There is more to a successful data scientist than mere knowledge
Look at Data scientist "definitions" with a wry smile: the "essential" skills very much reflect those that a short time ago were quite novel, and are being used in applications to problems that have recently become solvable.
By Chris Barnes, Data scientist at University of Canberra, Australia.
I would like to disagree with the tenets of the article on 10 Signs Of A Bad Data Scientist.
I am one of those born with an aptitude for numbers, for whom maths was a daily escape from the grind of school homework. I have never worked in any Maths/Stats Department or group, since graduation: I have taken the route of the generalist quantitative problem solver who is aware of a wide range of techniques, but only learns new what is necessary to solve the (applied) problem at hand (e.g. Statistics). As a generalist, I have participated in (and led) numerous projects where the expertise provided morphs depending on the expertise of the other people involved - basically filling in for missing skill ls (e.g. emptying the rubbish).
Hence I view data scientist "definitions" with a wry smile:
the "essential" skills very much reflect those that a short time ago were quite novel, and are being used in applications to problems that have recently become solvable.
I believe that scientists of very different hues can share similar conceptual models and philosophies, but bring different sets of complementary party tricks (skills) to the project table. Some of my most enjoyable (productive) years were spent working closely with Bob B., a creative scientist who began as a computer programmer in 1963 (still active in genomics), and neatly complemented my lack of programming expertise. Bob would not necessarily know the latest solution to a particular problem today; but because he had lived through the development of computer science, tomorrow he would give a creative solution that integrated current techniques with others that were currently out of favor, but improved upon what now passed as the accepted solution. For example, today an experienced data scientist might spot that, where speed was of the
essence, Hadoop etc. is a quite inefficient solution for some problems and avoid it.
Because data preparation etc. required 90-95% of our time, Bob's optimal contribution was in automating this step, while I was the one that presented the polished analysis to the client. When the usual budget panic arrived one year, the somewhat ignorant bean-counter manager decided that Bob was not very productive in terms of final results, and that they could economise by getting rid of him. Subsequently, when our (my) output had dropped to less than 25% of our previous total (Bob was about 5x faster than me at programming), I also was "let go"! The method used was strictly illegal, but I was not stupid enough to fight it.
There is rather more to being a successful data scientist than mere knowledge!
Bio: Chris Barnes is a senior data scientist with extensive national and international experience in the areas of business, sport, environment, maths and engineering in the public and private sector.
Editor: this post was originally made by Chris Barnes as a comment on 10 Signs Of A Bad Data Scientist, but it raised important ideas that deserve a separate post.