I have taught over a thousand students about research methods and data analysis over the past 26 years. When you teach that many students, you begin to pick up on their common stumbling blocks to learning. And one of the biggest stumbling blocks is understanding the core concept of all analytics: variation.
Variation is a critical concept in all analytics. Some researchers might say it’s the most important concept in all of analytics. For this reason, variation deserves a specific place in our discussion of analytics.
What is Variation?
The Merriam-Webster online dictionary lists 11 different definitions of the word variation. The one definition most relevant to data analysis is:
A measure of the change in data, a variable, or a function.
A simpler explanation is that variation is the collection of differences in a characteristic across a set of observations.
If you have measure of individual heights for each member of your family, there will be differences. These differences are the variation in height across family members.
And of course, you can find variation across many types of observations and characteristics.
When you think about variation, it is useful to contrast the idea against its opposite: a constant.
A constant is a characteristic that doesn’t change across a set of observations.
For example, if you collected data on the physical characteristics of a group of dogs, the species of your sample would be a constant.
With an understanding of what we mean by variation and constants, you are prepared to understand their importance in analytics.
The Importance of Variation in Analytics
When it comes to data analysis, you sometimes need constants to help frame your projects.
For example, you might choose to limit your data to only analyze certain types of people, places, things, or events. If you do this, then you are restricting some characteristics to be constants.
Typically, however, your primary focus will be on the variation in your data.
You want to understand how your data differs from one observation to another. For example, a hospital might want to understand how its available bed count changes from day-to-day or week-to-week.
You want to know if an outcome can be explained or predicted by a set of inputs in your data. For example, an e-commerce sports equipment company might want to predict weekly sales based on local, state, and national sporting events, weather reports, and extracurricular league schedules.
Or you may want to know if implementing a specific treatment might cause a change in activity. For example, a company might test out a variety of marketing campaigns across different markets to identify those campaign elements that resonate most with different customers.
In each of these examples, the focus of the analysis is on the variation in the data. Whether you are using descriptive, explanatory, or evaluation analyses, the differences across observations are generally where we find the most interesting information.
Additionally, virtually every analytic methodology available focuses on variation. From calculating a simple average to the most sophisticated deep learning neural network, variation provides the fuel that analytics thrives on.
Incorporating Variation into Your View of Data
Based on the discussion so far, I hope it’s clear that variation is a big deal in analytics.
In fact, no one talks about it much, but variation is a big deal in problem-solving and asking questions. And the more you incorporate variation into your perspective on data, the better you will become at working with analytics.
Problem-solving and critical thinking are incredibly important skills for every professional. Both of these skills require an implicit understanding of variation in people, places, things, and events.
After all, if there were no variations, you would need to think critically about anything. You would simply find a solution that works once and implement that solution every time to get the same result.
Unfortunately, many problems are not created by constants…rather, it’s the variation that creates problems.
The same goes, to a slightly lesser extent with asking important questions.
Sometimes, important questions will focus on a concept that is constant. More often than not, however, it is the variability you find in life that spurs the most important questions.
- How do you prevent patients from having adverse reactions to medications?
- Where should we deploy limited police resources to help prevent crime?
- What factors are most important in driving consumer purchasing decisions?
None of these questions have an answer that is a constant. You won’t find a one-size-fits-all solution.
The best analytic questions focus on variations that are the most costly, inconvenient, and undesirable. If you understand the who, what, when, where, why, and how of these questions, you’ll be in an excellent position to mitigate the problem.
The next time you look at some data or analytic results, think about the impact of the variations.
Conclusion
Before you read this article, you might not have given much thought to the concept of variation. I don’t blame you. It’s a word that doesn’t get much airplay in popular culture.
However, variation is a core concept in all analytics. It is implied in every act of critical thinking and problem-solving you engage in. Heck, variation plays a hidden yet pivotal role in forming great questions. No matter where you are on your analytics journey, the concept of variation is not far behind. From the simplest descriptive statistics to the most complex artificial intelligence applications, variation is the fuel that feeds your analytic machine. Keeping this in mind will help you build your knowledge base faster and more effectively as you grow.
