Credit scoring in the wider world is well documented, and loved and hated in equal measure. But when it comes to accurate analysis in the alternative finance industry, there is huge variability across the platforms.
Because the direct lending industry is relatively young within the finance sector, few platforms have had the chance to build up data-rich, seasoned loan books to use for developing risk prediction models. As a result, many rely on scorecards developed using information from the wider UK universe of companies, partner with credit reference agencies, and use a mixture of off-the-shelf credit risk scorecards, their own metrics and human judgement.
Commercial Credit Data Sharing (CCDS) is expected to kick off later this year; a scheme launched in April 2016 that requires nine major banks to share credit information on all their (willing) SME clients, furnishing finance providers, including alternative lenders, with a wealth of current account and credit information not previously available. This is expected to provide a considerable uplift in the accuracy of credit scoring models over time, and hopefully will facilitate the alternative finance industry in serving a host of currently overlooked smaller businesses.
However, it is not as simple as just having access to information; not all grading systems predict the same event. Some are calibrated on publicly-available data, to predict formal insolvency events; others are trained to predict all forms of company closure, insolvency and dissolutions; still others are trained on proprietary (i.e. not publicly available) customer data, including events such as late payments, not necessarily associated with insolvency, as practiced extensively by the main banks. This is the current challenge that data scientists face: building risk models that predict very specific outcomes; accurately reflecting investors’ experience of risk and return, but also affording borrowers fair and objective assessments.
ThinCats has allocated a considerable amount of time and resources to these issues, and the company is in a position to give UK SMEs more than just a number crunching, ‘computer says no’ experience, whilst also protecting the interests of the lenders.
As a secured lending platform, investors’ risk exposure and net returns are driven by both default risk and the ability to recover capital given a default. The ThinCats grading system makes the distinction between these two risk components, providing every loan on the platform with two grades; a number of security ‘padlocks’ and credit ‘stars’.
In order to produce these relative gradings, multi-layered processing models have been developed in-house. The credit grading model consists of an in-depth analysis of the company’s financial health, the ‘Hybrid Financial Score’ and its dynamism, the ‘Dynamic Score’. These scores are combined through multivariate analysis of the observed levels of insolvencies within the calibration data set (approx. 500,000 borrowing companies in the UK) to give a rating of one to five stars for each applicant. Over time, specific information about P2P defaulters as a differentiated segment, will be integrated into the model.
This is complimented by the security grading, represented by a maximum of five padlocks and determined by the asset to loan ratio, based on the value of the borrower’s assets relative to the outstanding loan amount. All loans listed on the ThinCats platform are then professionally qualified by the credit team.
This all combines to produce an award-winning analysis of information, ensuring that businesses looking for loans are given a fair and balanced hearing, and that investors know that each loan has been thoroughly assessed and vetted based on the most accurate information available; a complex system, but one that proves beneficial for borrowers and lenders alike.
John Mould, Chief Executive Officer, ThinCats