This is not about attacking a guy - a friend of mine - who, at first glance, seems extremely overpaid, like any top executive. Indeed, the question is about whether data scientists should be coders (spending 50% to 100% of their time writing code) or not.
I believe the answer is negative. There are many different types of data scientists, and no real data scientist, in my opinion, spends more than 50% of his/her time coding. But you are welcome to post your point of view. Data scientists spend much of their time producing measurable added value, and many times, it involves intuition, vision, and gut feelings, and stuff that you don't learn at school.
Some interesting comments that I've read include
The latter is about the top qualities any data scientist should have: creativity, and the disruptive element to think about products that no one before would ever imagine it would work. And also the fact that data science is a blend of several disciplines.
Here is an example of a data scientist, still writing Perl programs and relying on vision, yet creating bridges (API's) between platforms, to automate growth hacking, lead generation, and everything that comes with it, to the point that it is described as IoT (Internet of Things) for digital media. Almost without coding. Some top automated trading algorithms are just like that, being mostly machine-to-machine communications, involving very little coding.
Data science is not just about coding, and my friend makes money from algorithms that deploy machine-to-machine communications, with very little coding involved: instead, it's about high-level API's and web apps, many times leveraging vendor platforms where much of the code resides. Many top data scientists actually do not code at all: they either manage a startup, or supervise coders. Those who spend their days coding are not real data scientists.
As an hiring manager, if you interview candidates, be aware that the data scientist job title has been abused, and do your due diligence to identify candidates that will make your client happy. Today, someone who can barely write an R program call herself data scientist and demands a $100k salary just out of her training. I think the data scientist job title should not be legalized like doctor or lawyer, but when hiring a so-called data scientist, ask for success stories, coding samples, and references. My 2 cents.
DSC Resources
Additional Reading
Follow us on Twitter: @DataScienceCtrl | @AnalyticBridge
Comment
I only take issue with one or two things:
"A good data scientist is a bad coder"
There is no excuse for being a bad coder. Incapable? That's fine. But a bad coder is a lazy coder - poor style, useless comments, etc..
In as much as I support the ability of data scientist to know how to code very well, I have seen many that are doing excellent job with software such as SPSS, Rapidminer and Weka which do not require any element of coding except understanding the techniques and drag and drop procedure.Hire manager should look for the person that can easily get the job done with good references . Nevertheless we still have some "cowboy data scientists" capitalising on the lucrative nature of the career to exploit employers, hence I advocate establishment of body that defines the requirement, qualification and certification for the profession
There are many quants or traders in the financial domain , who work with their own tools (no R, no Python, no SAS) who could easily call themselves data scientists. All they need to do is develop 1 hedging strategy which yield multiples of their 500K. Some really know their domain inside out, can quickly develop their own ideas, and implement them quickly.
I work as a data scientist (PhD) in the area of sales forecasting in FMCG, fraud detection, I was a teammate of the team which won the Kaggle competition (our #1st place solution here). I cannot imagine how I can do my work as a scientist only without programming skills in Python and R. If I see some predictive analytics problems, I understand what approaches can be applied to solve it. But to get good solutions I need to analyze distributions, generate a lot of new features, select most effective of them, do a lot of validations of models, build ensembles of models, etc. I can do all of these things only using my programming skills. Before I get the final effective model which can be implemented into production, I need to create a lot of prototypes for the analysis and if I ask some engineers to do programming work, in my opinion, it will be a very long process.
This is very inspiring. Thank you!
Why are folks so hung up about what what a data scientist should be or shouldn't be? As far as I'm concerned, get the individual that will get the job done, and call him whatever you want to call him/her with a salary to match the contributions/expectations.
Also, designing a brand new system that leverages various data to deliver high value and beat all competitors, is creativity and craftsmanship: you don't learn that stuff at school. Coming up with the concept might just take 30 minutes while sipping a glass of wine at midnight, but is has far more value than days of coding - coding can be learned at school or on Github, stackexchange, Google etc. And what you can learn for free on the web will one day get outsourced to the cheapest bidder that delivers good enough quality. That's why the $500k, in this case, is by no means overpaid.
If your friend knows his company's business, competitors, and market well, and knows what data to gather, what analyses to run, and what models to build so that he can discover new knowledge, provide strategic advantage, and communicate these results to company executives in a clear and understandable way, then I would say he is under paid.
I have seen many amazing programmers working on the wrong problems. Being able to code is one thing. Getting value out of the techniques that Data Scientists wield is another.
© 2021 TechTarget, Inc.
Powered by
Badges | Report an Issue | Privacy Policy | Terms of Service
Most Popular Content on DSC
To not miss this type of content in the future, subscribe to our newsletter.
Other popular resources
Archives: 2008-2014 | 2015-2016 | 2017-2019 | Book 1 | Book 2 | More
Most popular articles
You need to be a member of Data Science Central to add comments!
Join Data Science Central