This website uses cookies for analytics purposes. Learn more
Open Banking Model Build
Our client was an established unsecured lender who wanted to maximise the value they were getting from their Open Banking data
Although the lender was an early adopter of Open Banking data, they were using it only for income verification and assessing affordability
Two of our client's biggest acquisition channels were now providing customer Open Banking data upfront in their API journeys, enabling it to be used for pre-approved decisioning
They were therefore keen to use this opportunity to expand their use of this valuable data source, by leveraging an existing Open Banking data sample to build a new credit scoring model
Vestigo were engaged by the lender to build a machine learning (ML) model that utilised Open Banking data alongside existing credit bureau data points. The model predicted the probability of default
The project included working closely with the lender to sample a suitable dataset for modelling. Our team also liaised with the credit bureau to provide referenced loan performance on the lender's rejects and not-taken-ups
We investigated several modelling options from the more traditional Logistic Regression to more complex approaches such as Neural Networks and Random Forests, to find the best model for the objective
The final model significantly outperformed the existing model that used credit bureau data alone. The outcome highlighted the way Open Banking data can complement traditional data points to achieve more powerful predictions
For more on information on Open Banking, check out our article looking back on challenges and lessons learned in the first 5 years of Open Banking, as well as the opportunities for lenders going forward.
Get the latest news from Vestigo