Last year, the European Central Bank imposed sanctions totaling nearly 18 billion euros on supervised credit institutions. On average, according to Gartner, organizations lose 12.9 million dollars a year due to insufficient Data Quality: this is why, especially in an open digital ecosystem, it is crucial to ensure the quality of data acquired from external sources (infoproviders) and to maintain a strong focus on the governance of this information.
This was highlighted in a recent ABI Lab Observatory and reiterated today by Monica Ripoldi, Head of Data Quality at Banco BPM, while presenting a case study with Irion at the Politecnico di Milano, as part of the Big Data and Business Analytics Observatory. As the manager explained, the process of selecting data providers takes place in four phases:
- First evaluation step on the infoprovider itself
- Methodological assessment of the data
- Actual Data Quality assessment
- Overall technical-functional evaluation
Anomaly detection with AI
Through the anomaly score detection method (the identification of potential anomalies, for example in the analysis of historical series of the requested information), it is possible to implement these controls in addition to those already in use in the process. The outliers (anomalous values) identified by AI are sent for verification to a domain expert.
As part of the financial group’s Data Quality framework, Irion EDM acts as the “core system” underpinning the governance and execution of data quality controls. The Banco BPM system is structured around four classification dimensions (process quality, internal quality, relational quality, and temporal quality), which are further divided into seven quality criteria, based on the main regulations and market best practices.
There are three critical success factors: collaborating with the colleagues involved, continuously evolving and integrating the system, and ensuring data quality oversight through structured processes. Through summary KPIs, the company ultimately arrives at an objective evaluation of the qualitative state of the data and the level of adherence to the defined quality objectives.
