The Data Analytics Blog

Our news and views relating to Data Analytics, Big Data, Machine Learning, and the world of Credit.

All Posts

How To Adopt Credit Scoring

May 2, 2019 at 8:15 AM

Bringing automation into the credit assessment process through credit scoring brings about significant benefits. Some of these benefits include:

  • Consistent objective decisions,
  • Quicker decisions,
  • A lesser requirement of manual intervention
  • More profitable decisions (normally 20-30% better than purely subjective decisioning)

The scorecard is the mathematical tool used to bring about these benefits. In terms of introducing advanced decisioning in the originations space, in particular, the following diagram shows (from left to right) the natural advancement of credit assessment through the use of different credit tools.
Natural Advancement of credit assessment through the use of different credit tools

For more information on retail scorecards, read our blog on What is a Scorecard? 

Key considerations when adopting scoring into your environment

There are many considerations that one should deliberate when adopting scoring into your business.  Most of them are discussed below:

  1. What type of scorecard – generic, expert or custom?
    1. Generic – this is an off-the-shelf scorecard. Credit consultancies such as Principa sell these scorecards.  They have been built based on a selection of other companies’ data and reflect the trends of data in predicting risk and are relevant for a particular product (asset finance, personal loans, mortgages, etc.), region (e.g. Sub-Saharan Africa), segment (e.g. low/high-income, locals/foreigners, young/old, etc.).  As they are off-the-shelf, they are not adjusted for your specific bank.
    2. Expert – these are scorecards built when there is little or no data, but usually a history of lending. These scorecards are crafted based on both other scorecards that were built elsewhere for a population that best reflect the profile of customer at the bank, and bank-specific information (gathered normally from a questionnaire and engagement between bank staff and consultancy)
    3. Custom – these are the strongest scorecards, and they are built based on the actual mathematical relationship between the data and its ability to predict risk. The key requirement is there is sufficient data:
      1. over a significant period – minimum two years typically
      2. a number of accounts – ideally at least 500 “good” and “bad” payers
      3. Significant enough number of characteristics – for originations scoring this is the data captured at application time. We typically look to use 10-15 predictive characteristics in a scorecard
  2. How are you going to use the scorecards (in originations)?
    1. Set the cut-off (this should be in-line with your risk appetite and also take into account all the other policy rules you wish to use). A cut-off will determine who you will automatically decline, who you will refer and who you will approve based on the scorecard.
    2. Risk-based-pricing – you may wish to create multiple cut-offs each reflecting a different risk-grade. At each risk grade tier, you can restrict the term, loan amount and price (interest rate) – all in line with your risk appetite.
  3. How you are going to deploy the scorecard – the scorecard should be deployed in scoring software or module that is-part-of or can be integrated into your originations work-flow. (Read our blog on the key properties of a good scoring engine)
  4. Should you use credit bureau scores? – Yes; if they are available. You should look to combine an application score with a credit bureau score for more precise decisioning.

Retail lending organisations tend to first encounter scoring in the originations space. However, scorecards can be and are used in other areas – most notably:

  1. Behaviour score for account management – for transactional accounts with facility/limits this could be used to regularly review your customers and to offer them a facility/limit in line with their risk profile.
  2. Basel II/III Regulatory capital requirements – for those adopting a Standardised or Internal-Ratings-Based (IRB) approach, scorecards are used to segment the book for “probability of default”. For very young accounts (less than four months on the book) the originations score is used; for mature accounts, behavioural scores are used.
  3. Behavioural score in collections – behaviour and collection scores can be used in collections to help collectors prioritise which accounts to collect on first.
  4. Accounting provisions (IFRS9) requirements – if account-level provisions are being calculated, then behavioural scores can be used for the calculation of the probability of default.

What is a behaviour score?

A behaviour score is a credit score representing the likelihood of an account defaulting based on its prior repayment history. Variables used in these scorecards include information about the age of the account, the number of missed payments over a period of time, the amount paid over a period of time in relation to the outstanding balance, etc. 

The B-score is normally deployed in a scoring engine, and a full batch is sent monthly from the host system into the scoring engine, and the resulting score sent back to the host system (or account management system).

The behavioural scoring process is depicted below.

Behavioural scoring

Behavioural scoring can also be used for repeat customers applying for another credit product.  Remember that Behavioural scores (if well built) are typically more predictive of credit risk than both the application score and the bureau score. You should, therefore, look to introduce behavioural scoring into your application assessment process if possible.

For more information about how credit scoring could benefit your business, please contact us.

Contact Us to Discuss Your Business Requirements

Thomas Maydon
Thomas Maydon
Thomas Maydon is the Head of Credit Solutions at Principa. With over 13 years of experience in the Southern African, West African and Middle Eastern retail credit markets, Tom has primarily been involved in consulting, analytics, credit bureau and predictive modelling services. He has experience in all aspects of the credit life cycle (in multiple industries) including intelligent prospecting, originations, strategy simulation, affordability analysis, behavioural modelling, pricing analysis, collections processes, and provisions (including Basel II) and profitability calculations.

Latest Posts

Mathematical Optimisation In Five Phases

For businesses wishing to improve their credit decisions, the adoption of Mathematical Optimisation is an important consideration. Mathematical optimisation is more than a straight data-driven strategy design as it incorporates prescriptive analytics.

Amazon Web Services Vs Microsoft Azure: Which Cloud Provider Should You Host Core Business Systems On?

Deciding on which cloud service to host your core business systems on can be a daunting task. Amazon Web Services (AWS) and Microsoft Azure are two of the biggest players around, while Google Cloud and IBM Cloud are also gaining market-share.

Truth Seeker: How To Avoid Logical Fallacies And Cognitive Biases In Data Science

We have released a new eBook titled Truth Seeker: a guide to avoiding logical fallacies and cognitive biases in data science.