“Statistics is not math!” This is a statement that was made to me very often by an engineering leader back when I worked at Motorola. I was brought into an acquired company to teach them how we used quality tools. We were troubleshooting a product issue and trying different solutions. I would compute some basic statistics or create charts to see if the solution was working. The “fuzziness” of my answer would cause this particular leader to say, “statistics is not math!” He did not literally mean that statistics is not math, rather he was expressing a common challenge with statistics, which is that we often do not arrive at a definitive answer.
Statistics is important! In an age when opinion and fact are constantly blurred, statistics is valuable to help us separate strong causal relationships from weak ones. In an increasingly complex world facts get muddy because we can always find scenarios that challenge the so-called fact. Try this as an example: Driving distracted causes accidents. That is a fact, right? Years ago, I drove with a colleague on a business trip. He would literally take his eyes off the road and read his phone for what seemed like minutes. After seeing him do this three times I told him that he could either stop looking at his phone or stop driving. Fortunately, he stopped looking at his phone. Looking at his phone while driving was normal for him. As far as I was aware he had never been in any serious accidents. So, this must mean that the distracted driving warning is a complete fallacy, right? Hold your answer, please.
In physics events are highly deterministic. Gravity, (on earth) acts the same way on an object every single time the object is dropped. Modeling this phenomenon is a straightforward mathematical approach. In my earlier story, this is what my engineering colleague referred to as “math.” Statistics help us understand phenomena that are not quite so deterministic. Outside of hard sciences, pretty much any event having to do with humans or nature can occur multiple times with different results. Anyone that has taken a statistics class or has done any sort of gambling knows all about dice. You roll a dice once and you get a 6. Then you roll it again, keeping everything the same and you get a 1. Why? What changed? For people that like the consistency and stability of a deterministic world this is highly disconcerting.
Statistics is math; it just uses different rules. Statistics is important because it allows us to step outside the hard sciences and continue to model and predict what will happen given variation of unknown factors. We model based upon what we know about an event. We can create a mathematical model just the same, only it has an error term on it and the result has a confidence interval associated with it.
“But why can’t you just tell me if this worked?!” is what I often hear from business leaders or engineers that do not get it. If it were only so easy… You flip the switch off and the lights go off. You flip the switch on and the lights go on. You tell an employee to turn the lights off every time they leave work and most of the time the lights are off at night, but some of the time they are left on. Why did the employee not maintain the habit? What changed on this day versus all the other days that they turned the lights off?
We may not find the root issue for this individual occurrence of not turning the lights off. The statistics help us understand the likelihood that the employee will turn the lights off on a consistent basis. Then we can further investigate the organizational system to determine what we might change to increase the likelihood of the desired outcome (the employee turning the lights off). If we provide no guidance then we can expect that the employee will turn off the lights 8 out of 10 times. If we instruct them properly they will turn the lights off 9 out of 10 times. If we put a sign up as a reminder and provide feedback on a regular basis they will turn them off 99 out of 100 times. Furthermore, if we implement a motion sensor the lights will be turned off about 99,999 out of 100,000. This is a fictional example, but you can see how we use statistics to determine if our interventions are moving the organizational system to higher levels of performance.
This becomes more important when we start thinking about a call center and helping our call center agent enter the correct customer address in the order management system. It becomes important when we think about stocking the products that our customers are most likely to order so that we can meet their delivery expectation and out perform our competitors. It becomes important when we think about the practices needed to establish a safe work environment.
As for my question about distracted driving, the drive with my colleague was quite a long time ago. I do not know if his distracted driving behavior changed or not. I hope it did and I hope that he never got into an accident. Thankfully, we did not get into an accident that day. Regardless, this one sample certainly does not mean that distracted driving being dangerous is some sort of fallacy. It is quite the opposite. The likelihood of an accident increases in magnitudes. Statistics generated from good research illuminates this risk. They allow us to study and model events that do not have simple deterministic outcomes so that we can predict the likelihood of achieving the desired outcome. The answer is not going to be definitive and the prediction is not always going to match one individual’s experienced reality. However, it allows us to study and improve in the non-hard sciences which is the environment that most of us work to improve.