Underrated skills in Data Science

5 Minutes Read

So you want to break into Data Science. Excellent career choice! There are plenty of opportunities for capable professionals in the field. Notice, however, the emphasis on “capable”. One of the reasons why companies are struggling to hire is that most candidates they meet are just not good enough. Maybe surprisingly, the most common areas of deficit are not technical.

In this article and its follow-up, we discuss some crucial skills that you need to develop to find the job of your dreams. Bad news: You will not learn these in a Data Science MOOC. Good news: You can practice them every day in your real-life™.

The talent shortage

Talk to any company leader or investor, and they will tell you that hiring in the fields of Data Science or Machine Learning is a genuine headache. How can that be? Here at LyRise, we receive dozens of CVs for every job we post. It cannot be that hard to find one that fits the role! 

Sadly, there are three fundamental problems:

  1. Yes, we receive lots of CVs. But each of these candidates has sent their CV to multiple other companies. We are all competing for the same resources.
  2. The practical experience of most candidates is restricted to the projects they completed in online courses or in their academic studies. One way to stand out as a beginner is by presenting a few meaningful projects that use real-life data and solve real-life problems.
  3. Too many candidates consider Data Science to be a purely technical field. This notion is reinforced by the vast majority of the material available online. In truth, a Data Scientist is primarily a business problem solver who uses technical tools.

Points 1 and 2 are well documented elsewhere, so in this article, I will focus on point 3. 

Misconceptions around the role of a Data Scientist

Aspiring Data Scientists typically come from a STEM background, such as computer science or engineering. As a result, they tend to display a strong bias towards quantitative or process-oriented skills. This bias is strengthened by most online courses and blogs, which focus primarily on the technical elements of the role: mathematics, coding, machine learning, etc. In the end, most of us are in this field because of our interest in such activities.

But why do companies need Data Scientists in the first place? What delivers value for them? Certainly not iterating through 10 different ML models, at least not directly. What adds value is the end result: recommendations for executives or a customer-facing product (whose customers can be internal or external).

In both cases, the skills required to deliver go well beyond the technical. Imagine your company is launching a new product and has asked you to recommend a pricing structure for it. The project might go something like this:

  • Meet the main stakeholders to fully understand the project goals and deliverables;
  • Define performance metrics related to the business goal;
  • Meet the product manager to understand the product, its target audience, the market segments it addresses, the competition, how the product will be sold;
  • Meet with the pricing specialist to understand the pricing structure of the target market, competitor pricing, etc. Collect historical pricing data on similar products;
  • Brainstorm with team members about the data and information that we might need;
  • Research previous approaches to similar tasks;
  • Meet with the sales department to gather historical sales data and future sales targets;
  • Meet with the communication manager to collect data on past and future campaigns, as these will impact sales directly and add noise to our model;
  • Explore the available data. Form hypotheses about the price sensitivity of different market segments and influencing factors;
  • Come up with helpful features to inform the recommendation model;
  • Develop a proof of concept and measure its performance;
  • Present the proof of concept to stakeholders;
  • Iterate and gradually improve the prototype until you meet the goals;
  • Present the final outcome. Explain how it works and draw attention to potential caveats.

Obviously, in real life, things would not go linearly as above. We have not yet found a way to write cyclically in human languages, although aliens have solved this issue. Most steps in this list are not technical but are about connecting business issues with technical solutions, working with colleagues, and thinking about how to approach problems.

Of course, statistics, programming, and a knowledge of ML algorithms are all essential and require commitment. However, these are systematically assessed in job interviews, so it is safe to assume that most short-listed candidates have them. What might be trickier to master and measure are behavioural qualities, commonly referred to as soft skills.

Most underrated skills in Data Science

This preamble brings us to listing the most underrated skills in Data Science. And I mean underrated by candidates and education providers, certainly not by employers! This list reflects our interactions with companies having difficulty finding qualified candidates. If you are one of these employers, register on our platform today, we will help! 

To keep this article short, we will only cover one group of skills today. Future posts will complete the list.

Business acumen

According to Wikipedia, business acumen is “keenness and quickness in understanding and dealing with a business situation (risks and opportunities) in a manner that is likely to lead to a good outcome.” More generally, it is one’s ability to understand the business environment they are operating in and the business consequences of their involvement.

This is the main pain point of many recruiters. Candidates are technically savvy but do not understand how they fit within the company's business model, nor the consequences of their work.

Understanding marketing operations is essential to analyze sales data, for instance, in the pricing example above. Otherwise, sudden bumps in sales might be misinterpreted as emerging trends when they are just the result of one-off promotional campaigns.

Here is another example: if your company sells its product through a distribution network, the only price it will be able to dictate is what distributors will pay, not what the end customer will pay. Therefore, your company has an indirect influence on the retail sales price. This is likely to lead to noisier data.

A key element of business acumen is the ability to switch between business and analytical considerations. More specifically, you need to be able to translate business problems into Data Science questions, and Data Science predictions into business recommendations. For example, how will you measure the quality of your pricing model? Sales figures? Revenue? Gross profit margin? Customer satisfaction? And how do you translate these metrics into an ML performance score? How do you ensure that the two sets of metrics correlate?

How to develop business acumen:

Obviously, a lot of it will come with experience. But you can speed things up by showing curiosity and interest in what other teams are doing. Go and meet them, ask them to talk about their jobs, and understand how different functions interact with each other.

"But", I hear you ask, "I am not employed yet. How can I do what you suggest?" It is a little harder, but it is also much easier than before. Say you are interested in a job at some car manufacturer. You can first learn about the automotive industry – not the available car models, but how these products are engineered, who produces them, how they are sold, who the customers are, how they are financed, etc. The answers to these questions are much more complex than you might think. Then you can use LinkedIn to reach out to industry professionals whose roles are relevant to the position you are chasing. Inquire about them (everyone's favourite topic! ), their role, the company, and how the Data Science function would affect them.

To be continued

That’s enough for today. I hope that I have started to convince you not to focus only on technical competencies, but also to be aware of the wider context in which you will be working. In our next article, we will continue to go down the list of underrated skills in Data Science. See you soon!

Picture of Luc Frachon

Luc Frachon