References /

SAS Institute Inc.
Tel: +44 (0) 1628 490490
For more information on SAS, please visit


Printer-FriendlyPrinter-FriendlyPDF PDF

UK References


University of Southampton Uses SAS® to Help Credit Card Company Test Out Statistical Correlations

A leading credit card company has been making use of SAS software and the SAS® Enterprise Guide® solution, in particular, to monitor and control credit risk over many years. The business recently engaged with the University of Southampton on a project where SAS was used to test correlations between different statistical models built by the credit card company.

The credit card company's statisticians initially used SAS to build statistical models and carry out data analysis. Over recent years, this usage has significantly expanded. The majority of analysts around the business now have access to SAS, and much of the reporting and nearly all of the analysis that happens across the organisation takes place using the software.

As part of an ongoing programme of engagement with UK universities, the credit card company recently ran a project in conjunction with the University of Southampton, which itself has been running projects with businesses for around thirty years. A student, Melissa Franzetti, working on her MSc final project in Operational Research and Finance, spent three months with the credit card company working on a project to test the correlation between different types of statistical models.

Gillian Groom, industrial liaison officer, notes, “We actively look for external placements for all students on our  business analytics, operational research and management sciences Masters courses, where they can focus on tackling real business issues. Melissa’s project was the first we have carried out with this particular company, although we have worked with many other financial services organisations. Typically, they all look for SAS skills from the students to carry out the requisite data processing and data modelling – it does appear to be the tool of choice in this area.

"For the students themselves these projects form a key part of their Masters' degree courses," continues Groom. "The feedback we get from most students is that these projects are the hardest but the most rewarding part of their Masters' courses because they give them some tangible work experience that ultimately also helps them to secure a future long-term position."

For the credit card company, evaluating the level of correlation between different statistical models and then using this information proactively to drive business strategy has historically proved to be a significant challenge. The issue was clear. The company has historically often built two separate models: one predicting risk, analysing the likelihood of an individual missing payments; and a second predicting utility, in this case how likely an individual is to use their card. Experience indicates that these two models are likely to be highly-correlated. Typically, as risk increases, so does utilisation. Lower risk individuals tend to use their credit card less frequently.

This is an area of concern for any credit card company looking to effectively 'split out' populations. Clearly, the ideal is to find low risk populations that have a low possibility of defaulting, but at the same time are likely to use their cards extensively. The company naturally wants to maximise its profitability and to do so, it needs to target those customers best equipped to help drive that profitability.

So, the better the issuer can disaggregate these two models, the better the decisions it can make are likely to be. As Franzetti commented: "By minimising the correlation between the two scores, the company will be able to target those customers that are going to be high utilisers of their card for a given risk score."

Carrying out the Process

Franzetti used the credit card company's data to look at different statistical approaches to this problem and employed SAS solutions, principally SAS Enterprise Guide version 4.3, to carry out the analysis and then summarise the results at the end of the project. The student also utilised the various regression procedures within SAS.

The project lasted a total of three months. During the first, the student focused on exploration, a process that included data cleansing and data transformation. With this process completed, she used SAS to analyse the underlying data sets; to create statistical models and ultimately to create credit risk models, in particular.

The second month was focused on completing the analysis and the third concentrated on the final write-up and reporting. By making use of SAS, the student was able to achieve the desired results.

Delivering Results

"The work carried out on this project gave us a thorough overview of different approaches to this problem, together with a good understanding of all the different options available, while helping us to understand how we model these different options," says a spokesperson from the credit card company.

Franzetti was delighted with the opportunity to use the SAS solution, describing it as user-friendly and highlighting the opportunity to use sophisticated software solutions without being blinded by complexity. Underlining the success she achieved, she even won a university prize for the best SAS-based project. Having now left the university, she is confident that the experience of working with SAS will be a major benefit in her future career.

"First, it has enabled me to decide exactly what I wanted to do. Adding these skills has made me more employable," she said.

"I have a background in mathematics, I have my MSc but, at the same time, I am now proficient in the use of SAS," she continued. "Many of the companies I interviewed for were looking for these kinds of skills."

In her new job, Franzetti is now focusing on online fraud and using SAS on a daily basis in order to create codes that help prevent people from acting in a fraudulent manner.

Copyright © SAS Institute Inc. All Rights Reserved.

University of Southampton

Business Issue:
Test correlation between two different statistical models: risk – analysing likelihood of an individual missing payments; and utility – how likely an individual is to use credit card. Explore possibility of minimising correlation between the two scores.
SAS Enterprise Guide for model building and related analysis
Comprehensive report delivered. Overview of different approaches to problem provided, insight into different available options and how to model these options.

Adding these skills has made me more employable. I have a background in mathematics, I have my MSc but, at the same time, I am now proficient in the use of SAS. Many of the companies I interviewed for were looking for these kinds of skills.

Melissa Franzetti

Former student, The University of Southampton

Read more: