| Scorecards – development, use and importance |
 |
 |
 |
| Friday, 21 May 2010 02:00 |
|
Scorecards are widely used models in various markets especially the financial market. The concept of a scorecard is simple. A scorecard is a model that relates a set of variables or attributes to an outcome. For example, a typical demographic scorecard for the assessment of credit worthiness will consist of variables such as marital status, income, age, years employed, years at home address, etc. The output of the scorecard is the probability of the individual honouring his/her credit commitments or defaulting on them. The variables of a scorecard are selected through a rigours statistical process in relation to the desired outcome. The statistical process also assigns specific scores or weights to each value of each variable. The more a variable and its values are capable of separating the values of the outcome, the more weight it gets. For example, if we consider years employed with the outcome of a Good or Bad credit worthiness, we may have the values of 2 years or less employed having low scores (higher probability of being Bad) and values of 5 and more years employed having high scores (lower probability of being Bad). The relationship between the variables and the outcome may be weak or strong depending on the nature of the variable and its relevance to the outcome, the correctness of its values and the completeness of its values in relation to the outcome. Many organisations only come to realise how poor their data capturing is when they attempt to use the data to develop scorecards. The data analysis and preparation is a major and fundamental task in scorecard development. In many cases, this task may consume 70% to 80% of the time required to build a scorecard. This task involves the understanding of the data, how it was captured, how it is populated, how it is related to the outcome, etc. These tasks are required to be carried out by a qualified analyst and may require the processing of thousands or sometimes millions of records. Another related task to data analysis is the transformation of data from one form to another and the creation of new variables from other variables. A simple example of this is creating ratios from two numerical variables or counting specific incidents within a time dimension. The task of transformation and creation of variables is critical to the development of a good scorecard. When the scorecard is used, the scorecard variables are evaluated and the specific scores of each variable are added together to create the overall or final score. The range of final scores is usually grouped into score bands. Each score band is characterised by a certain probability of the outcome and a certain percentage of records or cases falling into this band. It is very important to understand that a score of a single case does not mean much on its own. The score or a score band is only meaningful when applied to a group of cases. So for example if we have a score band with a 10% bad rate, then if we have 100 cases falling in that case, we will expect 10 of these cases to eventually turn bad. At the time of scoring, we will never be able to identify which case or application will turn Bad or Good.As the probability of a certain outcome (such as Bad credit risk) becomes higher within the low scoring bands, we will reach a level where there are too many Bads expected in that band enough to make us decide to decline all the cases or applications falling into that band. In other words, we are now sacrificing the few goods that may be in the band to avoid accepting the many Bads in that band. Scorecards are extensively used in the credit industry to assess credit risk at the time of application. It was proven in numerous case studies that the accuracy of a properly developed scorecard exceeds human judgments even by experience credit managers. Although the common usage of scorecards is in assessing applications for credit, scorecards are used in many other areas within the credit industry such as the assessment of profitability, attrition, fraud, responsiveness, etc. Many organizations are relying heavily on the utilization of scorecards because scorecards ensure: • consistency in assessment • Accuracy of assessment • Stability of model • Automation • Speed of usage • Ease of implementation • Ease of use • Ease of monitoring |
| Last Updated on Friday, 21 May 2010 11:47 |