Principles of Design #55 – Normal Distribution

Feb 24 2014

Normal distribution is used to describe any set of data that forms a symmetrical bell-shaped curve when plotted, and comes from the work of Karl Friedrich Gauss, and hence is also known as the Gaussian distribution or bell curve. Such distributions are found commonly in nature: annual temperatures and student test scores are two examples. The bell-shaped curve that results from plotting such measurements shows a small number of cases at the extremes and a large number of cases in the middle of the curve.

For example, measurements of men and women (e.g., height, weight) are normally distributed. However, there are many measurements that can be taken from humans and there are no truly “average” people. The distribution can in principle give guidance for design – for example, in human measurements around 68% of us fall within one standard deviation of the average of any measure and approximately 95% of us within two standard deviations.

In such distributions, the average is also the most common measurement, but this is not always true of other variables (such as income). Nassim Nicholas Taleb has argued that many variables, including income and stock exchange fluctuations, are not truly random. He states that the difference is that in a normal distribution most observations are close to the average (in his words, “mediocre”), and that there is a very dramatic decline in the odds of finding cases as you move away from the average – the odds of finding someone 10cm taller than average are 1 in 6.3, but of finding someone 60cm taller (i.e. 2.27m) they are 1 in 1,000,000,000 (1 in 1 billion).

By contrast, if you look at the chances of someone having a net worth of 1, 2, 3, 4, 5, 6 million, the decline in the odds is much flatter, and income distribution data show that while mean (average) wages have been rising in developed countries, the median (middle case) has actually been falling. Similarly, Taleb argues that the normal distribution fails to account for extreme events (as in recent stock market fluctuations) and treats such data with more certainty and predictability than is warranted.

Even without his concerns, the average will not always (or even sometimes) be the preferred design parameter or reflect the real world. For example, a car designed for the “average” person will at best be suitable for 68% of people. In fact, the percentage will likely be much lower than this, for cars along with most products and services comprise a number of different features and dimensions. A person who is average on one of these variables, will not be average on others. Even the statistics of Gauss say that the chances of someone being average on two measures in only 7% falling to less than 1% for 8 measures or more.

For designers, this means that the focus should be on the 98% of the population who fall within a reasonable range of the average, considering that catering for wider ranges of people almost always increases cost (except arguably in the days of localised 3D printing). Typically, American airline seats will accommodate around 98% of American men.

However, the myth of the average (wo)man is a fallacy, particularly when it comes to product and experience design. Designers and researchers should seek to understand different target groups and design according to their specific needs. After all, if some people like coffee with sugar, and some like coffee without sugar, would that mean that you should launch a “slightly sweet” coffee brand?

REFERENCE

Universal Principles of Design by Lidwell, Holden & Butler

Fooled By Randomness by Nassim Nicholas Taleb

No responses yet

Leave a Reply