IOTA Peer-to-Peer Technical Assistance Project 'Segmentation of Debts and Data Mining for Debt Management'

Both Lithuania and Portugal sought opportunities to adjust their debt recovery strategies to the shifting landscape and customer expectations and improve tax administrations’ function to better target taxpayers and offer highly personalized approaches in debt recovery. Belgium, with its extensive experience in utilizing vast amounts of debtor information scattered across multiple databases to provide accurate profiles and intervention strategies, agreed to organize a joint workshop for the Lithuanian and Portuguese experts. This way, information was not only shared between the provider and beneficiary, but also between beneficiaries.

Between 10 and 13 June 2024, both Beneficiaries’ experts had working meetings hosted by the Belgian Federal Public Service Finance in Brussels, during which the Provider’s team shared many valuable insights on how the different datamining-based applications and tools in use in the General Administration for Collection and Recovery are constructed. Both teams also met with the head of local tax collection units, to explore how the tools are being perceived and used in practice.

More specifically, beneficiaries were introduced to the predictive models currently used by Belgium in debt management as well as those still in development. The meetings covered not only the theoretical aspects of the predictive models but also included demonstrations of their practical application. A visit was arranged to a department directly engaged in debt collection, where Lithuanian and Portuguese experts observed firsthand the operation of the information systems used for debt collection. The internal debt recovery interface was demonstrated, allowing beneficiaries to see how taxpayers are displayed in the internal information system and which indicators prompt employees to take specific actions.

Providers also shared insights on the data sources for developing predictive models, the most effective indicators for forecasting taxpayer behavior, and the core principles of model development. Significant attention was given to the CRISP-DM (Cross-Industry Standard Process for Data Mining) model. Each step of the model was discussed in detail, with practical examples provided and questions answered.

In return, Lithuanian representatives presented the automatic recovery from bank accounts process via PLAIS and the use of automated calls to apply nudging techniques for PIT debtors. The Portuguese colleagues gave presentations regarding the Bank Accounts Seizure by data transmission, Mutual Assistance Management Application, and the e-Invoice (e-fatura) system, particularly regarding B2C solutions.

Find out more about the Project by visiting the dedicated Project pages for Portugal and Lithuania on the IOTA website.