Credit Scoring: The Development Process from End to End





Data Science


By: Natasha Mashanovich, Senior Data Scientist at World Programming, UK

Part 1: Why Do Credit Scoring?


“Buy now, pay later” is a tempting offer made by many financial and retail firms to their customers to increase their customer base. However, both parties need to be aware of the risks when taking such credit decision. It is important for both the lender and the customer that the customers will be able to honor the credit obligation and payback what is owed for the purchase by the end of the loan term. Lenders need to be able to assess the risk of default for each customer so the lender can decide to whom the offer should be granted.

What is a Credit Score?

Advances in technology have enabled financial lenders to reduce lending risk by making use of a variety of data about customers. Using statistical and machine learning techniques, available data is analyzed and boiled down to a single value known as a credit score representing the lending risk. This value can help guide the decision process. The higher the credit score the more confident a lender can be of the customer’s creditworthiness. Credit scoring is a form of Artificial Intelligence, based on predictive modeling, that assesses the likelihood of a customer defaulting on a credit obligation, becoming delinquent or insolvent. The predictive model “learns” from by utilizing a customer’s historical data together with peer group data and other data to predict the probability of that customer displaying a defined behavior in future.

The greatest benefit of credit scoring is the ability to help take decisions in a fast and efficient way such as to accept or reject a customer or increase or decrease loan value, interest rate or term. The resulting speed and accuracy of making such decisions has made credit scoring the cornerstone in risk management across sectors including banking, telecom, insurance and retail.

Credit Score Types and Customer Journey

Credit scoring can be utilized throughout the customer journey, spanning the entire customer experience during the length of the relationship between a customer and an organization. Although primarily developed for credit risk departments, marketing departments can also benefit from credit scoring techniques in their marketing campaigns (Figure 1).

As depicted in Figure 1, different credit scores are utilized at different stages of the customer journey:

  • Application score assesses the risk of default of new applicants when making decision whether to accept or reject the applicant.
  • Behavioral score assess the risk of default associated with an existing customer when making decisions relating to account management such as credit limit, over-limit management, new products, and the like.
  • Collections score is used in collections strategies for assessing the likelihood of customers in collections paying back the debt.

Figure 1. Credit Scores throughout the Customer Journey

Credit Risk Scorecards

Over the years, a number of different modeling techniques for implementing credit scoring have evolved. They range from parametric or non-parametric, statistical or machine learning, to supervised or unsupervised algorithms. The most recent techniques include highly sophisticated approaches utilizing hundreds or thousands of different models, various validation frameworks and ensemble techniques with multiple learning algorithms to obtain better accuracy.

Despite such diversity, there is one modeling technique that stands out – the Credit Scorecard model. Usually referred as Standard Scorecard, it is based on logistic regression as the underlying model. Compared to other modeling techniques, this method ticks many of boxes, making it the favored approach among practitioners and is used by nearly 90% of scorecard developers. A scorecard model is easy to build, understand and implement and is fast to execute. As a statistical/machine learning hybrid, its prediction accuracy is comparable to other more sophisticated techniques and its scores can be directly used as probability estimates and hence to provide direct input for risk-based pricing. This is critical for lenders that comply with the Basel II regulatory framework. Being very intuitive and easy to interpret and justify, scorecards are mandated by regulators as the exclusive credit risk modeling technique in some countries.

A scorecard model result consists of a set of attributes (customer characteristics) typically displayed in tabular form (Figure 2). Within an attribute, weighted points (either positive or negative) are assigned to each attribute value in the range and the sum of those points equals the final credit score.

Scorecard CriteriaRangePoints
AgeUp to 2510
26 to 4025
41 to 6538
66 and up43
IncomeUp to 20k-10
21k to 40k16
41k to 70k28
71k and up45
Bureau ScoreUp to 300-25
300 - 5000
500 - 65030
650 - 75050
Total ScoreSum of Points

Figure 2. Standard Scorecard Format

Would you like to discuss your requirements or arrange a demo?