If you ask 100 data scientists what data science is, you are very likely to get 100 different answers. Most of those answers will have some points in common. The use of algorithms to create machine learning models is one of those points, I would expect the most common of all. Big data is another. Most definitions focus on the skills needed and the technologies used. We find these definitions unsatisfying at Miniclip. We currently hold two different views that have evolved over time and that are not bound to a specific definition of what data science is.
The first view is the oldest and most pragmatic. It is interested in the output of data science. Every application that allows automation and/or interaction with data in non-traditional ways is a data product. Data products are the output of the Data Science Team. This view is very powerful since it opens our scope. There are many problems that most data scientists wouldn’t even consider because it is not in the narrow scope that they defined given a set of skills and technology. By removing skills and technology from the definition and focusing in the output we are able to pick up projects that have a large impact and often a very low cost. Our software engineers have access to visualizations that allow them to find quality issues in our games during development. Our quality engineers are able to automatically check the quality of the events our games generate. Our product owners have access to automated specialized reports. All of these are data science products.
This view impacts not only the scope of the Data Science Team but all the BI department since everyone is able to participate in the data science efforts and innovation. Many of the products we use and most of the automation the Data Engineering Team has created has this view in mind. The by-product of this is growth of all elements of the BI department. It is common to have data engineers experimenting with machine learning and even more common to have data analysts building automated systems and reports. We are all able to learn more and do more.
The second view is more recent and more philosophical. It is the view that business users without technical know-how can (and often do) have a vision that can be made reality by data science. The projects I mentioned earlier are a good example of this. My favorite one and to this day also my favorite data science project at Miniclip started with Nick, our CMO. He had a conversation with me where he said, “there’s a moment of truth in the lifecycle of our users where they’ll tend to make a purchase or to leave the game.” This sentence was the start of a family of machine learning data products that we collectively call The Moment of Truth.
This view opens up data science to the whole organization. It allows us to consider “what ifs” and discuss their use. It allows us to prototype systems and models and infers their future potential. It allows everyone at Miniclip to be part of the Data Science Team and contribute to the innovation we want it to create.
Essay By: Ricardo Vladimiro, Data Science Lead at Miniclip