I find your comment rather funny, but you are forgetting some very relevant skills a DS is expected to have like NoSQL knowledge, MapReduce (Hadoop), some decent language to interact with... I don't think it's that easy to find somebody with these skills.
Right, but BAs have been doing SPSS, SAS, Mathematica, MATLAB, R etc, etc for decades. Any BA worth his or her salt could pick up these new things without breaking a sweat - they are not rocket science, and the value of a BA is in their knowledge of statistics and industry domain knowledge.
A DS is a BA without sufficient industry experience to know there is already a job doing what they do...
I don't really agree with this. A "Data Scientist" (as the term is "properly" used [1]) generally needs to be a very good computer scientist in addition to understanding the data.
Ordinary data analysis generally lives in a world where the computer is fast enough, and you need domain knowledge + basic stats. The term "Data Scientist", when properly used, deals with cases where domain knowledge + basic stats isn't sufficient, and computer science knowledge is also required.
[1] Many people who advertise for a "data scientist" really just need a BA or a DBA. I recently turned down a job as a "data scientist" after telling them that Hadoop is massive overkill. `SELECT SUM(x) FROM table WHERE condition` took about 4 seconds to run. Admittedly, the ruby script they were using to do it row by row was a lot slower, particularly since they were loading the entire row rather than just field `x`.
Different people mean different things by the BA/DS distinction. But I've seen a team of 15 people with SAS experience try to get into machine learning applications, and they are making less progress than I would expect from a single good ML practitioner (with R or Python experience).
This is just an anecdote... And Perhaps some people make the BA/DS distinction so this isn't the relevant comparison. But the returns to smart people and sharp tools is astounding.
Geez, we seem to have this conversation every week.
If companies aren't being silly, business analysts != data scientist. You still need to be comfortable with various statistical distributions, learning algorithms, graphical models, genetic algos, and possibly some NLP.
Best example I can use: at Knewton I work with two data scientists who can jump from belief propagation or genetic algorithms to infrastructural discussions at the drop of a hat. I don't think business analysts can do that.
yummyfajitas gets it right below: most companies need an analyst, not a DS. That doesn't mean there's no need for a DS.
10 comments
[ 3.8 ms ] story [ 23.2 ms ] threadBTW thanks Airbnb for holding these talks!
A DS is a BA without sufficient industry experience to know there is already a job doing what they do...
Ordinary data analysis generally lives in a world where the computer is fast enough, and you need domain knowledge + basic stats. The term "Data Scientist", when properly used, deals with cases where domain knowledge + basic stats isn't sufficient, and computer science knowledge is also required.
[1] Many people who advertise for a "data scientist" really just need a BA or a DBA. I recently turned down a job as a "data scientist" after telling them that Hadoop is massive overkill. `SELECT SUM(x) FROM table WHERE condition` took about 4 seconds to run. Admittedly, the ruby script they were using to do it row by row was a lot slower, particularly since they were loading the entire row rather than just field `x`.
This is just an anecdote... And Perhaps some people make the BA/DS distinction so this isn't the relevant comparison. But the returns to smart people and sharp tools is astounding.
If companies aren't being silly, business analysts != data scientist. You still need to be comfortable with various statistical distributions, learning algorithms, graphical models, genetic algos, and possibly some NLP.
Best example I can use: at Knewton I work with two data scientists who can jump from belief propagation or genetic algorithms to infrastructural discussions at the drop of a hat. I don't think business analysts can do that.
yummyfajitas gets it right below: most companies need an analyst, not a DS. That doesn't mean there's no need for a DS.