It seems like more and more organizations want to be involved with data science practices nowadays. A development I do not only encourage but also understand as I look around in the rapidly changing world we live in. Markets are becoming more transparent, which enables consumers to easily compare different offers and to choose the most attractive supplier. Getting the most value out of your data is essential to serve them according to their individual needs and in this way gain (or maintain) a competitive edge.
So we believe in the possibilities of data science more and more. However, many organizations struggle with successfully implementing it in their business processes. I do not think this is strange, since there is a natural gap between business needs and data science possibilities. To illustrate this, I would like to pose two fictive (but realistic) quotes of two essential stakeholders in a typical data science project: a data scientist and the CEO (or CFO, CMO, etc.), both working at a multinational organization.
“I want to develop and build the most epic algorithm that can automatically calculate the optimal solution of a mathematical model, that includes all kinds of different data sources.”
“This year’s goal is to increase our profits with 10% compared to last year.”
Although these two desires can fit into each other nicely, in practice they often result in a misunderstanding. The mindset of most data scientists is theoretical, thinking in the optimal solution. Let’s say the data scientist has found a way to implement an algorithm that calculates the optimal price of every product (optimal prices means that the organization generates, for example, maximal revenue or profit with these specific price settings). From a purely theoretical point of view it would be unwise to not implement this algorithm immediately since it does not get any better than optimal.
However, a typical C-level thinks more pragmatic and is often skeptical about such an optimal solution, because it takes a lot of time and effort to implement it. A manager first needs to see and feel what such a solution means for the organization in order to be convinced for making such an important investment. But, translating the optimal solution into an added value story is not necessarily included in the skillset of a pure data scientist (if you are interested in the skillset of a typical data scientist, it is certainly worth reading the earlier released data scientist counterpart blog of this one).
The solution to this problem is not to change the data scientist’s mindset, or to learn the CEO more about the possibilities of data science (many C-level manage better things to do than to take an online data science course). The solution is to add another essential player to the team. Someone who understands quickly how an organization, its market and its customers work operationally and tactically, and knows what are the important performance indicators and what influences them. On the other hand he or she has to be up-to-date with the (im)possibilities of the latest data science techniques and has to know which available internal and external data sources are relevant for a specific business problem.
Such a player can translate business needs into a clear data science vision and a realistic roadmap towards it. A roadmap that keeps the optimal solution in mind, but also guarantees that the organization also profits from the steps towards it. Along the way, new unforeseen practicalities will arise and the roadmap can be adjusted in an agile way. Which eventually makes the optimal solution more robust and fits the specific business needs much better. Such a player we call a data translator.
So besides business understanding and knowledge of data science (im)possibilities, what other skills does such a data translator need?
▪ Empathy: to easily connect with people and understand what is the question behind the question.
▪ Creativity: to connect the many dots of techniques and data sources to fit the specific business need.
▪ Visualization: to translate a complex multi-dimensional situation into a visualization that is easy to interpret for a human brain such that patterns are easily derived.
▪ Storytelling: to translate a complex optimal solution in a clear vision and realistic roadmap that is understandable for people with no data science background.
So we need a translator to effectively use data science in a business environment. Although the data translator has a crucial role to play in a data science team, I do not think that he/she should necessarily have a vast mathematical background. Of course, a data translator should have affinity with data science possibilities and potential threats. But this can be learned through experience. What makes a data translator so valuable is the ability to combine this affinity with the (in my opinion) more important soft skills of his/her profile.
Most mathematics/computer science/statistics graduates with a data science interest find their way into a nice data scientist or data engineer position. In general, data scientists do not aspire a position where talking (but more importantly listening) to board members is a significant part of their job.
However, when data scientists get more experienced and get more feeling with many different business processes, they typically learn to see things from different perspectives. Then they understand that getting added value out of data science is more than finding an optimal solution. Data science is not a goal itself if you want to successfully apply it in a business environment. It is a very powerful instrument that can help organizations to achieve their goals in a data driven era. But you need to close the gap first between business needs (C-level) and data science opportunities (data scientist). And the data translator is the man for the job.