ONE of the universal complaints I hear seems to be engendered by a simple statistical misconception regarding averages. In the context of networking, it is computer departments declaring that network loading is only 20 per cent, and so there is obviously plenty of bandwidth. In the case of hospital beds and unemployment, it is get the patients to the right hospital, or get on your bike and find a job - there are plenty of opportunities out there. For road networks, it is more white and yellow paint, traffic-calming schemes and speed restrictions that will be sure to save lives.

Unfortunately, such assumptions are generally wrong because averages indicate very little and are insufficient to quantify any problem or solution. As evidence, we see recurring logjams on computer networks, a crisis due to a lack of hospital beds, people still unemployed while there are plenty of vacancies, and billions lost in traffic jams on the roads. On all counts, the bottom line is increasing human inconvenience, injury and death.

If only the people making decisions could grasp the concept of one more statistical parameter - standard deviation or variance - we might see more considered judgments, and a far more satisfactory outcome. As a general rule, it is almost irrelevant where the Gaussian Bell Curve sits - the average value - but the width or spread (standard deviation or variance) is vital.

In the 1960s and 70s, when most motorway systems were being planned, the people responsible for the traffic models got their statistics right and designed the new roads to be three or four lanes wide. But such over-provision at the time was considered politically unacceptable and so two and three lanes became the standard for much of Europe. Just 30 years on, we are reaping the reward of those decisions - massive fuel, time and loss of life costs.

We also face an infrastructure upgrade bill that may be beyond the economic limitations of individual countries. The same is true of healthcare and education facilities, and almost every one of our basic infrastructure investments and services. Unfortunately, the cost of correction is generally far greater than the original provision, with an extreme impact on ecosystems and lifestyles.

In the computer sector, planning errors of this type are also expensive in terms of poor service and time wasted, but overall they are relatively cheap to correct. Probably the biggest misconception here is that by saving on bandwidth, storage space and processing power, money will be saved. I have yet to find a single case where any of this is true. There is here only one issue of importance: the loss of human time, creativity and opportunity.

So what is happening? When computer departments say the average traffic is 20 per cent, ask for the measurement period. If they say it is for an hour, day or week, ask them to repeat the measurements over 10 minutes, one minute, one second and one millisecond. They may well come back looking puzzled with results that climb rapidly along these lines: 20, 85, 220, 650, 10,000 per cent. The same is true for road traffic, hospital beds and jobs, but not on such an exaggerated scale. So the low cost of bits over atoms may just save us from bad decisions in the future.

Peter Cochrane holds the Collier Chair for the Public Understanding of Science & Technology at the University of Bristol. His home page is:
http://cochrane.org.uk