Let’s talk about something that has been bugging me lately. You know that feeling when you’re like a kid in a candy store, except the store is filled with statistical tools and machine learning algorithms? Yeah, I am sure that we have all been there.

Faced with heaping amounts of data, one can be tempted to go straight to one’s arsenal of statistical and machine learning tools to make sense of the data and tease stories out of them.

Here’s the thing: Just because one have this amazing toolbox filled with t-tests, ANOVA, regression analyses, and fancy machine learning algorithms doesn’t mean we should throw them all at our data like we are having a statistical party.

Trust me, I’ve learned this the hard way.

The real magic happens when we start with the business question first.

And that can only be gleaned by talking with the business and problem owners first, thinking through the problems, and developing structured data and analytic strategies that will lead to a potential solution. It is like using a hammer to open a door: Sure, you could open the door with a hammer, but probably, you just needed to turn the knob.

I think when data becomes available, one should resist to dive in and just apply every single test and algorithm to it. Instead, one should be guided by what I can probably call as North Star of any analytics projects — two questions that need to be answered before any analytics project is undertaken:

1. What is the actual business question or business problem that we are trying to answer — and why is it important to answer this question?
2. What decisions and actions will be made based on our findings?

Without answering these two questions, it seems like one is only doing statistical gymnastics for the sake of showing off one’s expansive toolbox. Relevance goes out of the window — and potential frustrations can surely creep in.

The “So What?” Factor: If you can’t explain how your analysis will lead to actionable decisions, you might need to step back and reassess your approach.

Remember: The most sophisticated analysis in the world is worthless if it does not help the business and problem-owner make a better business decision. As analysts, our job is not to show how many statistical tools and machine learning algorithms we know — it is to solve real business problems and drive meaningful actions.

So… the next time you are tempted to dive straight into your toolbox of t-tests and ANOVAs and regression and XGBoost and SVMs, pause and ask yourself: What is the real question here? What are we going to do with the answer?

Your stakeholders will definitely thank you for taking the time to answer these questions.

And if you are like me, you would be less frustrated.