The end of the year always provides an opportunity to reflect over the past year. Let's face it, 2007 has been a challenging year for the South African credit industry - enormous operational pressure to meet National Credit Act (NCA) compliance, a certain amount of intrepidation as we tested the new legislative landscape and its subsequent impact on credit spending.

There continues to be a great deal of debate on the impact of the NCA but it still remains difficult to determine its tangible impact as well as predict its effect over the coming months, and years. Certainly initial feedback from the South African Reserve Bank is that the new legislation has only marginally curbed growth in consumer spending.

To provide some insight into the South African consumer's financial health and spending over the last year, Experian conducted an analysis of its data over the period 1 October 2006 to 30 September 2007 to determine whether any significant trends are evident in the lead up to 1 June 2007, and thereafter.

A snapshot of consumer credit information

Generally, consumers' credit profiles appeared to improve following 1 June 2007 (Refer to Figure 1). Our analysis shows that the percentage of consumers with negative credit information has decreased from one quarter to the next and through the implementation of the NCA. This was primarily fuelled by the implementation of the first phase of the credit information amnesty where certain default and judgement information was removed from the credit bureaux. Consumers seem to be in a better position regarding credit activity and positive credit information.

Figure 1 : Consumer credit information per quarter

Impact on consumer spending

Our findings show a significant impact on the opening and closing of credit accounts in the lead up to 1 June 2007 (Refer to Figure 2). The analysis indicates a sharp increase in accounts closed (primarily of NLR accounts where over 30% of accounts were closed) in March 2007 compared to the previous month. As interest rates remained constant the impending NCA implementation is the likely cause. Similarly, the months preceding 1 June also showed a significant increase in new accounts – a likely result of aggressive marketing by credit providers prior to ‘1 June' as well as consumer fear that it would be difficult to open an account once the regulation came into effect.

Figure 2 : No. Accounts closed vs Accounts opened

Subsequently, there appears to be a slow down in new accounts being opened and also consumers maintaining existing accounts with fewer accounts being closed. Although it's likely that interest rate hikes and increased inflationary pressures have also impacted the slow down.

Figure 3 : Average overdue balance

Impact on negative credit information

In the months following the implementation of the NCA, our study shows an increase in outstanding payments (Refer to Figure 3) as well as in the average months in arrears. Again this is likely to be a result of interest rate hikes and inflation, as opposed to the effect of the NCA.

Figure 4: Percentage Negative account

As was anticipated, the analysis further shows a significant decrease in negative credit accounts on Experian's records following the credit information amnesty (Refer to Figure 4). This decrease was particularly significant in terms of the percentage of default deletions. The 1st phase of the credit information amnesty (implemented 1 June 2007) resulted in over 16 million records being removed from our records, benefiting more than 6 million consumers.

Furthermore, the deadline for paid up judgement listings of up to R50 000, granted before 1 September 2006, has been extended to 31 December 2007 which is likely to further impact negative information on our records.

Conclusion

Perhaps not as drastic as may have initially been predicted the impact of the National Credit Act, interest rate hikes and increased inflation is impacting South African consumers. There appears to be a gradual slow down in the boom of consumer credit spending that we've experienced over the last few years.

	Experian Decision Analytics, part of Experian, the global information services company, announces that Toyota Bank, which specializes in car loans and is owned by Toyota Motor, is to automate making credit decisions and carrying out applicant’s analysis with an Experian Decision Analytics’ AcquireSM solution and a car loan generic scorecard.

The decisioning systems will enable Toyota Bank to process applications in real time in order to quickly assess the risk level of applicants and provide customers with an immediate decision in an efficient and cost-effective way.

The generic car loan scorecard will enable Toyota Bank to accurately assess the risk of car loan applicants and make quick decisions while identifying optimal customers. Along with AcquireSM, this solution improves operational efficiency by enabling the automation of manual processes which results in a significant cost reduction. It improves decisioning quality through combining application data with comprehensive information from external and internal sources, thus providing a complete profile of every applicant. It also helps increase profitability by maximizing revenue, reducing referral rates and minimizing risk.

