5 Analytics Specialisms - How Data Science is Changing

It's been a few years now since Data Science has been the new kid in town.  It's an evolving discipline which was once staffed with the resident “data geek” yet now colleges are churning out fresh faced data analysts and data scientists by the truck load, all intent on changing the way businesses solve difficult problems.

The evolution of data science is being driven by technology, which is creating specialisms within the discipline.  Where once a simple database supported a data science project, there are now options for big data, small data, structured, unstructured, on-premise, in the cloud, SQL and NoSQL.  The trend is true for model building, visualisation and dash-boarding too! The “data geek” can’t keep up and is no longer capable of being proficient in all the possible tools.

If we think about the component parts of a typical data science project we can see 5 specialisms emerging:

The Problem Solver

The problem solving capability comes with experience but also requires subject matter expertise and an adeptness (or at minimum an appreciation) of all 5 specialisms in order to deliver a solid analytical approach to difficult business problems. Everything from predicting which customers churn to reduce the effort in collecting outstanding debt, the person tasked with providing an innovative approach to problems needs to be the analytics teams star performer.

The Data Wizard

Every data science project needs data.  It's the raw material; the ingredients! Most analysis involves taking data from numerous business systems, spreadsheets and often external data.  Oftentimes, data requires cleansing to rectify known errors or different sources may need to be merged. All of this is in an effort to build a data set ready for analysis.

The variety, veracity, volume and velocity (the 4 V’s) are all considered before deciding on the right tools for this data crunching. The data master must be equally comfortable with Databricks, Hadoop, Oracle, SQLServer and MongoDb.  Data Mastery is a key skill which can no longer be left to the “jack-of-all-trades”.

The Master Modeller

A recent article on RBloggers.com, pointed out that a Neural Net model performed 15% faster in Python than in R.  In days of old (!), it was perfectly reasonable for programmers to know a single programming language. In analytics, you’ll need at least SQL, R and Python to even be considered for a modeling position.  Add to this, that a good working knowledge of statistical modeling is also a prerequisite and you begin to understand why such a specialism is required.

Visualisation Expert

We can all create a variety of graphs in Excel.  We can even paste them into PowerPoint. That said, we’ve also been to conferences where the presenter seemed to have absolutely killer visualisations, charts, annotations and infographics.  We’ve all seen incredible interactive dashboards which have the ability to engage users and impart information in a way we could only dream of creating ourselves.

I’ve seen so many CVs claiming that it's author has visualisation skills and in each case the author has been deluded.  The ability of a visualisation specialist will help analytics teams to convey complex and difficult results to audiences in a way that is memorable.

Storytelling Supremo

After potentially months of analytics work, we’re typically given only 30 minutes to stand in front of a senior management team and bring them on the analytics journey!  Problem Statements, Approach, Results, Impact, Next Steps etc. It's not for the meek and requires a good degree of skill to craft a story that resonates. It's also a role for a salesman who is capable to talking the language of senior management whilst also fielding questions about the statistical significance of results or the appropriateness of a modelling technique (there’s always one SMT member who read something about linear regression once!)

Time to Change

In any mature business with an analytics team, it's no longer appropriate for a single “data geek” to deliver all of these specialisms. Even many large consultancy companies offer single-flavour data scientists without first understanding the specialisms required.

Recruitment companies too, advertise for data scientists and unreasonably expect candidates to be masters at each specialism.  Data Science is changing, lets see who keeps pace with the evolving model.