Arthur Von Eschen

Hello Arthur, you are Senior Director, Game Analytics at Activision. Please, tell us about the relevance of predictive analytics for your work.

Predictive analytics is relatively new to the video game industry. But it has quickly emerged as one of the main resources for ensuring game quality, understanding player behavior and enhancing the quality of the player experience.

According to the Harvard Business Review, being a data scientist is the sexiest job of the 21st century. What makes you or your profession so sexy?

Complex systems (e.g. human behavior) are difficult to understand. Companies often make decisions that impact these systems using intuition, personal opinion and data points which are very misleading. Data scientists are able to understand more about these systems, to a greater depth, and with greater precision than any other profession. We expose how these complex systems work, we predict how they will change and we provide insights on how best to influence them. This new level of understanding that we provide is what makes our profession so sexy.

Moreover, you are a speaker at the upcoming Predictive Analytics World on 4 & 5th of November in Berlin. The title of your session is “Cheating Detection in Call of Duty“. What makes this topic so important and relevant?

Predictive analytics has created tremendous value for industries such as banking, finance, insurance and retail. But predictive analytics is moving into new industries. In each of these new industries we as analytic practitioners must prove that analytics can provide enough value to make sure that businesses continue, if not grow, investments in our discipline. Cheating Detection in Call of Duty is one of the first examples of a large scale analytic service for the video game industry that is able to improve product quality and drive a strong ROI.

What can the participants of your session expect from your presentation – and what do you expect from the attendees?

An industry that is new to analytics poses a number of challenges to a data scientist. It impacts how you sell your ideas, how you build a case for ROI / value and even what analytic techniques you leverage. Attendees can expect to learn how I went from no-one wanting / understanding analytics to fully embracing the build out of a large scale analytic service, what techniques we used to build it, what challenges we ran into throughout the process and the business impact we made. I fully expect that after attendees watch my session they will want to play Call of Duty so much they will rush out and purchase Call of Duty Advanced Warfare ☺.

Big Data is still a hot topic in 2014. What do you think: is it pure hype or is it a substantial change for predictive analytics?

I think it’s both.

Analytics has driven substantial benefit in many different situations in many different industries. This of course generates hype. It has lead more companies and more industries to start investing in analytics, but much of this investment is poorly done. Companies often invest beyond their ability to execute, they don’t hold analytics accountable for measurable ROI or they lack the maturity and resources to execute properly. Some vendors are taking advantage of the hype and far too often sell services or solutions that don’t provide measurable improvements to the business or decision making.

However, many of these organizations and many of these industries will find benefit from their analytics investments. It is these situations that will drive the long term substantial change for predictive analytics. The hype that’s being generated now will open the door for analytics to more industries, to a wider range of company sizes / business models and to more problem domains.

Please make a prediction for predictive analytics: what comes beyond big data and what trends drive predictive analytics in the upcoming years?

The biggest trend that we will experience is that analytics will become ubiquitous. I believe analytics will follow a similar path to the computer. Computers were initially used to address specific scientific and industry needs, then moved into mainstream business, then made its way into the household, then into many devices (cars, toasters, stop lights, etc) and finally into personal items (the smart phone). It’s almost impossible to get through a typical day without using or interacting with dozens, if not hundreds, of computers. In the future, analytics will be imbedded almost everywhere that a computer is today.

I believe that it is this overarching trend of ubiquity that will drive how analytics evolves, what problems we try to solve and what those next big problems are. I think one of the first sub-trends we will see is the trend of small data. Today in analytics we are focused on building the super computers. Some companies are working on the personal computer equivalents. But analytics needs to get smaller, faster and be able to be just as effective with smaller and smaller sets of data. We will need to create the equivalent of laptops and finally embedded computers.

And one more personal question: what is your favourite prediction tool or algorithm? And why?

It’s very difficult to pick just one prediction tool as my favorite, but I can settle on two: R and KNIME.

KNIME provides a rapid analytic prototyping environment and a framework to build analytic applications that can include models and processes from other tools. It’s a great tool for building out actual solutions, instead of just models.

R is an incredibly flexible analytic tool. I like R for a great number of reasons, but probably the biggest reason is that it’s incredibly easy to integrate with and to embed in software. This makes it really easy to use R to create standalone software and services.