Credit attributes: Analytical decisioning building blocks

As building blocks to any predictive model development, it goes without saying that the better the predictive variables, also known as attributes, the more accurate the predictions. For many predictive analytic modeling shops, attributes are a key differentiator to model performance versus the competition.  It not only takes skilled analytical minds to create the attributes, but years of industry experience combined with in-depth knowledge of the data, and the ability to interpret and translate raw data into essential packets of information.

In the consumer credit world, credit attributes capture the essential elements of credit data to enable lenders to evaluate consumer credit worthiness.  The additional layer of complexity for lenders in utilizing credit data from all 3 major U.S. credit bureaus makes tri-bureau leveling of attributes an absolute necessity.  By leveling the attributes, the meaning of an attribute is the same across all 3 credit bureaus so that greater efficiencies in model development and decisioning can be achieved.  Imagine having to make decisions off of 3 models based on 3 different sets of attributes.  With tri-bureau leveling, only one model needs to be developed across one data set, resulting in consistent decisioning across all 3 credit bureau data.

Developing the best tri-bureau leveled credit attributes is important, but as important is attribute governance, which includes ensuring the continued integrity of the attributes. With the ever increasing regulatory scrutiny, it’s no wonder that lenders have identified attribute governance as a priority.  For instance, OCC Office of the Comptroller of the Currency’s (OCC) 2011-2012 Supervisory Guidance on Model Risk Management requires the annual validation of existing risk models as well as the validation of the attributes of custom models.

Whether you outsource or develop credit attributes in-house, here’s a high-level checklist to help you assess whether the development and management of your credit attributes is up to par given the data complexity and increasing regulatory requirements:

  • Keeps pace with industry trends

Does the attribute development and management team have industry and tri-bureau credit bureau data expertise and experience? 

  • Attribute governance

Can the attribute documentation stand up to regulatory and compliance audits? 

  • Stability analysis 

Do the attributes undergo a rigorous monthly protocol for testing and maintenance midst dynamic credit data and environment to ensure continued accuracy and stability of attributes across all data sets and feeds?

  • Tri-bureau leveling

Are the attributes tri-bureau leveled to enable the development of one model that can be run on one data set?

Credit attributes that continue to evolve and keep pace with industry trends are essential to model performance.  Attributes are also extremely valuable in segmentation, overlay to rules and policy definitions, which I will talk about in my next month blog.  Be sure to visit again!

 

Comments for Credit attributes: Analytical decisioning building blocks

blog comments powered by Disqus