Key Challenges in Analytics: Interview with Véronique Van Vlasselaer
Véronique Van Vlasselaer (Data & Decision Scientist; Analytics & AI Lead South West & East Europe at SAS) obtained her Ph.D in 2015 with the department of Information Management and Decision Sciences mentored by prof. dr. Bart Baesens. During her Ph.D she focused on the development and implementation of fraud detection solutions from a data science perspective. She researched how network analytics can augment traditional machine learning algorithms in the fight against fraud.
Since 2015, she is employed at SAS, world leader in software for advanced analytics and AI. Within SAS, she is part of the Global Technology Practice, a black-belt team that is not afraid to tackle any data science challenge, and she leads the Analytics & AI community of South, West and East Europe.
What was the topic of your Ph.D?
During my Ph.D, I researched how traditional data science techniques for fraud detection can be enriched and improved with information extracted from social networks. I had the pleasure to collaborate with the Belgian Social Security Institution (Smals) and Worldline, validating my work against real data and having the opportunity to get direct feedback from field fraud investigators. Social security fraud is defined as the intentional evasion of companies to pay their social security contributions. One of the main modus operandi in social security fraud is when companies go bankrupt and do not (have to) pay their social contributions. After some time, a new firm is founded with a similar structure to the bankrupt company, but with a clean page for their social contributions. The newly founded firm is just a re-birth of the bankrupt firm, continuing its activities as nothing has happened. The advantage of social network analytics is that these techniques are extremely powerful to expose such kind of fraudulent behavior. The hit rate of the detection algorithms that we have developed was very high.
What was the most important lesson learned of your Ph.D?
Back then, the use of social network analytics in data science was rather new. Our research group was at the forefront of how social network analytics can be applied, specifically to fraud. I am very proud that many of the fundamental principles on network analytics for fraud that are used today are based on the work we did. Our research resulted in many papers presented at international top conferences and published in high-quality journals. The paper ‘GOTCHA! Network-based Fraud Detection for Social Security Fraud’ was published in the journal Management Science, one of the most renowned academic journals worldwide. The findings of my Ph.D research opened my eyes on how powerful networks are in our daily lives and how we often underestimate the information that we can extract from it.
Although the insights on the power of social networks seemed to be my most important lesson when I just graduated, time and experience has taught me that there is much more value in my Ph.D than just the technical findings. Being exposed on how the industry applies data science and maturing in the practicalities of data science, I understand now more than ever that the most important lesson of my Ph.D is that true benefits and value from data science are only realized by carefully listening to your end users and focusing on communication and explanation of your findings. And that is exactly the reason why my Ph.D delivered such successful results. Unconsciously, I consulted the experience of fraud investigators and translated their gut feelings and expertise in technical requirements. It is because of their input that we succeeded in building such powerful algorithms. In the data science world, we often say “Let the data speak!” The more experience I have, the less I believe in this statement. The human factor in the loop is and will be always crucial in making good analytical models.
What did you do after your Ph.D? Was it a big change?
I joined SAS, world leader in software for Analytics & AI, as an analytical consultant. I worked on a variety of projects, having the opportunity to expand my skill set to other analytics domains like forecasting, operationalization, decisioning, etc. I have deliberately chosen to not only focus on fraud, as I would like to have a feeling what analytics can do for other domains. I also strongly believe in cross-pollination. Techniques and best practices that work well in one domain (e.g., fraud) might be very promising and useful for other domains (e.g., risk) as well. One big change that I immediately felt when I joined SAS is that people’s believe and trust in your story completely changes. I still remember my first conference I attended when I worked for SAS. It was after a couple of weeks I left university. I was invited as a presenter, but as I did not have any projects from SAS to share yet, I decided to present some insights from my Ph.D. It made me quickly realize that if you tell a story as an academic researcher, you are considered as neutral and innovative. Your mission is to make the world better. When I presented at that conference, I was no longer perceived as the neutral and innovative researcher, I was a representative from SAS, a software vendor. There must be a catch in my story – and of course there wasn’t. Same story, same presenter, but different association. Everything suddenly changed. It was a tough reality shock. It motivated me to re-think the story and bring even a stronger one. But it has to be said: Perception is everything. Bringing a believable story when you are standing in vendor’s shoes is an impressive achievement.
What do you like about your current job, and is your Ph.D useful for it?
I am currently part of the Global Technology Practice at SAS, a black-belt team that focusses on the most challenging analytical problems. I also lead our Analytics & AI community consisting of more than fifty Analytics & AI experts across Europe. My job is extremely varied, and that is what I like about it. SAS has many customers worldwide, operating in different sectors and domains, each with their own requirements and expectations. It inspires me to hear from our customers how analytics is used to bring value to their business. At the same time, I love to inspire customers about the potential and opportunities of analytics. The most energy I get from making my hands dirty and diving into a practical analytical problem. Thinking out of the box and creatively handling data and algorithms to achieve the end result is what drives me and makes me happy in my job. And that is exactly what I learned during my Ph.D: Nothing is impossible. Everything is feasible, you just have to think it through. Working with people with the same can-do attitude is what makes the job at SAS so wonderful.
What do you consider as key challenges in analytics?
First of all, the analytics world is flooded with so many hypes and buzz words. If you talk today about data mining, you are old school. When I started my Ph.D, hype terms like data science and machine learning surfaced. Although those terms still enjoy their popularity today, if you don’t talk about AI (artificial intelligence) you are out of the game. Many articles are published trying to define what each term means and how each term differs from another. I sometimes have the impression that the discussion about which term to use when receives more attention than getting value from analytics (or data science, or machine learning, or AI).
During the last years there is an important shift from developing accurate analytical models to the development of accurate and trustworthy models. Analytics and AI is getting more and more integrated in our daily lives. That has an enormous positive impact, but exposes us at the same time to some risks. Stories pop up where analytics and AI has been proven to be biased, unfairly treating minority groups or people with uncommon tastes or behavior. Addressing these issues and creating trust in analytics and AI will be a challenge that we need to urgently tackle.