The evolving scope and nature of data science
Historically, the evolution of science shows us that specialist fields emerge from more general areas as knowledge and expertise grows. Leonardo da Vinci would probably not recognise what he knew as ‘science’ among our multitude of expert scientific fields. Data science is no exception to this rule. It has emerged from, and sits within and among, many other fields. Its main parents are statistical mathematics and computer science, but there is also considerable crossover with business, and close links with business analytics.
Digging into data science
However, even within data science, there are many different elements. We might start with the premise that data science is all about creating intelligence from data. To do this, we need to create models—but we also have to put them into production. This process contains several steps, from finding and organising the data through modelling, testing and deployment, via validation and exploration.
One way to think about Data Science is by breaking it down into about 15 separate areas or steps:
Each of these steps or stages requires different expertise, and few data scientists are able to do everything. For example, in an interview, SAS data scientist Robert Blanchard commented that he tends to spend most of his work time building models using artificial intelligence. He explained that knows how to operationalise them, but it is generally more efficient to leave that step to others who are better at it.
One interesting way of looking at each area, and the expertise required, is to consider whether it is a ‘business’ function or an ‘IT’ one. This is my take on that, with business shown in light green, and IT in blue. You can see that most of the modelling work is probably best positioned as an ‘IT’ function or as leaning heavy on IT competence, as is the technical matter of wrangling the data. However, finding the best data is a business function, because business teams are best placed to understand what questions they need to answer, and therefore what data will enable this.
Positioning data scientists
The next question is where do data scientists sit on the spectrum between IT and business? The short answer to that is ‘it depends’. There are many factors involved here, including both individual and organisational preferences. Some data scientists prefer to be on the ‘IT’ end of the spectrum. Like Robert, they like to focus on the model building and training. Others want to understand more about how their work contributes to the business. They have moved closer to the middle of the spectrum. My colleague Caroline Payne made an interesting point about specialism vs. generalism in the era of AI. She suggested that in a fast-moving world, the ‘ideal’ may be to have a broad general knowledge, but also deep specialist knowledge in a particular area. Data scientists might therefore have a broad knowledge of the business and statistics, but perhaps be experts in building models using artificial intelligence.
I think one of the most important factors in the positioning question is the organisation itself, and especially its size. In my experience, in larger organisations, data scientists tend to sit closer to the business end of the spectrum. They have more contact with business functions, and more insight into the business. I suspect this may be driven by company philosophy, and an aim to get ‘best value’ from investments in data science. After all, you don’t want to waste your valuable assets’ time doing anything that doesn’t directly relate to a business problem—and the best way to make sure of that is to ensure that your data scientists understand the business. This would make sense. Evidence is growing that data scientists need both hard and soft skills to survive in business. They have to be able to deliver beyond pure technical skills like coding and statistics. Communication skills, and especially the skills needed to enable them to explain the outputs and outcomes of analytics, are increasingly important.
A combination of skills
The value that a data scientist will provide is generally not quantifiable at the point of hiring. Instead, it is the sum of the value generated from the problems they solve—and that will be greater if they combine hard and soft skills, or generalist and specialist knowledge. However, it is also an individual issue. Each of us has our own skills and expertise, and we combine them in slightly different ways. The questioning of positioning may be company policy, but it is also a matter of individual capability.
