What does a Data Scientist really do?

September 12, 2016

According to the Harvard Business Review, “Data Science is the sexiest job of the 21st century”[1]. Despite suspicions this conclusion was drawn by a Data Scientist, there is no doubting that this new job title is both exciting and in serious demand.

Plenty of confusion exists as to what a Data Scientist actually is, which is unsurprising given the Wikipedia article about Data Science was only created in 2012[2]. Confusion is further compounded by lots of analysts and engineers desperately trying to re-spray themselves as Data Scientists, hoping to access the salary premiums currently available.

Let’s start out with the data roles your business already has, and thus your Data Scientist role shouldn’t be:

• Every business has a team of people who answer questions about the company’s data, these people typically answer questions like “How many widgets did we produce last month”, “Which is our best selling widget in south east England”, or “What is the year on year change in margin from our largest customers”. Fact based questions with minimal room for interpretation, this is your team of data analysts. These are not Data Scientists even if they do answer these questions with fancy interactive dashboards in Tableau or QlikView.
• Sometimes you ask a question and get the answer back quickly, other times someone sighs telling you that the business doesn’t capture that data, or it isn’t easy to put in the place where analysis happens. This is a problem for your Data Engineers, they do a lot of plumbing behind the scenes to capture and put datasets together. Another job your Data Scientist shouldn’t be doing.

Data Scientist’s skills differ, they thrive answering open questions which have opportunity for interpretation, requiring a mix of skills to solve. With help from the business, their approach will involve breaking questions down into a number of hypothesis to test – some will pay off and some will be dead ends, this is the science part! Time will be spent understanding the business, the data, researching appropriate techniques, building complex analytics and an appropriate strategy to validate their answers.

If you’re asking questions like the below, you need to ask a Data Scientist, and you’ve absolutely worked out what they do:

• “Who are our most valuable customers?” A good data scientist would take this open question any start out determining “What makes a valuable customer?” as it may well be more than just your gross profit, it might include predictability and stability of their orders, the value of the customer’s brand association, and future revenue potential. Once they’ve built a model they can apply this to your customer base to get some answers.
• “Where should we build our next store?” This would involve understanding your current customer base, competitor locations, population demographics, business growth strategy. The end result should be a number of options each with risks, rewards and appropriate confidences.
• “What external factors influence our revenue?” A brilliantly open question which could have answers including combinations of weather, competitor marketing, online reviews, world economic factors, twitter sentiment etc… getting a good answer to his question could be very valuable, and is well outside the skill-set of a standard analyst or engineer.

Given the confusion of the role of data scientist; It’s currently quite common to mis-hire Data Scientists, dropping them into an analyst or engineer type role, or at least make this a significant part of their role. They can probably do this work, but they won’t enjoy it and will likely leave through boredom – you could also get that work done at a much lower cost. Love your data scientist, hire a really good one, and let them focus on what they’re good at, supported by the analysts and engineers where needed.