We continue answering questions for the nontraditional data science job candidate who recently asked us about how to become a data scientist when their educational background was in marketing and social media.
In the previous article, we gave take-a-step-back advice. That is, we started with answering the more holistic question about which hiring path to take as a nontraditional candidate. Then we covered how to think about education versus knowledge in the data science job search. In this article, we’ll cover what specifically the candidate will have to learn.
One specific question the candidate had was There’s an answer on Quora for the question “How do I become a data scientist?” that lists a ton of math classes and programming languages and techniques that I need to learn. Do I need to learn it all before I start looking for a data science job?
The answer is an emphatic No. No, you do not need to go out and learn everything that is directly or tangentially related to Data Science, Computer Science, Machine Learning, Statistics, or otherwise. What you have to do is go out and learn specifically what the job you want requires you to know. And the best way to do this is to do a project that would be directly related to the job you want.
That said, here’s the short version of the advice I gave the candidate:
- Ignore the Quora answer
- Figure out what companies you want to join that are doing data science
- Figure out what projects they have done
- Try to replicate one of their projects
Now, let’s take a closer look at those four steps.
Step 1: Ignore the Quora answer
This answer, while it looks good, is not very useful for someone who is coming at it from not being in school. Especially if it is someone coming at it from a Masters in Marketing degree. Yes, of course, you have to eventually learn all of those things. That's just the ticket to get in the door. BUT, right now, you don't need to know everything. RIGHT NOW, you have to figure out a way to take 1 step forward. Then maybe tomorrow, you'll take another step. Then maybe in a few months, you'll come back and tell us about all the cool awesome work you are doing. So for now, ignore Quora.
Step 2: Figure out what companies you want to join that are doing data science
Everybody and their uncle is hiring data scientists. This means that you can go work in the agriculture, social networking, finance, video, audio, text, media, fashion, etc. Every single company out there has data and needs someone with "data chops" to do something valuable or insightful with that data. So go figure out who you want to join forces with. It might even be a company that you're already familiar with and/or are working at.
Step 3: Figure out what projects they have done
All companies have data. Take a few minutes to think about the data they could potentially have - sales, web, supply and demand for their products, sales teams, etc.. From there you can start thinking of "hand-wavy" ways you could use data science. For instance, given their sales data can you predict next quarter? Given their web data can you build a recommendations engine? Given their supply and demand can you build a classifier for their best/worst customers and can you think of how to predict it sooner than later? etc?
Step 4: Try to replicate one of their projects
This will force you to go out and a) learn what tools are available to use, b) learn what state the data is in (how raw is it and how much "wrangling does it need"), c) what math you need to use.
So the answer we gave was that “No, you do not have to learn all of the math, computer science, statistic, etc that the Quora answer said you needed to be a data scientist. Forget about math. Find a project and start working on that. Everything you need to know will fall into place as you work on it.” This makes it much easier than buying a ton of math and computer science textbooks in the hope that you’ll learn it all by the time you have a data science interview.
Similarly, if you are a candidate with a nontraditional background, who is worried about what math, computer science, and statistics you need to know, as a first step, we suggest the following:
- Figure out what company you’d want to work for
- Figure out what projects they have done / would do given their data and goals (usually either increase profit, cut costs, or both)
- Start working on the project and as you go along, it’ll become apparent what math, computer science, and statistics you need to do the project.