Banks Need to Invest in Digitisation to Better Serve SMEs

If a person from the 1980s was somehow transported to present day, a lot about the world would astonish and amaze them. Our current era is nothing like the future that was imagined in popular culture at the time — there are no flying cars, we do not have the ability to teleport, and, sadly, as the last 22months have demonstrated, we still haves ignificant improvements to make when itcomes to healthcare and social equality. One thing, however, that this time traveller would not find surprising at all is the decision-making process used by banks when a small or medium-sized business applies for a loan.

Indeed, much about that process has changed little (if at all) over the last four decades despite the enormous advances in technology and society in that time. Banks still make credit decisions based largely on intuition and experience of their credit officers, and if they use models for forecasting at all, the forecasts are based on historical data and share fundamentally flawed root assumptions: firstly, that tomorrow will be a lot like yesterday and, secondly, that most businesses are more or less alike. Over the years, this has proved to be good enough for the most part, and, as a result, these models are considered sacrosanct.

The Traditional Approach

Credit modelling in highly commoditised markets such as credit card, mortgage or auto loans is based on large volumes of historical data showing the performance of similar loans. From a pool of loans of a certain origination vintage, it is possible to observe the historical rates at which loans enter various stages of delinquency and eventual default and thereby estimate a multistate Markov model. This can then be used to simulate the trajectories of loans in a given portfolio and estimate the probability of default for these loans.