The COVID-19 outbreak is causing a lot of turbulence in present times - experts across the world are expecting a big crisis that will permanently change the world. This situation not only has an impact on almost every aspect of our lives but also forces us to rethink many business areas, including credit risk management.
COVID-19 is putting credit models to the test - they might lose their predictive power when faced with the kind of unique pandemic circumstances. Some companies are trying to fix, modify or replace them to fit the new reality. But if we build our risk assessments on the solid data ground, changes in the environment should not affect the process. Check how solid data ground should be prepared.
That's a continuation of describing how to work with data to drive your business based on transaction history. In the previous part we described how rules-based systems and categorisation might be moved to the next level with machine learning.
Standard credit scoring usually includes only a set of basic rules which measure default rates. Traditionally models include basic information e.g. income, loans, account balance, spendings or debts without a deeper understanding of financial behaviour patterns and without tracking changes. Usually, scoring model is defined once, at the very beginning of work with risk and it is never changed (there is always a lack of resources or no one is eager to touch such a sensitive area).
It's working in many lending companies right now, but it may be not elastic enough in the dynamically changing world where the number of customers behaviours patterns is growing rapidly and crises (like pandemic circumstances) appear. Moreover this kind of approach is time-consuming. Usually, it's calculated manually and we are always limited by the time and volume of data we can analyse. And as every manually performed work, its efficiency is not verified as often as needed.
The easiest way to enhance credit scoring decisions is to feed it with data and keep them the most informed. Open banking gives us access to real-time banking data - we can use them to confirm identity and credit assessment. The current financial picture is the key factor in proper risk management - especially in dynamically changing pandemic environment.
But it is only the first step - even if we have tremendous amounts of information, without proper analytics and turning data into meaningful insights we are not able to take full advantage of it.
One of the most compelling examples of enhancing credit scoring decisions by machine learning-based automatisation are features. Features are created by combining all available data in every possible way. The best scoring solution feature has a pool of over 500 000 parameter combinations. Majority of them are not valuable insights. So in the next steps, we should choose the most useful features which measure changes in critical areas describing creditworthiness and cut the insignificant ones. These choices are not based on subjective decisions but as part of an automated process, and the main factor is the impact on the model effectiveness.
Well defined features allow us to measure changes in crucial areas describing creditworthiness (eg. changes in income structure), track changes when the environment and show the most current and complex picture of financial conditions, without delays characteristic for traditional databases. Whatsmore this kind of model is updated and evaluated automatically wherever is needed.
What's more, a good scoring model supported by machine learning algorithms allows not only to lower the default rate but increase the acceptance rate as well. (Data from Kontomatik scoring use cases shows that acceptance rate increased av. about 2 p.p. , default rate lowered about av. 4 p.p.).
All this will happen in real-time - so in a few minutes, you can prepare a data grounded risk assesment. Your clients' user experience remains unchanged and you are able to offer competitive offer with reduced risk for your business at the same time.
Would you like to talk about how to enhance your scoring model? Write to us and let's talk about your business!
APIs are essential to the process of Open Banking because they pave the way for data sharing in a safe, secure, standardized and efficient w
A package of new draft proposals, PSD3 and PSR, includes plans to strengthen customer rights, combat fraud and improve APIs
Open banking is extremely popular around Europe, not just the UK