Finnova Analytical Framework

Finnova is a leading provider of banking software in the Swiss financial centre. Finnova supports banks and outsourcing providers in achieving digital transformation and open banking, and their solutions are used by more than 100 banks.

Finnova is recognised for their powerful innovative banking solutions, in development, in operation and in advisory. Together with their partners, Finnova supports over a hundred customers from the financial services sector in being profitable and competitive. Through Finnova's Open Platform, which enables open ecosystems, banks are able to be more innovative, agile, and better satisfy their clients' needs. Finnova supports their customers' individual digitalisation strategies, both with their own products and with a broad-based partner network.

Banks have to meet complex regulatory requirements. Rule-based systems can only provide limited support. Without the use of powerful, precise analytics, these requirements are managed with high risks and costs.

The solution was to create the Finnova Analytical Framework, a business solution based on Prospero's software platform for predictive analytics, DetectX®. It is embedded in the Finnova banking software and responds to requirements from all suites and modules. DetectX® can be used to map the entire range of applications of machine intelligence for optimising the business processes of a bank in a uniform platform. Examples are anti-money laundering, fraud prevention, risk management, robo-advisory and potential-oriented sales management.

DetectX® offers analytical approaches such as supervised learning, unsupervised learning, and rules based on expert knowledge. Supervised learning uses a target variable on which the model is built: a representative sample is drawn and separated into train and test datasets, and the optimised model is then applied to the universe of data in the productive environment. Within unsupervised learning the entire data is taken and used in the analysis and modelling process in an unbiased way, so unknown and suspicious behavioural patterns can be revealed. The combined use of both methods makes the models accurate and stable. Expert knowledge is no longer used in isolation but in combination with the other two, with profiling as a preliminary processing stage based on static and dynamic customer data.

The heart of the framework is the DetectX® Analytical Engine, which validates, optimises and calibrates models continuously in the background. Whether the models are for fraud detection, transaction analysis, link analysis, robo-advisory or analytical CRM, the engine ensures maximum precision and stability of the modelling process. Data from different sources is loaded for processing, checked for completeness and distortion, transformed, supplemented and enriched. In a complex optimisation process, millions of combinations are calculated and the relevant factors are continuously determined with their weightings, all in an automated process so that expert intervention is reduced to a minimum. Model creation and model application run in parallel in an unlimited, scalable process.