After dismantling the complex network of 380 ghost companies that vanished leaving behind debts of 43 million euros to the state, the Independent Authority for Public Revenue (AADE) is not limiting itself to previous findings. Instead, it is expanding the application of the advanced business connection detection model it developed for this specific case, marking a new era of audits for company networks that change managers and tax registration numbers with great ease.
At the core of the new system lies a mathematical algorithm, the Jaccard similarity index, which measures the degree of similarity between different data sets. Through this method, patterns and pathways that are not apparent with conventional audits are identified, allowing tax authorities to map connections and movements that until now remained in the “dark”.
A decades-old wound thus appears to be finding a solution, with a permanent digital “wall” against a form of tax fraud that, although many knew and discussed for decades in the market, was difficult or almost impossible to detect and prove technocratically and systematically — until now at least.
AADE: The end of impunity
In the business world, for decades it was considered an open secret: many would set up companies, “load” them with debts, “burn” them and open others in their place. However, the store continued to operate normally, turnover flowed, transactions took place, customers noticed nothing. But in the tax office registries, the tax number had changed, debts remained uncovered and unpaid, those responsible disappeared, but the network continued to drain taxes and social security contributions insatiably for many more years, simply changing tax identity and registration number.
Until now, the exposure of such networks usually happened fragmentarily, coincidentally or after a complaint. Even when they were detected, it was usually too late: the money had been lost to the public sector, the “straw man” did not exist or knew nothing, while another was already “running” the network’s next company with a new tax number. In practice, a network could never be fully exposed, as it managed to “pull” enough of its tentacles, even when it lost some.
The reason was twofold: an auditor entering a business finds it difficult to grasp the complex scheme that has been set up and operates in these cases: different companies, with different tax numbers and different “responsible parties” and representatives, successively “produce” debts of millions to the state, without one appearing as a continuation of another.
The algorithm that “sees” networks, not individual companies
For the first time perhaps, this “efficient” form of fraud appears to be finding an answer: a new “tool” will now identify such recurring patterns with multiple tax numbers and successive debts of different businesses, radically changing the form and logic of audits.
And this is because:
– traditional tax auditing “saw” one business at a time.
– the Jaccard similarity index, conversely, does exactly what human auditing could not do: it measures how much two or more businesses “resemble” each other in terms of their people, addresses and behavioral patterns. When this similarity is repeated across dozens of companies, all the tentacles automatically emerge — that is, the network.
Particular emphasis is also placed on so-called “single-use tax numbers”: foreigners without strong identification or even non-existent persons – often with stolen or fake identities – offer their tax numbers (with impunity), create companies for the network, “load up” with debts and then disappear.
Thus, instead of the major debtors that AADE and EFKA seek to collect debts of millions, in their place they found other persons and other companies with different tax registry numbers, often however with the same activity and at exactly the same address!
In the past, this “succession” of companies and debts was almost invisible in classic auditing. But even if a “suspicious” auditor could match the pieces of the “puzzle” by identifying the suspicious pattern in a business audit, the same could not happen on a mass scale. Even if he identified dozens of puzzle pieces, dozens or hundreds of others might escape him.
However, now the algorithm will automatically flag these traces, since every violator leaves a digital footprint in every “birth” and “death” of a network company.
How the Jaccard index works
In practice, the system will now record en masse:
- which person appears in which company,
- how many times the same person is repeated in different schemes,
- which tax numbers succeed others,
- which businesses show common or neighboring addresses,
- which companies have common natural persons in management or representation roles.
The result is a network interconnection diagram, which is not visible in a simple tax audit. Only when relationships are mapped comprehensively does the organized network become apparent.
However, for the “trick” to be revealed in its entirety and not fragmentarily, the new strategy provides that on-site auditing will be activated only after the data has already revealed the complete structure of the network.
The strike will thus be sweeping: auditors will raid simultaneously all points of the network – and not just one – having already formed a picture and not to search, but to confirm! “The data showed the structure. The investigations simply confirmed it” as characteristically noted in a special AADE strategy document for the new audit model (which also led to the exposure of the network with hundreds of “single-use” tax numbers). And this now becomes the rule for fraud with “straw men”.
