For this week’s publication, we have received an already-made interview from the Belastingdienst to publish. This interview has been translated from Dutch to English.
The story of Markus, data scientist:
“After my bachelor in Physics and master in Econometrics, I started working at the Belastingdienst. Someone who took assessments for the Belastingdienst that I knew thought: “Hey, someone with such a profile can be useful in the Data Foundations and Analytics department (DF&A)”. She was very enthusiastic about the department, so here I am, already for a year or two, completely satisfied.
That is mainly because I can develop myself in all kind of directions. I am currently working as a data scientist at the ad hoc analysis department. This means of course that you deal with ad hoc questions, but the diversity within those questions is enormous and can come from all sorts of parties. They can come directly from within the Belastingdienst, but also from the AFM, the Ministry of Finance, the House of Representatives (Tweede Kamer) and the Freedom of Information Act (Wet openbaarheid van bestuur). If citizens have questions about the actions of the government, they can submit a Wob-request. If that request is of a quantitative nature, it may be possible that we will answer that question.
Data lead of a data foundation
Within answering those questions, I have a diversity of tasks. In the beginning, I was working on a data foundation, as we call it, for most of the time. This means working up the source data in such a way that you can make reliable analyses and robust products out of it. I am still working as a data lead for one such data foundation. This means that you are responsible and that you are the first point of contact. My foundation has to do with workflow control. I make sure the information from this data foundation reaches the right people and I answer questions if there are any.
Insights from analyses
In addition, I develop information dashboards and make chain- and process-analyses, for example for value added tax (VAT). If a company does not file a tax return, the Belastingdienst estimates the tax due. Analyses that I have made show that this process is not optimal for some target groups. With the products we develop at DF&A, we provide more insight into our customers and unnecessary and incorrect estimates can be prevented.
Choose your projects
My tasks differ considerably and that’s what makes it interesting. You can often choose which project you want to be involved in. These projects range from impact measurements (do our projects yield results?) to developing risk models (where should we audit more or less?). We use modern analysis techniques for this, such as gradient boosting and neural networks.
With these techniques, we deliver usable products for the ‘business’, as we call it. One example of these is a product called dynamic monitoring. If someone owes the Belastingdienst money, and several warnings have been given for too late payment, a seizure can be initiated. However, it could be that someone has nothing to seize. In the past, we would perform manual checks repeatedly with some interval to see if there was still nothing to seize. We’re doing that dynamically now. The system automatically notifies us when someone has a job or buys a car again. In that way, it automatically becomes clear when there is something to be collected, because everyone has to pay taxes for a fair and well-functioning society. This is also a nice aspect about working for the Belastingdienst.
Department under development
Especially the versatility of projects, the enormous development within our department and the many solutions that have not yet been discovered, are things that really make it a wonderful job at a nice place, because that is how we work within DF&A at the Belastingdienst. The atmosphere is great, colleagues get along very well and do a lot together. In addition, we have a state of the art analytics platform. You can run SQL queries on the Teradata server in Apeldoorn, which gives you an answer in no time, while in the meantime 60 billion records have been consulted. Bizarre, right?
And you don't do that for nothing, because the impact of the things we develop here is big. Clever things that we develop in SAS yield savings or extra collections of many millions. And of course, we do all this with a PIA: a privacy impact assessment, because we have to work responsibly every day. Which data can we use for which purpose and how do we minimize impact on people’s privacy? For example, we are working on pseudonymising data: the end-user can only see what he needs to see and no more.
As you can see I am busy with many different activities and my curiosity is not fulfilled yet, so I will stay here for a while.”