Essential Data Science Skills

There was once a time when all you needed to get a job in data science was basic R or Python skills and a knowledge of machine learning equivalent to the content of your average MOOC.

Or so I heard. I wasn’t lucky enough to get into data science that early.

I first heard about data science back in 2015, almost three years after Thomas H. Davenport and D.J. Patil named data scientist “the sexiest job of the 21st century,” and by that time, competition for data science roles was already starting to become stiff.

With interest in data science growing with each year, and many universities now offering Masters degrees in Data Science and Analytics, I think it’s safe to say the competition has only become more intense since then.

It’s also safe to say that completing the Coursera Data Science specialization is no longer enough to make your resume stand out from the crowd.

So, if you can already program like a boss, and you’ve done so many machine learning courses that random forests and neural networks haunt your dreams, what skills should you focus on next to get your foot in the door for your first data science job, and level up further once you get there?

Why don’t we ask the employers?

The Most Important Skills for Data Scientists

To find out what employers are looking for in data scientists, I examined 100 data science job advertisements, across four English-speaking countries (Australia, Canada, UK and USA), found on LinkedIn between 22 April 2019 and 5 May 2019.

The job ads were selected to represent a broad cross-section of employer types, sizes, industries and job levels, and purely managerial roles were not considered.

From these job ads, I manually extracted details of the skills listed as either selection criteria or in the day to day duties of the role, and determined the top 20 most requested data science skills (excluding programming languages and technologies), based on the proportion of the job ads in which they appeared.

Here they are:

These skills can be broadly divided into three categories:

Modelling and Statistics: including machine learning, statistical modelling and model deployment;

Data Engineering (and Programming): including data wrangling, working with databases and distributed computing (if we had included programming languages in our analysis, this is where I would also put them); and

Communications and Expertise:including providing insights, industry knowledge, and leading and mentoring junior staff.

From there, it is possible to identify six skills you can work on developing to build your data science prowess across these three categories, dramatically improving your employment prospects:

Modelling and Statistics

Statistical Modelling

Model Deployment

Data Engineering

Working with Databases

Communication and Engineering

Providing Insights from Data

Leading and Mentoring Junior Team Members

Communication

Which Skills Should I Focus on Learning First?

Of the 100 data science job ads that I extracted, 15 were for entry level roles (defined as roles with “Junior”, “Graduate”, “Intern” or similar in the title); 44 were for mid-level roles; and 41 were for senior roles (defined as roles with “Senior”, “Principal”, “Lead” or similar in the title).

The table below shows the proportion of job ads that mention each of the six skills previously identified, at each level, along with the ranking of each skill.

At all job levels, “presentation and communication” consistently ranks within the top two skills sought out by data science employers. Furthermore, the communications-related skills of “explaining technical concepts to a non-technical audience” and “working with clients/stakeholders” both become increasingly important as job seniority increases.

So, it stands to reason, that if you had to choose one skill to focus on learning first, communication would be the way to go.

But why stop at just one?

“Statistical modelling/techniques” also ranks within the top three skills at all job levels, making it a good candidate for skill development activities, particularly for data scientists at the early stages of their careers.

Moving from entry level roles to mid-level roles, the proportion of job ads that mention “statistical modelling/techniques” as a desirable skill increases from 53.3% to 86.4%.

However, for those looking to move into senior roles, developing the skills needed to lead and mentor junior team members may be a better alternative, particularly for those who have already developed their skills in statistics.

Whereas only 6.7% of entry level data science positions require leadership/mentoring skills, these skills are required in 58.5% of senior roles (an 873% increase), making them the 5th most requested skill at this level.

Conclusion

Mastering all the skills mentioned above is likely to take some time, but that’s OK. If you could master everything there was to know about data science in a weekend, would you really want to devote years of your life to working as a data scientist?

However, by focusing on just one or two skills at a time, and devoting just a few hours per week to developing them, eventually you will build up an in demand skill set, which will help you to stand out from the crowd and increase your chances of getting the next data science position or promotion you apply for.