August 30, 2011 | Mark Paradies

What is a Real Trend?

I was reading an article about industrial safety in the UK. They were trying to make a link between the budget cuts for the UK Health and Safety Executive and increased fatalities in the UK in from June 2010 to June 2011. The 16% increase in fatalities caused some to call for a change in policies by the government.

This started me thinking … What is a real trend?

Of course, to each of the 171 workers who were killed and their loved ones, each fatality was a disaster that needed to be prevented. But does the number of fatalities increasing from 147 to 171 represent a real trend or just common, everyday variation – noise in the system – that doesn’t prove anything.

There is a way to tell. The math and trending techniques to judge trends were developed back in the 1920s by Dr. Walter Shewhart. It’s the same mathematical techniques used by Deming to improve quality and most six sigma programs.

All one has to do is to develop a Process Behavior Chart (see Chapter 5 of the TapRooT® Book for examples of these charts and the math required) for the past 20 years or so of data from the UK. One could then calculate the standard deviation of the data from one year to the next, and the chart’s limits would show if a statistically significant trend exists.

Of course, a trend would not prove the relationship between fatalities and UK HSE budget cuts. Other factors could be more important. For example:

  1. The general increase or decline in economic activity.
  2. The influx of non-English speaking workers, especially in more hazardous industries, into the UK.
  3. Company budget cuts that influence safety performance.
  4. Management attention shifting from safety performance to other activities (especially in a bad economy).
  5. Changes in overtime after downsizing or layoffs.

That brings up another point about trends. The more statistics are “averaged,” the less the statistics mean.

For example, what if we looked at industrial fatalities worldwide? What would year-to-year variation mean to us? Almost nothing. Increases in one country could be overshadowed by decreases in another. I believe that national statistics are almost useless for improving performance. At best, they are a fairly insensitive barometer to the general trend of safety performance if viewed over several years. (in general, you will see a decrease in the numbers, but the progress probably won’t be even.)

That should start you to think about your statistics and how you are using them.

Corporate statistics aren’t as useful as site statistics. And breaking site statistics down by looking at statistics by work type can make the trends even easier to detect. Once a significant trend is detected, one can draw causal implications much more easily.

Do you want to learn the math behind using trends properly and then plan to apply trending to your company’s operations? Then read this book:

Root Cause Analysis
Show Comments

Leave a Reply

Your email address will not be published. Required fields are marked *