CompuScore is the product of a joint venture between Compuscan credit bureau and ScoreSharp, an innovative risk management consultancy. CompuScore is a leading edge suite of credit bureau scores which predicts the probability of customer default throughout the credit life-cycle. In excess of 3,000 modelling variables, across 26 vertical market filters, were evaluated during the modelling process to ensure the most predictive bureau scores available. CompuScore summarises the key predictors from Compuscan’s extensive databases (CPA accounts, NLR accounts, Proprietary Loans, Enquiries, Collections, Administration Orders, Debt Restructure and Public Adverse) into a 3 digit score which can be readily integrated into credit risk and marketing decision points.

The CompuScore suite is currently comprised of three distinct score versions: CompuScore.A for risk assessment at the customer acquisition stage, CompuScore.B, a behavioural score for use in the on-going management of customers throughout their life-cycle and CompuScore.C for collection management. A supporting comprehensive library of 60 adverse reason codes ensures that credit grantors are assisted in complying with their obligations as stipulated within the National Credit Act.

CompuScore.A was developed to predict the probability of a serious default event occurring within 12 months of an account opening. A random sample of 1,000,000 customers who opened a CPA account and 1,000,000 customers who opened an NLR account, between November 2008 and October 2009, were used in the modelling. Out-of-time validations, performed on customers who opened accounts in November and December 2009, ensure the robustness and stability of the finished product.

CompuScore.B was developed to predict the on-going default risk in a 12 month outcome period. The development sample was created using customers having open CPA or NLR accounts at specific observation points. A random 1:50 sample of all consumers holding open CPA accounts and a random 1:15 sample of all consumers holding open NLR accounts were selected. Importantly, multiple observation points spanning a 12 month sample window were used to ensure that the possible impacts of seasonality were accommodated.

A consistent segmentation approach was used when developing both CompuScore.A and CompuScore.B, resulting in a total of 21 scorecards in the full model suite. All scorecards, regardless of version and segmentation level have been scaled identically to ensure the same risk to score relationship.

The adopted segmentation strategy filters customers based on the type of credit provider originating the enquiry. ScoreSharp has found that the predictive trends for key variables can be distinctly different for consumers applying to an NLR institution as opposed to those applying to a CPA institution.

Both the trade-off curves and default rate % by score band charts for CompuScore.A and CompuScore.B on their respective development samples are shown below. To maximise the benefits of implementing CompuScore, ScoreSharp will undertake a detailed customised calibration project, whereby the performance of the score suite on a credit granter’s own customers can be measured and suitable strategies defined.

CompuScore.C was developed to predict the probability of receiving payment on overdue accounts based on key performance indicators on the Compuscan credit bureau.

The score was tested against a hold-out sample to measure its performance. As an example, one of the score elements is the number of 0 arrears updates in the last 12 months. Customers who have at least one account which has been 0 in arrears have a higher likelihood of activation since this is an indication that the customer has funds which can be diverted to the handed over matter. In this case the higher the number of 0 updates, the higher the score.

Similarly, the number of satisfactory trades that have been opened for more than 12 months is used as a score element. A satisfactory trade is defined as being less than 3 months in arrears. Again this indicates that the customer does have available funds which could be used to pay the handed over account.

By placing CompuScore.C into probability bands and overlaying these bands with the hand over amount, we are able to set the priority on which matters should be targeted to yield the best return on collection effort.

For ease of interpretation we have created a matrix depicting the methodology used to set the priority. Please see Figure 2 for an illustration of this matrix. The hand over amount banding is customised depending on the product and industry for maximum effectiveness.

Please note: CompuScore.C is only available to CPA members or organizations with written agency agreement consent from the CPA member (their client) on whose behalf they are collecting.

To learn more about CompuScore ABC and  to arrange a customised evaluation please contact Compuscan at Tel: 021 888 6000 / e-mail at This e-mail address is being protected from spambots. You need JavaScript enabled to view it or ScoreSharp at Tel: 011 317 3822 / e-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it .