Three Lessons CrossFit Taught Me About Data Science

Coming back to the blog after a long hiatus hurts almost as good as coming back to CrossFit after a week off of workouts while in San Diego for the Gartner Digital Marketing Conference (#GartnerDMC).

Among the many topics we covered last week – and my personal favorite / frequent soapbox – was Big Data and Data Science Basics for Marketers. If you missed it, or were so captivated (or horrified) by my on-stage demonstration of air squats and thrusters (in heels!) that you forgot to take notes, fear not. I’m bringing you the cliffs notes version this week.

With all of its intimidating equipment, elite [m]athletes, insider jargon and acronyms, cult-like following and ridiculous hype, it’s no wonder many marketers consider data science out of their league.

I used to feel the same way about CrossFit. It’s worth noting that despite my attachment to this metaphor, I’m no good at CrossFit – you’d laugh if you saw me try to do a pull up with that giant rubber band…let’s not even talk about snatches.

But here’s the thing, I don’t have to look like or aspire to become the level of athlete we see competing in the CrossFit Games to get the benefits of the workout, to build some new muscle and maybe even prolong my life.

Learning about data science is like cross training for marketers. You don’t need to go pro, but getting the basics down will improve your performance and prolong your career. You can’t beat the algorithms, so you might as well understand them.

Here’s what you need to know:

Data science is multi-disciplinary. Most of the exercises that make up a CrossFit workout aren’t new. What makes the workout effective, not to mention ridiculously difficult, is the combination of moves which draw from three existing disciplines: high-intensity bootcamp, gymnastics and power lifting. Data science similarly sits at the intersection of three business disciplines.

You’ll often find data scientists who are deeper in one area compared to others. Many start with a statistics or computer science background and build up the business expertise on the job. Conversely, a marketer with a bit of a quant background and plenty of substantive expertise may take online courses or self-teach to develop the coding skills required. Regardless of how she acquired them, the bona fide data scientist will have mastery of skills from all three.

It’s not all about size.The prevailing misconception is that big data is synonymous with or requisite to data science. (It’s not.) Big data and data science are complementary, but they aren’t the same thing. Big data can refer to a type of storage or a type of input, while data science is what you do with the data. Basically, big data is the equipment; data science is the workout.

While we are on the topic of size, let’s get a few things straight: Data science doesn’t need big data. In fact, marketers have been doing data science on what we might call small data for at least 25 years (it was called modeling back then). Think of it like a bodyweight workout – data science burpees, if you will; no fancy gym or Hadoop environment needed, though most data scientists feel pretty comfortable with the “heavy” stuff these days. And by the way, size, or volume as Gartner refers to it, isn’t the only characteristic of big data. Big data might be high velocity or high variety too.

A data scientist needs to have the strength, speed and agility to work with all kinds of data, be it big or small, structured or unstructured, historical or streaming.
Step aside, meat heads.

It’s a scalable foundation for you to build on. Many marketers – like beginner CrossFitters – approach data science with huge expectations and even bigger misconceptions. Organizations don’t magically transform from luddites to Amazons and Netflixes overnight. Muscle and peak performance are developed over time, starting small with a focus on mastering foundational movements. Understanding supervised and unsupervised machine learning and regression and classification problems is like getting the right form down by doing air squats and press ups. As you get more comfortable and natural with those movements, more complex ones will become available to you. Start with a small project and a manageable data set, learn the basics and build a plan for increasing your maturity over time.

Hint: A great way to get comfortable with foundational algorithms is to check out the cheat sheets on scikit-learn.org and Microsoft Azure Machine Learning Studio.

Remember, in business, just as in CrossFit, you should leave the heavy lifting to the pros until you get your form right. Ask your vendors smart questions about the data science that goes into their solutions, work with service providers to augment your internal capabilities (we have a Market Guide for that, clients only) or talk with us about when and how to hire a data scientist for your team.

If you’re feeling inspired to build some data science muscle, be sure to check out this week’s content line-up (for clients only).