedule. Scheduling errors of this magnitude set a project up for failure. Plans created for a project that turns out to be more than double its estimated size cannot be effective. The most serious consequence is probably that, if upstream activities are abbreviated proportionately to the abbreviated schedule (more than 50 percent), the average project might be doing as much as half of its upstream work downstream, at 10 to 100 times its nominal cost. Overly aggressive schedules also put excessive schedule pressure on developers, which ultimately hurts both morale and productivity.

Adding developers to a late project. Perhaps the most classic of the classic mistakes is adding developers to a project that’s behind schedule. There are notable exceptions to the rule, but, generally, when a project is behind schedule, new people subtract more productivity from existing staff than they add through their own work. Fred Brooks likened adding people to a late project to pouring gasoline on a fire (Mythical Man-Month, Addison Wesley, 1975).


This list of mistakes is hardly exhaustive. I have simply identified the mistakes I have seen most often. Your list might be different. A complete list would undoubtedly be pages longer.

Regardless of the exact contents, keep some list of classic mistakes in mind. Conduct project post-mortems to identify the classic mistakes particular to your organization. Exchange war stories with colleagues in other organizations to learn about the mistakes they’ve made. Create checklists of mistakes for use in your project planning. Post lists of classic mistakes on your group’s bulletin board for use in project monitoring. Appoint a “classic mistakes watchdog” to sound an alarm if your project begins to succumb to a classic mistake.

The classic mistakes’ seductive allure brings them into play again and again, but we as an industry have gained enough experience to recognize them for what they are. Now that we recognize them, we just need to be hard-headed enough to resist their appeal.

Editor: Steve McConnell, Construx Software  |  More articles