Greece’s tax authorities are turning to sophisticated data analysis to uncover hidden corporate networks, using a mathematical algorithm to identify connections among hundreds of businesses suspected of avoiding taxes and social-security payments.
The approach helped investigators uncover a network in Thessaloniki, Greece’s second-largest city, involving 380 companies and 205 individuals linked to €43 million in outstanding public debts.
At the center of the investigation is the Jaccard similarity index, a mathematical tool that measures how closely two sets of data overlap. Greece’s Independent Authority for Public Revenue, known as AADE, uses the algorithm to identify businesses that may appear unrelated on paper but share owners, directors, addresses, accountants, bank accounts or commercial relationships.
The system is particularly useful for detecting suspected shell-company arrangements in which businesses close while owing taxes, only to reappear under different names or tax-identification numbers. Rather than examining companies individually, investigators analyze entire networks of relationships, combining corporate-registration records, electronic invoices, value-added-tax filings and banking information.
In the Thessaloniki investigation, the algorithm identified 719 connections among 369 tax-identification numbers when applying a similarity threshold above 5%. These relationships formed 29 groups, the largest containing 262 tax numbers. Raising the threshold above 10% narrowed the results to 504 connections across 330 tax numbers, while a threshold above 20% identified 358 connections involving 270 tax numbers.
The process allows investigators to isolate smaller clusters of businesses with particularly strong similarities. A newly established company serving the same customers and suppliers as a recently closed business, for example, could trigger closer examination, particularly if the two also share directors or contact details.
The algorithm doesn't establish wrongdoing by itself. Instead, it highlights relationships that investigators can examine for evidence of tax evasion or fraudulent activity.
































