Q&A: How hybrid fraud detection cut losses by 98%

Special Tax Inspectorate has nearly eradicated carousel fraud in Belgium

Thieves steal an estimated 100 billion euros (US$132 billion) a year from governments across Europe in a scheme known as carousel fraud. This high-return, hard-to-prove crime involves multinational networks whose members conspire to exploit weaknesses in the value-added tax (VAT) system.

Here's how it typically works:

A conspirator in one country sells goods to an accomplice in a tax-free country - a business-to-business transaction. The accomplice sells the goods to a new buyer and collects the VAT, but does not give it to the government. The new buyer eventually sells the goods tax-free – usually to the original conspirator – and then asks for a refund on the VAT the government never received in the first place.

In Belgium though, carousel fraud is under control thanks to an advanced analytics technique called the hybrid detection model. This innovative method uses multiple analytical techniques to expose even the hardest to find fraudsters. With this approach, the Belgian government saves nearly 1 billion euros each year.

Tens of billions of euros are lost every year through fraud carousels .... Not implementing this solution in this difficult budgetary period would be almost criminal.

Yannic Hulot

Can you tell us how you uncover the fraud?

We're looking for the very essence of the fraud – for example, someone who has certain characteristics, is known for other acts, for whom we received information from abroad, has put into place a company, and engages in transactions that are not declared. So it takes a bundle of elements to identify. Even if these characterizations are well-defined, the difficulty remains in managing and verifying this information. So we are facing a huge amount of information.

What's the magnitude of VAT carousel fraud?

In Europe, the VAT gap is estimated at nearly 100 billion euros a year. One-third can be attributed to missing traders intracommunity fraud [a form of carousel fraud that claims tax refunds for an unidentified or missing trader somewhere in the export chain]. The major difficulty is in finding the fraudulent transactions quickly from among a huge number of transactions. One company in 10,000 participates in these fraudulent transactions, and there are hundreds of millions of transactions within the European Union annually.

The problem of VAT carousels is now under control in Belgium. How was that achieved?

Risk analysis using advanced analytics helps us quickly identify companies at risk so we can take action before significant damage has been done. With the hybrid detection model, we really have the tools that are in line with our needs. The results are so good that we can now restart some old projects that were originally evaluated with traditional techniques.

Social network analysis is used in the hybrid detection model. Is it an adequate tool for the detection of VAT carousels?

With the amount of data we have from intracommunity transactions, company data, unstructured data, etc., we have a cluster of data that is totally unreadable because of the large number of links. The challenge is to extract from this chaos the data that is relevant. This is exactly what we can do with social network analysis and other methods of the hybrid detection model. But key in all this is to also combine business knowledge during the analysis.

Is this approach applied to all companies or is it limited to areas at risk?

It's a holistic approach and not confined to one area. We focus on certain transaction profiles rather than particular sectors. And it's not limited to specific types of transactions. It's very important to have a comprehensive approach. If we focused on a specific area, the fraudster would simply change tactics.

What is the current situation for the rest of Europe?

The situation varies greatly from one country to another. It's not widely known, but European taxpayers lose billions of euros. Some member states, and in particular the large countries, have difficulty managing the massive amount of information. A solution such as the hybrid detection model is the answer. It lets you view the essential, prioritize cases and then allocate resources. Today, in the 28 EU member states, the data available for analysis is the same. This means that the Belgian experience is completely transferable.

So we can dream that one day there will be a common approach at the European level?

A common analysis is the ideal solution. But some member states consider it a loss of sovereignty to entrust the analysis to others. Centralized analysis could be entrusted to Eurofisc, which is already well-structured and has a lot of business knowledge. A common cross-border analysis would make it possible for governments to save enormous amounts of money. Tens of billions of euros are lost every year through fraud carousels .... Not implementing this solution in this difficult budgetary period would be almost criminal.


  • Combat loss of Value Added Tax (VAT) by criminals (estimated at €100 billion/year).
  • Detect, prevent hard-to-prove carousel fraud.
  • Stop crime rings from exploiting weaknesses in VAT system.
  • Prioritize investigations of organized fraud.



  • In 2002, losses in Belgium from VAT fraud amounted to €1.1 billion.
  • In 2012, the losses were a negligible €18.5 million – a reduction of 98%.
The results illustrated in this article are specific to the particular situations, business models, data input, and computing environments described herein. Each SAS customer’s experience is unique based on business and technical variables and all statements must be considered non-typical. Actual savings, results, and performance characteristics will vary depending on individual customer configurations and conditions. SAS does not guarantee or represent that every customer will achieve similar results. The only warranties for SAS products and services are those that are set forth in the express warranty statements in the written agreement for such products and services. Nothing herein should be construed as constituting an additional warranty. Customers have shared their successes with SAS as part of an agreed-upon contractual exchange or project success summarization following a successful implementation of SAS software. Brand and product names are trademarks of their respective companies.

Back to Top