Leave a comment
Get the GH Bookmarklet

Ask GH

For a relatively new field, there does not seem to be a specific path to becoming a Data Scientist. How does one get there?

  • JG

    Jim Gray

    about 6 years ago #

    In my case...I spent over fifteen years obsessively playing with data, doing computational astronomy research, doing computer vision, machine learning, studying HCI and UX, while absorbing everything I could about business and marketing by studying it & working with existing experts. At which point, it started being the trendy term for someone with domain knowledge in business & marketing, who tries to improve results via scientific research & computational analysis.

    Most DS I've met got a PhD in some computational science field, then took their career in the direction of applying the same skill set to business problems. I run into a lot of astronomy PhDs, but that's likely observation bias on my part.

  • AS

    Amit Sonawane

    about 6 years ago #

    I wouldn't say that this book will make you a data scientist, but it does a great job of teaching how to convert information into insight. John is the Chief Data Scientist at Mailchimp so naturally a lot to learn for marketer and growth hackers as well!


  • VV

    Visakan Veerasamy

    about 6 years ago #

    My favorite joke answer– "Become a statistician, then move to San Francisco's Bay Area."

    But more seriously, "Data Scientist" is this big scary phrase that really means– a person who is very comfortable looking at data, manipulating data, running tests to collect more data, and figuring out what the implications of those are.

    As with regular scientists, you don't actually need somebody else to give you a label in order for you to start practicing it. And it's not nearly a new field so much as it is a new label or a new approach to looking at a field. I'm a writer before anything else, but I've been dabbling in "data science" just keeping track of the number of words I've written, when I publish what I publish, what blogposts do well, which don't, what the lengths are, etc.

    I would say... you want to get as much skin in the game as quickly as possible. So first start with whatever data you already have, then look up Stack Overflow or Quora or other places with specific questions to figure out how to get data from your information. Once you've turned some data into information, congrats! You've done some (rudimentary) data science. Then do it again.

    If you work in an organization, see what you can do with the data that you already have access to. If you can come up with anything interesting to show for it, you can then ask for access to more data.


    Start small, do tiny acts of data science (look for patterns, avoid correlation=causation, test hypotheses, prove things, look cool, get all the ladies/gents, earn the legitimacy you'll need to do more data science)

    Shopify's CMO @craigmillr actually has some interesting things to say about doing marketing from an engineering perspective:

    "“My experience has been doing it in a more incremental fashion. A lot of this didn’t happen overnight. When we started the website itself, Shopify.com, it was considered the design team’s domain. For me to ask for anything was, well you know, they’ll consider it…

    You ask for a few small things and then you show that you’re adding value. ‘Why don’t we just change the title tags and some content on the page’– they were fine on that, and then they see that you make a big impact because you get more traffic to the site– so you you show them Google analytics, here’s before and here’s after.

    And then they start to gain a bit of trust in you, and they ask you, so, what else can we do? A/B test the copy, great results.

    Same thing with product– there was a clear demarcation inside Shopify. We kind of pushed them and just said, “Why don’t we just start testing some different things in the onboarding– the first thing you see when you signup?” We tried that out, had some good success there, and then tried more things.

    I think you have to prove yourself over time– but I think it’s very easy to do if you have a lot of data. Because that will show the way.” –https://growthhackers.com/shopifys-10x-growth-in-3-years-takeaways-and-transcripts-from-interview-w-shopify-cmo/

  • KR

    Kamil Rextin

    about 6 years ago #

    Natural science (physics comes to mind) or mathematics background with computational skills like developing simulations would help IMO. Languages like R are super helpful to learn as well.

    • FA

      Faisal Al-Khalidi

      about 6 years ago #

      Thanks @kamilrextin. It seems a strong quantitive background with some technical skills are important.

      • KR

        Kamil Rextin

        about 6 years ago #

        Yep, but it also depends on what you want to do or figure out. In most cases if your looking at historical data for trends or using tools like Mixpanel/ GA etc then you definitely don't need a PhD in statistics to make some descisions about what is working and what is not working. However if the work required is to build a predective model for demand for a Uber or Sprig or ... another of the on demand/delivery logistics company then it gets a little complicated there :)

  • BM

    Barry Mueller

    about 6 years ago #

    Check out reskill.me

  • RS

    Rob Sobers

    about 6 years ago #

    I'm curious: what's the outcome you're looking to achieve by becoming a data scientist?

    • SE

      Sean Ellis

      about 6 years ago #

      Would be great if he could clarify the intention of his question... I read this discussion with interest just to try to understand the difference between a data analyst and a data scientist. I don't want to become either but definitely see myself hiring some of each over the next few years.

    • FA

      Faisal Al-Khalidi

      about 6 years ago #

      @rsobers @sean I asked this question because I'm interested to learn more about using data to drive growth decisions. My lack of knowledge on the subject led me to thinks data scientist/analyst are the same thing! But after this discussion it seems they're not. So maybe a follow-up question is needed later :)

  • AD

    Aditya Dugar

    about 6 years ago #


    This thread on Quora answers almost everything you would want to know.

  • AS

    Ankit Shukla

    about 6 years ago #

    May be this could be a helpful visualization
    Just found on quora.

  • SJ

    Sarah Jukes

    about 6 years ago #

    Great question and thread. I've wondered this too over the years.

    Without re-training or investing in a PhD program, I could never call myself a data scientist professionally. But I understand how hot the appeal of the title is right now and the increasing value of skill sets involving experience with data collection, storage, manipulation and visualization.

    Where I've lacked a strong quant need in my regular day job, I've looked for skills and experience outside of my day job via side projects and independent learning.

    For example, Coursera is great (there's a bunch of short-term data science courses starting today) and books like the one below. The other thing I've done is set up a recent collaboration with a data scientist professor to run a small but useful data mining, analysis and programming project. It's win win! I get to learn and draw on his expertise plus do something fun and interesting outside of my day job.


  • GB

    Gennady Barsky

    about 6 years ago #

    Would you say being an expert in SEO is the same as being a 'data scientist' since both professions require rigorous data analysis?

  • GW

    Grant Wilson

    over 1 year ago #

    NYU Center for Data Science was the best choice for me. A couple of years ago, I didn't even think about getting a PhD. Now, I'm proud of it.