Why must we enhance Open Banking data with Machine Learning? (Part 1)

Open Banking data is considered the most reliable when it comes to lending decisions.

It is difficult to falsify since it is sourced directly from the bank, and access is controlled by a bank authentication process necessary to obtain the data.
To summarise: Open Banking data is protected by regulations (PSD2) and the technology used to access it.

However, direct access to bank data does not guarantee its quality.

Indeed, consumer use has evolved, and an individual’s banking data is now spread across several financial institutions.
We no longer speak of a main bank: the vast majority of consumers have several banks. Salary might be received in one account, and bills might be paid out of another.

It is easier for consumers to hide financial information that “embarrasses” them by sharing a “neutral” bank account rather than all of their accounts as part of their loan application.

Therefore, as a lender, it is important to adjust and identify these different accounts so that they are appropriately taken into consideration in the risk analysis.

At Algoan, the data science team has developed an activity score that enhances our lending decision support tools based on Open Banking data.

The activity score allows lenders to identify accounts with abnormal or insufficient activity to make a lending decision. What is nice is that a lender does not have to set up complex business rules that would confirm the relevance of synced accounts.

For example, an account with only internal transfers could be considered healthy (as graded by a behavioural score) due to stable cash flow and the fact that the account is balanced.
However, that activity is insufficient to make a reliable lending decision. So, the lender should request that the borrower provide another account.

How does the Activity Score work?

The activity score is based on a combination of strict rejection rules as well as indicators that evaluate activity in a bank account.

The strict rules cause immediate rejection of accounts that are obviously inactive. If any of these strict rules is triggered (for example, no outgoing transactions identified in the accounts), the activity score suggests an immediate “No Go.”

However, if the strict rules are not triggered, the evaluation automatically moves to the next level. Algorithms for detecting the type of payment (card payment, direct debit, and so on), the category (salary, loan repayment, shopping, and so on), and regularity are then used to calculate a set of indicators based on the following three factors:

• Account activity, such as number of transactions, depth of history, and so on;
• Income, such as salary, allowances or other regular income;
• Expenses, such as card payments, rent, utilities, and so on.

Our algorithms weight these indicators to produce a score between 0 and 100 that can be interpreted as follows:

• Less than 20: insufficient activity;
• Between 20 and 50: limited activity;
• 50 and greater: normal activity and characteristically active accounts.

Today, between 5 and 10% of users have an activity score below 20 (insufficient activity).

This detailed interpretation of data was made possible by conducting extensive Open Banking data analysis using enhanced algorithms, beginning in 2018. Because we can precisely identify different Open Banking profiles, we can provide a reliable analysis of a loan applicant’s banking activity. This analysis complements a behavioural score, which evaluates the probability of the loan applicant defaulting based on his financial behaviour. This combination allows the lender to single out those who present the most significant risk to the lender.

With the advent of Open Finance, greater access to individuals’ savings accounts and investment data will undoubtedly further improve the performance and reliability of lending decisions. In this context where several sources of financial data can be cross-referenced, fraud—already jeopardised by Open Banking—will be even more difficult to commit, and that suits us just fine!