Mean Times

We like symmetry. Most of us do. There is something in our wiring or programming that finds symmetry attractive, pleasing, and embodying some balance that might actually communicate harmony. We see it in the YinYang and the Taoist philosophy. We often characterize justice as a balanced scale, and countless studies have measured our perceptions of beauty among individuals and found facial symmetry the driving attractiveness variable. When we measure and analyze to find meaning in data, there is also an underlying “hope” that we find symmetry. When we see a “normal” distribution, or bell curve, we enter a comfort zone. In fact, I know countless people who work terribly hard at converting data that is not symmetric or normal into a set that is. Some, actually too many, take out data that does not fit the beauty of symmetry, proceed to insult it with names like outliers and dismiss them from our view.

Loving the comfort of symmetry and filtering out that stuff that isn’t is often dangerous. This need for comfort leads to blinders we create (or are taught to create) when looking at our data. For example, the stuff we often call continuous and plot on a line graph isn’t really continuous, but it feels better look at it as if it was (this is for another day). So let’s get back to our problem with the quest for symmetry. Let’s look at two problems.

The first is with the (insulted) data that we label as outliers. Often the outliers that don’t fit the symmetrical gang are in fact the early messengers that a big change from what we expect has happened and that a whole new family of subsequent consequences may need our attention. There is a phrase that I will paraphrase, “One bad event can erase a whole bunch of good ones.” When the consequences of one bad are really big, seeking symmetry is very dangerous. There are many people today who were devastated in the Midwest from flood damage for which no insurance coverage existed (Katrina victims had their own hell.) There are guilty parties who gave bad advice. While some will argue that “the rules” say to investigate the outliers, it seldom happens, even with those that give that advice. So, with advice, the advice to us is caveat emptor.

The second problem is one that I frequently rant about; while performance or inputs often are symmetrical, consequences are typically not. Consequences are more asymmetrical. It is true in the markets, weather patterns, traffic accidents, air traffic congestion, business deadlines, modern warfare, and many more. An easy one to visualize is the process many use to estimate time and materials for a proposal. Many organizations estimate using what is called a “unit rate” from past performance which is either the mean (average) or another calculation of central tendency (frequency or likelihood). We add a whole bunch of those unit rates up to come up with a total estimate. With all that estimated symmetry, we should expect that half the time we will below that value and half the time above (almost never at that value). These symmetrical probabilities always yield very asymmetrical consequences. We often forget the symmetry and actually expect the estimated value (not a good choice). The benefits that may ensue from coming in below the expected value are typically smaller in impact than the downside effects of being above the value. Being late or falling behind are often really bad news.

• A proposal submitted a day late is rejected without opening.
• Missed delivery dates are subject to penalties or loss of business
• Running behind triggers overtime, expedited shipment, errors and stress
• Keeping an executive waiting for a report can have career altering effects.
• The fear of being late typically results in producing too early, inventory and stale information.

So, on average, we may get more bad news than good.

Does that mean that mean brings meaner?