Statistical Thinking and Leadership Potential

Jeanne LiJeanne Li is a senior research scientist at Amazon, who applies her research and statistical skills to workplace health and safety. Before joining Amazon, she worked as a research statistician in biomedical science and published numerous peer-reviewed articles.
Outside work, Jeanne Li is passionate about using statistical thinking and knowledge to augment day-to-day decision-making. If you’re interested in such topics and would like to collaborate on similar articles or book chapters, you can connect with Jeanne via LinkedIn.
The opinions expressed in this article are her own.

Statisticians and data scientists (I will call them “data scientists” in this article) are in demand more than ever to transform ever-increasing data into insight so leaders can make informed business decisions. However, data scientists’ value does not stop there. While I agree with John Tukey’s famous saying, “The best thing about being a statistician is that you get to play in everyone’s backyard,” I’d like to argue that data scientists get to play in their own backyard, as well. Their training and experience with statistics equips them with exceptional potential to become leaders in their businesses.

Reference Group/Baseline

Data scientists can, for example, make more meaningful inferences from data by calling out reference groups for comparison. For instance, according to a 2017 statistic from the Center for Talent Innovation, women in tech in the US leave the tech field at a rate 45 percent higher than men. Here, the reference group is men. In contrast, a reference group could also be oneself. For example, compared to Kate’s normal body temperature, her temperature is a half degree higher today.

The same principle could be applied to making sound judgments in one’s career, as we often take into consideration the reference group or baseline. I have a mentee, let’s call her Ava, who is usually an upbeat person and never speaks ill of anyone. One day, she came to me confiding that she didn’t know how to approach a particularly aggressive colleague. My radar tuned into this right away, knowing where her baseline is.

That same sensitivity could be appreciated in a people manager, too. If a generally positive, no-fuss employee came to you with critical feedback, you would know you need to pay utmost attention, as opposed to just brushing them off. On the other hand, if a regular complainer complained to you, you might choose to tune out a bit.

Machine Learning

As you probably know, machine learning was initially developed based on human cognition. Now that machine learning has matured over the last decade or so, there are principles data scientists can appreciate and apply beyond data science projects.

Our statistical training not only equips us to work on data science projects using our statistical expertise, but it transfers to larger business environments where we can flex our leadership skills.

One of my other mentees—let’s call him John—recently joined a new company and came to me wondering if he is bugging his new manager too much. I asked a few questions and came to understand that John’s new working environment is fast moving and there is a lot of ambiguity surrounding his projects. Knowing that his baseline is more autonomous and self-reliant, I encouraged him to continue to “bug” his manager so he could get better clarity on his projects. Then, once he’d gathered the necessary data and become acquainted with his new environment, John could return to his baseline and make small behavioral adjustments from there.

Similarly, in machine learning, we start out with training a model using a ton of data, and once we have validated the model, we only need to continue fine-tuning its parameters.

Probability and Sample Size

Probability has a whole range of applications, especially in risk assessment. We know we should not rely on an outcome to infer the quality of a methodology used. For example, Method A with a 40 percent success rate may yield a success, while Method B with a 70 percent success rate may still yield a failure. Especially with a small sample size. We don’t rush to conclude that Method A is superior to Method B based on a handful of outcomes only, but rather focus on evaluating the quality (i.e., success rates) of the methods or testing them with many events.

The same principle can be applied to talent management. In most companies, internal promotions could be viewed as conservative and time consuming to complete (especially in the eyes of the employee waiting to get to the next level). Most companies do this to ensure employees promoted to the next level will continue to perform successfully in their new position.

I heard from a people manager colleague that one of his employees pulled off an important project with huge success shortly after coming onboard, and my colleague was considering promoting that employee; however, it didn’t take long for the employee to start creating more disasters than contributions. It stands to reason that we say career (or life for that matter) is a marathon, not a sprint. From a statistics perspective, this can be explained by probability and sample size.

In summary, our statistical training not only equips us to work on data science projects using our statistical expertise, but it transfers to larger business environments where we can flex our leadership skills. Don’t discount leadership opportunities just because you’re technical!