A bunch of cowboys are all working a ranch. Miles from anywhere. Bored. They need something to pass the time when they are not herding cattle.
Now imagine that one of the cowboys on the ranch goes out to an old barn and decides to let off some steam by using the side of it for target practice.
Over the course of a year the side of the barn is peppered with bullet holes. Up, down, left, right. Most people looking at it would see it for exactly what it was: a load of random holes in a wooden barn.
Now the cowboy ain’t so stupid. He wants his buddies to think he’s become a real sharpshooter. So he grabs some paint, a brush and goes over to the barn and starts drawing circles around any clusters of bullet holes he can find that are relatively close together.
He knows how the mind works. He knows the human brain likes to find order and pattern in chaos.
So when the rest of the gang come to look, they will see a bunch of target circles that he has (apparently) been firing into. Each hole beside another one. Neatly grouped together.
They will walk away thinking he’s amazing, so accurate. That he was aiming for a target, rather than just the whole side of a barn. They won’t know that each bullet hole was randomly created.
It’s an example of the Texas sharpshooter fallacy. Random facts being grouped together to give, more often than not, inaccurate meaning and conclusions. The element of randomness is overlooked.
You see it in numerous conspiracy theories. Proponents will focus on the small amount of similarities and attribute meaning to them, while conveniently ignoring the many, many differences.
Maybe it’s an entrepreneur who Is hailed as a genius when their two successful businesses get lauded and their 50 unsuccessful ones ignored.
Some folks suggest a link between the John F Kennedy and Abraham Lincoln assassinations. Both married in their 30s. Both married women who were in their 20s.
Both killed in public. Both by gunshot. Both shot in the head. Both, in fact, shot by killers who were known by three names (John Wilkes Booth and Lee Harvey Oswald).
And the letters in those names both add up to 15 (Kennedy and Lincoln also both have 7 letters). There’s more: Both killed next to their wives, on a Friday and neither assassin would ever make it to trial. Both successors named Johnson.
Got to be a meaningful connection, right?
Not really. What about all the differences though. There are thousands. Sure, if you think long enough you can make a few random connections fit.
But only if you ignore so many, many, many others that don’t. Their predecessors’ names, their height, religion, birthplace, education, cities in which they died, where they were sitting when they died, the type of gun and so on. So many other differences outweigh the connections.
And so it is at work. Very often we use data to drive understanding of our business. Once we identify and understand patterns we can start to interpret, model and predict them, hopefully allowing us to create improvement.
What if three people in a company’s top 10 absence figures are all called Connolly and are aged between 30-45?
What if they all had under two years service? Does that mean that you direct all your effort towards other employees also called Connelly in that age bracket, within that service span?
What if, however, you also discover that they are all working in different departments, are no relation to one another, one is a woman, two are men and all suffer from three distinctly different types of illness? Maybe the connection looks a lot more tenuous now.
The trick is to work a hypothesis. Start with a theory and see if the data supports it, then test it thoroughly and objectively.
Use your base data to try and make connections but allow for randomness to be a factor, don’t rule it out as a possibility. Use common sense. Don’t fall prey to the mind looking for easy groupings, for order from chaos.
Get peer reviews. Ask someone else to look at your work and your conclusions. Do they agree?
Ensure your data is being fulsomely presented. Don’t cut sections out that skew what you want to try and prove. Let the data speak for itself.
Don’t paint targets round random bullet holes.
