Depth Versus Breadth In Data Science Jobs Skills

Depth Versus Breadth In Data Science Jobs Skills


Where do you go after you've started learning basic data science? You are already doing some exploratory data analysis with Python or R. You've done some modeling and machine learning as well. You may even have built data pipelines with a combination of R, Python, SQL, Unix, and other software. The question is - where do you go next?

Ask 5 data scientists and you'll get 6 opinions

If you go to Reddit, Quora, Cross Validated, Hacker News, or other places to ask this question, you'll get so many different opinions and "what worked for me" that it'll make your head spin and heart sink. Now instead of what to do next, you have 20+ tabs open, half a dozen book marks added, a few GitHub libraries starred, and more than a couple project ideas to try to do to learn.

Why is that? Why is it that the next step is so hard to do? And, most importantly for your data science job search - what does this have to do with getting a job?

Depth Versus Breadth In Data Science Jobs

If you look at data science job postings, there are usually 10-30 "keywords" listed as being directly relevant to the job. They range from excel to Hadoop and perhaps even specific types of algorithms or areas of research. In the article What tools do employers want data scientists to know? we wrote earlier, we saw 38 completely different technologies listed in only 30 job postings. This is usually one part of job postings that induces anxiety.

The reason for all of these tools is that some of the jobs listed are "breadth" data science jobs and others are "depth" science jobs. As you can well imagine, the "breadth" data science jobs will require you to be familiar with many more tools than the "depth" data science jobs.

The issue with where to go next and what to learn next then depends on the what type of data scientist is offering you the advice - a "Depth" or a "Breadth" person. Further, their opinions will vary along this same fault line as will their choice of tools.

Breadth Data Science Job Skills

Recently on Reddit, a commenter suggested that a person asking about the next steps should do the following:

  1. Learn Python to scrape websites
  2. Learn SQL to take that data and store it in a database
  3. Learn PostgreSQL to best store that data
  4. Learn Unix/Linux command line tools to run a cron job
  5. Learn Python data libraries (Pandas, Scikit-learn, as well as others) for the analysis part
  6. Learn Flask or Bottle or Django to host the analysis of the database as a RESTful API
  7. Learn web hosting to host this application on the cloud
  8. Learn AngularJS to consume and present the results
  9. Learn D3.js to create visualizations from this data

That seems like a great deal of things to learn with varying technologies, methods, knowledge bases, and platforms. This encompasses not only the full data pipeline but also full presentation mode of the data insights. This learning roadmap sounds like you'd learn enough to be a data engineer, data scientist, a back-end web developer, a front-end web developer, and a data visualizer.

Yet, there are certain data science jobs and team where this would be your responsibility and well within your grasp. So all of these things would certainly be considered as 100% important to your data science job skills.

Depth Data Science Job Skills

Also recently on Reddit, a commenter suggested that a person asking about the next steps should do the following:

  1. Learn core statistics
  2. Read "Intro to Categorical Data Analysis" by Agresti
  3. Read "Introduction to Statistical Learning" by Hastie & Tibshirani
  4. Read "Elements of Statistical Learning" by Hastie & Tibshirani
  5. Read "Pattern Recognition and Machine Learning" by Bishop
  6. Read "Machine Learning: a Probabilistic Perspective" by Murphy
  7. Read "Probabilistic Graphical Models: Principles and Techniques" by Koller

Outside of a massive Amazon bill you'll have, you'll learn the in-and-outs of the statistics, math, and algorithms behind a good deal of machine learning and data science techniques. The commenter further suggested doing all of the problems in the books as well as building your own versions of the algorithms. This learning roadmap sounds like you'd know enough to do PhD level work.

Yes, there are certain data science jobs and teams where this would be the basic knowledge expected of you and well within your grasp. So all of these things would certainly be considered as 100% important to your data science job skills.

Unicorns do not exist

This is why "data scientists" are sometimes called "Unicorns". Because there are vanishingly few people who have all of the "depth" required for the "depth jobs" as well as all of the "breadth" required for the "breadth jobs". So - unicorns do not exist. Which is good and bad.

Bad because companies do in fact need both types of data scientists. And often spend too much time and energy trying to find the right person rather than hiring two people - one "breadth" and one "depth"

Good because when companies realize they need teams rather than 1 individual unicorn, they to hire more people and hire the right people with the right skills and right background and right expectations.

Back to the question - where do you go next?

The power to choose is all yours. Look at the two lists above and decide which one sounds more appealing to you. If algorithms, statistical models, and machine learning get your excited, then the "depth" option list is the way to go. If you get excited at the possibilities of the "breadth" option list, then that is the way to go.

The great thing is that there is no wrong answer and that there is no wrong answer when trying to find a data science job. All you have to do is own it and be comfortable with your choice. Companies are currently hiring tons of data scientists of each type - so the sooner you choose and decide, the easier your job search will be.

To that end, take a look at the lists above once again and really think hard about what sounds more appealing to you. This will be the answer you are looking for in regards to where / what you should do next on your journey to growing as a data scientist. And lastly, don't forget that you can change your mind. If after a while, you want to pick up some of the skills from the other list - go ahead, you may in fact become one of those mystical unicorns!

Good luck!

Receive the Data Science Weekly Newsletter every Thursday

Easy to unsubscribe at any time. Your e-mail address is safe.