Credit Scoring and Automated Credit Decisioning in asset finance companies

Data: 2020.02.12

Author: Kestutis Skrodenis

One of the first features that our customers request in our SOFT4Leasing implementation projects is, “do you have an interface to credit bureau?” It is then followed by – “can you automate the approval process?”. The first answer is easy – Yes, we do. The technical part of the interface is not a big deal and, of course, it needs to be configured depending on which credit bureau, which country, what services required. The most commonly used services are Experian, Equifax, TransUnion, Creditinfo, Creditsafe.

It is a pretty straight-forward process to send a request, receive credit file, parse information relevant to credit decision – age of credit file, bankruptcies, defaults, external credit score, etc. Additionally, it will work fine in terms of consumer finance and can be automated. Of course, there are cases when SOFT4Leasing would refer to a credit officer for manual review and subsequent credit bureau searches (further searches on financial connections, co-borrowers). In business finance, the process is typically more complex than that – search based on search results is often required to do check on not only business but also directors, guarantors, beneficial owners. So the process can be partially automated, SOFT4Leasing system would make the first credit bureau inquiry and refer to credit officer to make the final decision review. For large ticket deals (i.e., above certain limit defined by companies’ credit policy and configured in SOFT4Leasing setup), the application would be referred to credit officer, to check multiple credit references, prepare credit recommendation and escalate to appropriate approval authority (again, levels configurable in SOFT4Leasing, for example – credit officer, two credit officers, head of credit, managing director, credit committee, etc.).

For some financial products, the approval process can be highly automated (for example – small ticket consumer lending). Other products most likely would require human interaction. In this case, the system still does part of the job – deliver all the relevant information in one place, including the credit bureau file.

Credit decisioning process


General credit decisioning process goes as follow:

  1. “Submit application” action, the trigger for the approval process to start.
  2. Technical data check for data consistency and completeness. The system will execute a set of data rules – if borrower identification details are complete if required supporting documents attached.
  3. Check if you already have a recent credit file. The aim is to reduce the number of credit bureau inquiries if the customer submits multiple applications or application re-submissions – re-use local credit files.
  4. Call to credit bureau interface, retrieve credit file and parse data fields (in other words, get the variables, which are needed for decision rules)
  5. Calculate internal variables from customer history and application details. For example – the current exposure with an existing customer, arrear status of an existing customer, capacity to pay based on income details, and financial commitments of the customer.
  6. Run the internal “Decision Tree” engine, which is a set of rules for “Approve,” “Decline,” or “Refer.” End up with overall credit decision, which may be final, or maybe a recommendation to credit officer.
  7. Detect minimum approval authority – if the application can or cannot be auto-approved, and what type of workflow to run next.
  8. From this point, the process can go one of the five ways, which are listed below.
    a) Application auto-approved and approval letter issued to the customer and introducer (equipment dealer or finance broker);
    b) Application auto-declined and goes to the archive. Decline letter issued;
    c) Application referred to credit officer with recommendation to approve;
    d) Application referred to credit officer with recommendation to decline;
    e) Application referred to credit officer with no recommendation.

When a business grows, and number applications grow, companies are willing to rely more on automated credit decisioning. So, let’s analyze one real-life example from consumer finance. The application flow on average 4’000 applications per month. External credit scoring interface is used, and internal credit decision business rules apply. The diagram below shows rate or approval (number of approved applications as a percentage of received applications). In this data sample, the average approval rate is 72%, where 58% of applications were auto-approval by the system and another 14% approved by a credit officer. The remaining 28% of applications were withdrawn by the customer or declined by a credit officer.

Figure 1 Rate of approval (number of approved applications as a percentage of received applications) per month

In the implementation process, the business rules were fine-tuned to get the optimal balance of approval rate vs. credit risk. The application flow started with several hundred applications in the first few months. After six months, it reached over a thousand per month, and the following year reached a steady average of 4’000 applications per month. From the diagram, we can see how the approval rate evolved:

  • Months 1-5 – “go-to-market” stage (from 30% down to 7% manual approvals)
  • Months 6-10 – “all automatic” pilot (~3% manual approvals)
  • Months 11-17 – balanced stage (stable, ~20% manual approvals, 80% automatic)

The scoring and auto-decision rules have been fine-tuned, so the auto-approval rate varies from 47% to 72%. Overall approval rate varies from 60% to 86%.

Credit policy rules are a trade-off between credit risk you take and how much new business you generate. To get the balance right, you need to monitor the level of arrears (missed customer payments, late payments, underpayments). We need to look at the next diagram, which is the current level of arrears by the month of credit approval.

Figure 2 Number of accounts in arrears as % of the total by months of approval

For example, month #13, we reached the record level of approvals – 86% was reached (diagram above), and today 5.8% of then-approved accounts are in arrears. Another clear correlation is month #8, record low rate of approval – 60%, and today only 3.8% of then-approved accounts are in arrears. I took these two very simple examples to illustrate the two KPI’s, besides looking at how they can (or if they do) correlate. In real life, there are many more factors that will affect your credit decision rules and expected credit losses. The key point are:

  • Constantly improve and innovate
  • Balance your credit rules to get more business
  • Have flexible software to facilitate that.