Scoring yesterday, today and tomorrow 🚀

Imagine that you are a bank or an investor who has some free money to invest. If you want to lend someone money on credit, that is, issue a financial instrument with a predetermined payment schedule, you want to be sure that you’ll see a return on the money, and, of course, in a timely fashion. But any investment entails a risk, and you must be able to quantify this risk before making financial decisions.

The main instrument used in such cases involves statistical or scoring models that are developed on a historical sample of issued loans that are marked as “good” (paid on time) and “bad”, or defaulted.

There is a range of serious obstacles and limitations to this setup, which, as we will demonstrate, leads to a certain injustice in the banking sphere.

1️⃣ The first limitation is that all documented cases of non-payments by borrowers are in the past. By automatically using information from the past, lenders are like a driver who looks solely in the rear-view mirror when piloting his car. That is, the process of measuring risk entails seeking out a set of observations regarding a situation close to what is expected in the planning time-frame, as well as possessing the ability to foresee potential changes that cannot be discerned in the data culled from the past.

2️⃣ The second limitation is the information about the borrower which is available at the time of decision-making. There’s a reason (or cascade of reasons) for each default, e.g., the borrower took out a loan that he never intended to pay, i.e., fraud; the borrower has no problem managing a heavy credit load today, but let’s say he loses his job a year from now, or comes up against other unforeseen circumstances — things change, and none of us knows what’s around the corner. The problem with risk assessment is the difficulty, when deciding to issue a loan, of seeing signs of an increased probability that the borrower will someday stop making payments, and determining the likelihood that this will happen during the term of the loan. This is why the loan application from a bank is so detailed and broad in scope — the bank needs exhaustive information for risk assessment. And yet, if you don’t have an extensive credit history, your credit rating will be like a hospital patient with an average temperature. This isn’t fair, because it means that good borrowers with a solid credit history have to pay for risky borrowers.

3️⃣ The third source of injustice stems from the scoring models (logistic regression and other linear models) and the lack of data about any non-payments, which leads to averaging and simplistic classifications of borrowers based on their risk profile.

In the past, banks used scoring as an excuse to reject a potential borrower’s application for a loan product.

In the future, financial companies will use scoring to select a product with a fair price to serve the borrower’s interests or resolve a specific situation he or she faces. Currently, we are witnessing a transition period. Less information is required from the customer and in the financial organizations of the future, all that a potential borrower will be asked is to provide his or her credentials and interests (if they aren’t automatically evident). Now the price that the borrower pays for a loan product reflects operating expenses, expected risk, partner commissions and net returns for the creditors and shareholders. In this article we don’t address the issues of costs and funding, but the better the expected risk for each borrower, the more equitable the rate of the loan product.

Information which is now virtually never at the disposal of the banks is information about the borrower’s psychology, his or her social ties and their quality, and information on non-banking debt relations. Psychological signs are investigated very actively and soon — in a year or two will be considered in half of all credit decisions. Information about social connections is used on the principle of “tell me who your friend is and I will tell you who you are,” and the main thing that limits the development of this type of information is legislation on the transfer of personal data. Therefore, the owner of this information will have the advantage in the credit solutions market. The non-banking sector that features debt relations is not covered at all, and this is the main thing that we want to use as an advantage in how our company scores potential borrowers in the initial stages of building a credit rating system.