Once you’re a trained data scientist, what will your new career really be like? Trace the experiences a data scientist might encounter on a typical day, including mining for good data, cross-functional meetings, lots of back-and-forth with internal clients—and, of course, some tough math.
7:45 a.m. – Quiet Time
You arrive at your Breakthrough Beverages office early for some quiet thinking time before the phone starts ringing and your inbox fills up with messages marked urgent. Breakthrough is a rapidly growing sports beverage company that markets its products as “naturally healthy” vs. its competitors, who it derides for their reliance on “artificial chemicals.”
Your top priority at the moment is predicting the appeal, measured by sales, of three different product features in an alternative beverage line targeting millennials: glass bottles instead of plastic, a preservative-free drink, and a low-calorie formula with no artificial sweeteners. You glance at the crystal ball your co-workers at your last job gave you as a send-off present. At times, you wish you had a real one.
Your first challenge will be getting the right data, and that will include more than sales records from other launches and projections from the sales teams. Facebook, Google and Twitter are the obvious social media options, but you’re leaning toward Instagram for its younger demographics. Looking at mentions and likes should be enough. You make a note to take a close look at these four channels, and to add a line item in the final project budget to cover the cost of external social media data.
9:32 a.m. – Putting Out Fires
Your smartphone announces a priority call. It’s Phil Brooks from marketing, asking for an update on some A/B testing for a web marketing promotion tied to “emerging sports”—skateboarding events, parkour courses and even lacrosse matches. You had handed the project off to Sandy Speith, a newly hired junior data scientist who’s not in yet. You promise to call Brooks back with an answer before noon—and then send your teammate an urgent email requesting a progress report.
9:55 a.m. – The Big Meeting of the Day
You walk over to the main conference room, named “Quench” after Breakthrough’s signature beverage, for an input session on a major new project about optimizing distribution. Breakthrough Beverages has grown so large that seat-of-the-pants strategies such as overstocking warehouses to ensure retailers can get enough product is costing too much. The team is hoping that the data science team can help them revise the distribution network to get product to retailers faster.
The room is full of heavy-hitters, including the two founders. You begin with some very simple ideas related to the variables you’ll need, such as data from manufacturing, logistics and sales down to the ZIP-code level. It will most likely also involve streaming GPS data. Using production, warehouse and ZIP-code level consumer take-away data, the right predictive analytics algorithm could reduce the amount of excess inventory in the pipeline and thereby cut costs. You deliver your standard explanation of the need for clean data, emphasizing that clean data isn’t free, although AI tools have made it a lot cheaper than it used to be.
When the back-and-forth begins, you listen very carefully. One of the biggest parts of your job is figuring out what your internal customers really need, as opposed to what they ask for. Today, you can see that the problem is as much about concern for having the freshest possible product on shelves as it is about inventory management, although these two goals aren’t contradictory. From a data perspective, you’ll need time-stamped input to understand where any delays might be occurring. You feel grateful for those general business courses you took back in college. Data science is more than math.
You close the meeting with an informal outline of next steps, which will include an initial look at whether the available data is useful for this project, one-on-one meetings to get a better understanding of current day-to-day procedures, and a timeline for when you can get everything done.
11:15 a.m. – Coffee and Decisions on the Go
You stop by the coffee bar for a quick espresso and then return to your office to sit down with Sandy Speith to discuss the sports promotion. She wants to try a multi-bandit approach instead of standard A/B testing. She makes a strong case that this new approach will send fewer users to suboptimal experiences. You give her the go-ahead. Now, it’s time for a run in the park next to your office building.
12:30 p.m. – Cruising Through the Inbox
Back in your office, you eat a sandwich from the cafeteria, accompanied by a (plastic) bottle of Mango Madness, one of the company’s hottest beverages. While you’re eating, you go through your inbox and respond to half a dozen inquiries about the progress of various projects. Then, it’s time for your one o’clock meeting with Dave Chen, who is an excellent data wrangler.
1 p.m. – Dealing With Doubtful Data
This meeting brings you back to the top priority of the day: the appeal of the three different product features. Within a few minutes, you confirm that the company sales data is not in good shape. You’re going to have to combine data from a number of systems and sources to create a data set you can use for prediction. That’s how data science works. You have to be part wrangler, like it or not. You spend the rest of the afternoon working with the sales data, and with the inevitable interruptions to deal with from people inside and outside of your team.
5:10 p.m. – Staying Current
You set aside your query for the sales data to check out your favorite data science blogs for new developments and tips. You also stop by Quora. Once upon a time, you posted questions there about a career in data science. Now you answer them.
5:45 p.m. – Time for a Beer
Your brain is tired after a long day of thinking. You head to The Buffalo, a sports bar where beer dominates anything Breakthrough might sell. Several of your friends from college will be there, and you might talk about sports, but tired brain or not, you’ll more likely argue about algorithms.
Check out the Data Science Academy to read more stories about data science on the job.2