Predictive features, instantly.
Credit scoring is about finding predictive features. Identifying predictive and unique features in transaction data can be a time-consuming challenge. We have spent hundreds of hours looking for predictive patterns in transaction data so you don't have to.
Creditstar found a 6 percentage point GINI uplift with Nordigen scores.
Our behavioural scoring engine is built to recognise risk-critical behaviours and patterns in categorised transaction data and calculate a probability of default score from 0% to 100%. The score is built from 1,000+ behaviour features and it can be used as a standalone score for pre-scoring or as a predictive supplement to existing scorecards.
What is behavioural scoring?
It's being able to calculate a probability of default based on behaviours identified in transaction data.
Why it is important?
To improve the accuracy of credit scoring models, it's important to look for new predictive insights. Transaction data is the perfect source for new predictive insights as it contains real financial behaviours.
Behavioural scores can be valuable across departments.
Simple API, supports multiple file formats
Our RESTful API supports a wide array of file formats, including our own JSON format, as well as transaction files (JSON, XML) directly from account aggregators such as Yodlee, Instantor, Kontomatik, Pich Technologies, Saltedge as well as bank statements in PDF formats in selected countries (e.g. Latvia, Lithuania, Estonia, Finland).
Hosted on the secure Amazon Web Services cloud
Security and scalability are our top priorities. For this reason we've deployed our solution on AWS (servers located in Dublin, Ireland) to ensure that your data processed securely and the service is always running.
Fully customisable to any lending product and audience
To ensure the maximum uplift, our score can be trained for your specific use-case or you can use one of our pre-trained models. For your score, we will pick only the features that give the best GINI uplift for your models.
- expected GINI uplift on new loan applicants: 5 to 10 percentage points
- expected GINI uplift on returning loan applicants: 2 to 5 percentage points
Built from 1,000+ predictive features
Feature engineering is arguably the most painful part of credit modelling. We have invested a lot of our own time to generate a long list of features so you can move faster. The features are based on categorised data and include trends, predictions, events, and ratios.
- trends - "changes in disposable income over time"
- predictions - "expected household expenses within next three months"
- events "first five transactions after receiving a loan"
- ratios "average disposable income to average gambling"