A co-founder of Dataiku explains how a technology that serves the success of his startup will be used by Bercy to find tax evaders

The government has presented an anti-fraud bill to recover about 60 billion euros a year.
The text provides for several provisions strengthening the investigative and sanctions powers of the tax authorities.
Among them, the increased use of a technology derived from mathematics and artificial intelligence: data mining.
The French startup Dataiku is experiencing rapid growth thanks to this technology. Its CEO explains how data mining can be used to detect tax fraud.

The French tax authorities will make greater use of artificial intelligence (AI) to detect tax fraud, it was announced on Wednesday, March 28, 2018 after the presentation of the anti-fraud bill presented in the Council of ministers.

This text aims to strengthen the investigative and sanctions powers of the administration.

The challenge is to combat practices by companies and individuals that, according to estimates, cost the state almost 60 billion euros a year.

To do this, in addition to the creation of a tax police dependent on Bercy, the bill should make it possible to remove one of the obstacles in the fight against fraud: the exploitation of data, resulting from the crossing of files of the various administrations, hitherto impossible. To do this, the administration will further use data mining — a technology that consists in collecting and analyzing data by applying a statistical method to them to learn lessons.

“Data mining now accounts for 10% of controls. We are aiming for 20% in 2018 and the goal is to eventually reach 50%,” one explains in Bercy, indicating that the number of specialized teams will be strengthened, including by using external specialists.

Dataiku is not one of these service providers. But the nugget of French tech – which has raised $ 42 million in a year — owes part of its success to its mastery of data mining.

Asked by Business Insider France, Florian Douetteau, co-founder and CEO of the startup, explains how the technology could be useful in the case of tax fraud.

“One can imagine that data mining is used to detect patterns of false invoices between companies or detect suspicious peaks in revenue distributions over a year, comparing them with companies in the same sector. This is millions of data. It’s impossible to see this with the human eye,” he details.

Publisher of a software designed to improve the analysis of company data, the startup doubles its activities and employees — 140 to date — every year by allowing companies to detect customer fraud. L’Oréal, AXA, Aramis Auto, AccorHotels or Showroomprivé and a hundred other companies work with Dataiku.

Asked about the expected benefits of the technology, Florian Douetteau is convinced “that there will be real effects. We’re on an already proven technology. Its success depends on the ability to get people on board with this technology while they have operating habits in their professions.”

Pending the consideration in Parliament, here are the other developments included in the text carried by the Minister of Action and public accounts, Gérald Darmanin:

the taxman will finally have the opportunity to publish his sanctions, a recourse to the Anglo-Saxon technique of “name and shame” which will target companies in particular;
administrative sanctions against “intermediaries” who contribute to fraud schemes, with penalties of a minimum of 10,000 euros and up to 50% of their fees;
the establishment of a regularization window for “bona fide” companies and systematic publication of answers to companies’ questions about tax schemes if they are deemed to be of general interest;
obligations for internet platforms like Airbnb to report on the income they generate and the identification of their users.