Advanced Analytics & Scoring / Adastra

Risk Advisory

Advanced Analytics & Scoring

Advanced Analytics and Scoring is an integral part of modern risk management. Quantification of the expected risk associated with a customer, measurement of actual riskiness and constant challenging and fine tuning of the risk assessment process is essential for underwriting, fraud management and collections processes.

An effective way to assess the riskiness of a customer is to incorporate into carefully chosen places in particular decision making processes a scorecard considering different metrics associated with the customers behaviour. The scorecards can be purely statistical, expert-based (based on data availability) or a hybrid (mixture of the two).

Our Services

We help organizations to gain competitive a advantage by:

  • Understanding how to use Advanced Analytics to gain advantage
  • Outsourcing of experienced Advanced Analytics teams
  • Data processing and development of scorecards
  • Definition of a client segments based on the score 
  • Defining further actions in particular processes based on the classification
  • Creating processes for effective measurement of scorecard performance
  • Regular calibration of scorecards

To build the statistical scorecards we use standard logistic regression models or AI/ML methods.

Underwriting Scorecard

Typical example of scorecards usage is to determine expected probability of default (PD) of customers while still in the underwriting process. Internally developed scorecards fitted to particular portfolios will have significantly higher accuracy than external scores (e.g. from credit bureau). An accurate customer segmentation in underwriting allows organizations to increase booked volumes, onboard more a healthy portfolio, implement risk based pricing and set-up lower maximal credit limits for more risky customers.

Fraud Management Scorecard

Two typical fraud use cases are being resolved with advanced analytics. An application fraud, where segmentation helps to determine customers that are likely attempting to commit a fraud for further verification and transaction fraud, where anomaly detection helps to identify suspicious transactions.

Collections Scorecard

Understanding of the riskiness of clients in collections processes allows you to focus your resources on clients who need urgent treatment. The usual scorecards cover pre-collections segmentation and several collections segmentations based on the structure of the collections process (e.g. early collections segmentation, late collections segmentation etc.)

Behavioral Scorecard

Understanding of the quality of an organizations credit portfolio based on customer behaviour while on the books is necessary for educated managerial decisions. Absence of the understanding of the development of riskiness of the portfolio can have far reaching implications. On the other hand, the detailed understanding of riskiness of existing customers, allows organizations to target, cross-sell and up-sell campaigns on customers with the expected risk profile.

Outsourced Advanced Analytics Team

Do you have work for an experienced data scientist but missing hands with an expert knowledge of both credit risk processes and advanced analytics? We are ready to outsource experienced data scientists specialized in end to end implementation of advanced analytics use cases into credit risk processes.

Scorecard Development

Our services cover end to end development and implementation of appropriate scorecards tailored to your specific portfolio and processes by a team of experienced data scientists specialized in implementation of advanced analytics use cases and credit risk processes.

AI Driven Anomaly Detection Framework

With an increasing penetration of fully online customer journeys, new threats for organziations providing loans or services on credit are emerging. Without a human touch in the process, organizations usually rely on pre-defined rules and data-driven approaches based on machine learning algorithms.

Such approach was designed and works well for simple or complex known fraud patterns observable from historical data, but are prone to fail to new unusual fraud attempts. Therefore organizations more and more adopt 3rd level protections supplementing standard rules based engines and machine learning algorithms. These are trained on historical data with AI powered anomaly detection automatically detecting new threatening patterns from real time data. 

Our next generation solutions for fraud prevention combines:

  • Pre-define rules based engine to detect simple known fraud patterns
  • Machine learning algorithms detecting complex know fraud patterns
  • AI powered anomaly detection revealing potential new threats