Perils, Pitfalls, and Pareto Charts
In my last blog, we compared fatality rates on Himalayan peaks to determine which mountain provides the biggest challenge for would-be mountaineers in search of death thrills.
Of course, if you’re an adventure/trekking company, you have another goal: to guide thrill-seeking mountaineers to the peak safely. You want to protect your customers from experiencing the ultimate defect: death.
So you need a tool to clearly identify the most important problems or causes leading to fatalities on the climbs. That calls for the graphical workhorse of quality improvement analysis: The Pareto Chart (Stat > Quality Tools > Pareto Chart).
The Pareto chart below summarizes the causes of recorded fatalities on 15 Himalayan peaks, including Mount Everest.
Tip: To clearly show where the cumulative number of fatalities hits the critical 80% mark, I added a reference line at 80 on the secondary Y scale (right-click and choose Add > Reference Line).
As you can see, nearly 80% of fatalities on Himalayan climbs are caused by just four things: Avalanches, Falls, Disappearances, and Exposure. If you’re a guide, those are the dangers to focus on to make the biggest impact in safeguarding your clients.
You can’t prevent an avalanche, of course, but you might be able to anticipate on which route the avalanche risk is greatest, or under what conditions or time of year one is more likely to happen.
For example, after subsetting the data to focus only on avalanches on Everest, I used Stat > Tables > Tally Individual Variables to quickly summarize the avalanche fatalities on Everest by route.
Close to 60% of avalanche fatalities on Everest occur on just 2 routes. Pretty important info, and it was the Pareto Chart that pointed me in the direction to discover this.
Good Categories Make Good Pareto Charts
A Pareto Chart is only as good as the categories that you use for it. The most common mistake is not categorizing the data well. Watch out for these pitfalls:
• Categories are too broad. You may end up with too few categories, making the chart vague and oversimplified.
• Categories are too narrow. You may end up with too many categories, making the chart unfocused and complex.
• Some categories are broad, some narrow. An inconsistent level of generality may cause you to miss or incorrectly rank leading causes.
I found a good example of the third pitfall on a U.S. Department of Energy-related web site. The U.S. Navy wanted to identify primary causes of equipment failures. They created a broad category that counted Personnel Errors as one cause. But they subdivided Mechanical Failures into several narrower categories, based on the precise type of mechanical failure. As a result, Personnel Errors appeared to be the leading cause of failures, and became the focus of their quality improvements efforts. Mechanical Errors actually caused many more failures, but inconsistent categorizing caused them to miss this leading cause.
The categorizations in the Pareto chart for Himalayan fatalities could be better. There’s overlap across categories like Exposure, Exhaustion, and Cold and Exhaustion. They should be clearly divided into just 2 separate categories, or combined into a single category. Disappearance is a bit unclear—in some cases it could mean an Unknown cause. In other cases it may have been a Fall that wasn’t witnessed or a storm that caused a climber to be separated from the group.
For this Pareto chart, we're limited to the causes (categories) recorded by the climbers. Of course, it probably wasn't easy recording data in an environment where winds can reach 320 kph (200 mph), temperatures can plunge to minus 73 degrees Celsius (-100 F), and there's up to 70% less oxygen than at sea level.
I hope you aren't facing those conditions when you collect data. Luckily, a Pareto chart doesn’t have to be perfect to provide valuable insight. It's definitely a heckuva lot more forgiving than Mount Everest.