17.4 “Take this Dataset and…Do Something!”
We have spoken with quite a few data analytics job applicants who received a dataset during the hiring process with an open-ended set of instructions, along the lines of “take this data, do something interesting with it, and let us know what you find.” In many ways, this is far more effective than would be the case if the company instead handed you a dataset with a clear set of step-by-step instructions. After all, who is going to add more value to a firm – a new hire who requires specific guidance for each task, or someone who can work independently to arrive at an interesting, useful result?
In such a scenario, start with the Exploratory Data Analysis (EDA) fundamentals that you have seen in this book, or in other sources that you have encountered. Summary statistics, with or without groupings by category, and data visualizations can form the core of this analysis while demonstrating your ability to work with ambiguity.
Visualizations in particular are a great way to approach a scenario like this. Remember that plots and graphs tend to land well with audiences, whether they are data novices or data experts. Be absolutely sure to include titles and labels, and to be completely familiar with the content.
If the dataset includes a natural response variable (for instance, total sales for the year or quarter) you could explore possibilities for modeling, perhaps using linear regression to predict the total. If it contains a categorical variable like ‘churn’ or ‘renewal’, you could explore a classification model.
No matter what, do not answer this prompt with some fancy modeling technique just because it “sounds cool.” If your interviewer is a seasoned analytics professional, he or she is not going to be bowled over just because you said “extreme gradient boosting” or “convolutional neural network.” In fact, using a complicated modeling process may just invite more questions – if you feel ready to handle those questions, great; if not, you would be wiser to not invite that kind of extra scrutiny.
And especially no matter what, do not present some complex algorithm, and, when questioned about it, tell your interviewer, “It’s too complicated for me to explain to you.” Of course, we cannot speak for every company. But if someone said that to us, the interview would not last much longer.