Minimise your risk and drive your credit decisions with CompuScore
What is CompuScore?
CompuScore is Compuscan's advanced credit bureau score which provides you with a score or number indicating the level of risk associated with an individual, plus the reasons for decline, based on detailed credit bureau data, variables & weights.
The Business Challenge:
You, as a credit provider, need an effective tool which consolidates all credit information and delivers a simple and intelligent guide to assess the credit risk of new and existing customers.
Our Solution:
Compuscan's highly predictive credit bureau score, CompuScore, provides you with a complete view of new applicants and existing customers to assist in the effective assessment of credit risk. By using CompuScore you can decrease your credit risk, increase your customer retention and have the ability to come up with risk based lending strategies.
The Score Explained:
CompuScore is available to you as part of an individual credit report online or you can get the score information via our batch services. Please see Figure 1 below for an illustration of how the score is displayed on an individual credit report. It provides you with a score or number, typically between 0 and 900, indicating the level of risk associated with an individual credit applicant.
A high score indicates low risk, while a low score indicates high risk. The score slots borrowers into groups which indicate the likelihood of a loan being repaid The score is prominently displayed in the top left side of the image and adjacent thereto, the risk level in which the score falls is displayed. There is five risk levels and corresponding score ranges.
The black bar indicates the actual score. At the bottom of the image are the reasons for decline, based on a specific set of bureau rules to give you, an even better indication of the particular risk level for a specific credit applicant.
The Declined Reasons that is generated by CompuScore includes the following:
Negative Information
Poor Payment Behaviour for Previous Loans
Poor Behaviour reported by Branch
Limited Loans History
Poor Recent Behaviour on Previous Loans
Poor Recent Behaviour on Accounts
Limited Accounts History
Limited History
How Was It Developed?
CompuScore was developed using statistical techniques, analysis and the historical credit information sourced from data on Compuscan’s credit bureau database.
It was developed across various loan terms based on a sample of over 500,000 consumers. Multiple types of data were utilised including Credit Providers' Association (CPA) information and Compuscan’s micro lending specific credit information. Good and bad repayment behavioural trends were modeled, using the latest statistical techniques, which led to the development of a set of highly predictive scores that can be applied across all your loan products.
Why Calibrate CompuScore To Your Data?
In order to derive the maximum power of the score, we would recommend that the CompuScore is calibrated to your data, utilising your performance definition and portfolio trends such as bad rate and acceptance rate. This will enable us to provide accurate prediction of your future acceptance rates and associated bad rates at specific scores, as well as in assisting in the selection of the best loan product to offer your client. In addition, CompuScore can also be complimented with additional bureau rules to ensure that you acquire profitable business.
How Does The Data Calibration Process Work?
The purpose of this exercise it to analyse the results of your previous applications and tune the Compuscan bureau score that will identify applicants that are more likely to repay their obligations. Upon completion, a report will be produced of our findings, and a workshop will be held with the client, where proposed score parameters will be agreed to.
The score parameters (cutt offs) will then be tuned to your business requirements, allowing you to use these parameters within your decision systems. If using Compuscan’s decision software then these parameters will be coded into our decision engine, Codix, so as to automate the approval decisions by product. The Codix engine will enact the parameters at the point of enquiry online thereby allowing the client to make the appropriate decision.
Extract File Considerations:
In order to complete the calibration exercise, we will require a retrospective sample of accounts that have been booked, and where possible accounts that have been rejected. Of those accounts opened, we will require information that will identify the customer as a good or bad account. When providing the sample, the following key points must be borne in mind: When providing the sample, the following key points must be kept in mind:
The sample must be a mature one, where applicants that are likely to go bad have already gone bad, and those that are likely to be good have a repayment window long enough to indicate they are likely to repay the full loan. As such, we recommend that the sample is taken from applications made approximately 12-18 months.
To deliver a predictive result, a large enough sample of applications are required to provide results that are not impacted by anomalies on individual applications. Therefore, a minimum of 2,000 good accounts and 2,000 bad accounts are required, but the client should supply as large a sample as possible, particularly if the score is to be tested across different product types.
In order to ensure the sample is not subject to temporary fluctuations such as additional promotions, seasonality etc, it is important to spread the sample of applications across a number of loans over a period of time, geographical area, etc. Therefore, we suggest that a random sample across multiple, similar products is extracted.
If you have stored your previously rejected applications, minimum of 1500 rejects should be supplied. This will allow us to perform reject inference so as to allow us to select previously good accounts that were rejected.
Should you require any further information on Codix, please do not hesitate to contact us at Tel: 021 888 6000 or e-mail us at info@compuscan.co.za
