If you have to mention a single word that is currently embracing the whole finance and banking world, it would be "fintech". It generally refers to a new financial industry that applies technology to improve financial activities. Use of innovation and technology is the core of its concept.
It's like the Uber of financial services. You get one omnipotent app for financial services like the ride-hailing app via which you can buy things, transfer funds, invest in shares, take loans, explore myriad arrays of financial products and that of multiple financial institutions. And all that in the blink of an eye, in real time. Well, that's from the consumer-side idea of fintech.
From the perspective of a financial service provider, it's the challenge to build such a flexibility for its clients, which requires it to reengineer each of its core processes and business models. This is where innovation and technology become catalysts.
One of the core operations of financial institutions are lending. Lending money and earning interest is the primary source of earning of a bank and quality of the lending portfolio is the key indicator of the health of the financial institution no matter how versatile it is. Innovation and technology, hence fintech, have been aggressively changing the way banks conduct their most activities.
Traditionally, lending or providing credit facility, whether to individuals or business entities, was dependent on human subjective judgement. For instance, when an individual applies for buying a house, usually an underwriter is assigned. The underwriter interviews the customer, analyses his lifestyle, authenticates his earnings, evaluates the value of the house, assesses its liquidity in case of default, and then makes a judgmental call on the creditworthiness of the individual. This 'judgmental call' can also be attributed to the underwriter's ability to identify strengths and weaknesses (risks) of the borrower and his (underwriter's) perceptions. Therefore, it becomes a subjective call.
That judgment may not necessarily be 'yes' or 'no', but may vary in terms of loan amount, tenure, collateral coverage and so on. Advanced markets try to mitigate this by assigning credit rating to clients or businesses. Also, risk appetite varies among banks in such a way that one may say 'yes' to a client, who was said 'no' by another risk-averse financier. But nevertheless, the fact remains, if you assign a human underwriter to underwrite a loan, the decision is subjective.
Fintech is challenging this aspect in two ways. First, irrespective of subjectivity, a human underwriter reaches its lending decision by deducting its observations via several logical decision trees. For different factors, the decision trees are differently weighed, they are regularly modified and updated based on new insights, and financier-specific risk appetite is attached to the whole decision-making process with insight from historical learning. Now, given the number crunching capacity of today's technology, all these can be done using modern statistical tool in few minutes. Moreover, weightage, insights, observations regarding each material factor may vary from analyst to analyst. But for an algorithm-based risk management solution, all that remains the same, hence, objectivity can be introduced in the whole process.
Second, it's way cheaper. Any process dependent mostly on human interactions is costly, incrementally. Technology-driven ones mainly have one-time sunk cost. Then it's just maintenance and keeping it up-to-date. In a cost conscious and shrinking margin industry, this is a make-or-brake issue.
Automation of credit underwriting in retail lending is widely spreading and gaining acceptability around the world. It is more relevant in consumer loans (home loans, car loans, personal loans) underwriting, while its access to SME loans is also catching up. Automating underwriting in business loans is markedly more complex than consumer loans as there are increasing numbers of internal and external factors that impact the wellbeing of a business.
Automating credit underwriting is prone to be susceptible to major downsides. One factor can be termed as "garbage in, garbage out" phenomenon. The algorithm that encompasses the impact of different risk and reward factors of a borrower largely depends on the accuracy of the information inserted in the system. If the information is flawed or biased in anyway, the underlying result of risk management will also be erroneous leading to failure of the whole system.
In organisations at the earliest stages of adopting automation of credit underwriting, the credit analysts are assigned with the task of inputting borrower's information in the system. In this early stage, automating algorithm and human underwriting are run in parallel to ascertain the efficacy of the new system as well as building confidence in the organisation.
Another risk is that since the automating credit algorithm is a structured set of logical factors, borrowers may get smart enough over time to manipulate the system to gain better than deserving results. For instance, a grocery store with lower than average inventory turnover may indicate that it is operating and managing profitably. One such grocery store figuring out this logic may try to 'game' the system by manipulating lower than actual inventory turnover.
To guard the system against these pitfalls, it is important to regularly fine-tune the algorithm in such a way that its features do not become obvious over time. Also, there needs to be 'circuit breakers', intelligent checks and balances in the process making it ever-evolving and harder to penetrate. Designing the questionnaire or the borrowers' information collection interface needs to be intuitive and smart so that 'out-of-line' or 'unrealistic' or 'too-good-to-be-true' responses can easily be identified and alerted/ejected by the system.
Also, it is more important to construct the data set from more mechanical source than subjective source. Common sense dictates the arrays of information will be like business experience, ownership structure, banking habit, credit history, profit margin, debt-equity ratio, revenue growth, collateral coverage, education and age of owner, permanent residency etc. Many of these can be manipulated or 'dressed-up' for a better credit score.
So, are we heading towards an era, where all lending decisions in banks will be made through some complex and mysterious algorithm, devoid of human interaction?
Dr. Philip Tetlock argued regarding this phenomenon of system-driven forecasting in his bestselling book, "Superforecasting". While he appreciated the inevitable benefits of automation, the need of human interaction will never really cease to exist. It will always be 90 per cent system-driven, but human judgment will always determine the course of the algorithm, the factors to be incorporated, and expert opinions to be addressed.
The scope of scaling up credit underwriting using technological innovation is enormous. It's too big and too profitable to ignore. And, the only way to get this advantage is by getting started. This is fairly a new phenomenon in a developing economy like Bangladesh. Relatively inexpensive and easily available human resources are making it even difficult to most bankers to even make sense of automation. In addition, the technology of Big Data construction and analysis, and the data science technologies that require for automating complex operations like credit underwriting is yet to be readily available here. But it is high time to get going. Those who will invest in this at this early stage, will get a head start and will have unique upper hand by the time this will become mainstream.
The writer is Assistant General Manager and Head of Credit (Small Business) at IDLC Finance Limited. [email protected]