AcquireSM is a web-based application processing solution that allows an organisation to implement the solution at a significantly lower cost compared to deploying software. The system offers the functionality of the Experian-Interfax credit bureau which stores over 3 million credit histories and works with over 190 financial organizations, including major banks operating in consumer finance. Thus, when a credit application is processed by AcquireSM and a score is calculated, the system uses the records of both the bank and the credit bureau.

Dr. Olaf Neitzsch, President of Toyota Bank, comments: “Over 72,000 Toyota and almost 7,000 Lexus cars have been sold in the first half of 2007 in Russia, 30-40% of them on credit. We are set to achieve leading positions in financing their sales in Russia.”

“Toyota Bank already has its first five points of sale located in Toyota and Lexus authorized dealerships in Moscow. Launches of 16 points of sales in all Moscow-based authorized dealerships of Toyota and Lexus are planned for 2007 – early 2008. In 2008 Toyota Bank services will be available in Saint-Petersburg as well. Regional markets are part of further development plans. At the moment the bank product line includes car loans for new Toyota and Lexus vehicles. In nearest future ZAO Toyota Bank plans to extend its product line, develop new products stepwise, since high quality of own services is top priority for the Bank.”

Daniel Zelenski, Head of Experian Decision Analytics Russian office, says: “We are delighted that Toyota Bank has become our customer in Russia. Russia today is a particularly promising sales market and production location. The past year was record-setting for Russian car market in terms of growth rates: sales of foreign models increased by 280,000 units in 2006, which represents a 100% increase year-on-year. Sales of new imported cars stood at 720,000 units, a 76% increase as compared to 2005*. We at Experian Decision Analytics are confident that our advanced solutions will ensure Toyota Bank’s successful launch in this country and will provide for further competitive strengths and expansion of their business across this exponentially growing marketplace.”

	This article is the third in a series relating to scorecard development and types of scoring models, and focuses on the Split Analysis. The first article touched on the concepts of scoring and the different types of scorecards. The Masterfile Analysis; which is the preparation required before building a scorecard was the focus of the second article.

The purpose of the Split Analysis is to establish the potential benefits for the business of developing multiple scorecards to control exposure to bad debt across the portfolio.

Split analysis is normally done to separate areas of business so they can be treated differently. It can be done, for example by products, source of business, characteristics, new vs. an existing customer, new vs. used vehicles etc.

Reasons for the splits can be ;

•  Historical – used to split in the past or it just feels right.

•  Practical – business using separate systems or

•  Statistical – more powerful assessments, improved decisions (debt reduction/more revenue/more profit)

It is important to consider the relative size of each sub population when identifying optimal splits. It is recommended that sub population splits should account for at least 15% or more of the total portfolio size unless this is of sizeable volume. This ensures that the scorecards are statistically sound and will be robust over time. In addition, it may not be cost effective to develop and implement scorecards for small sub-populations.

It is important to verify that there is a significant difference in the risk level of each sub-population within each potential split. This ensures that it is appropriate to have different scorecards for each sub-population.

As mentioned, the objective of Split Analysis is to establish whether or not there is any benefit, in terms of improved scorecard discrimination and/or bad debt savings, in having more than one scorecard. This is achieved via the following process:

A database is created containing all information available as at time of application. All analyses conducted at the Split Analysis stage is prior to obtaining retrospective bureau information.

An overall scorecard is built on the entire database and scorecard performance measures are identified. These are:

•  Discrimination – which measures the separation between 'good' and 'bad' customers by taking into account the difference in the mean scores and the amount of overlap of the goods and bads. The bigger the difference in mean scores and the less the overlap, the bigger the discrimination. Discrimination is normally between 0 and 5000.

•  Gini – The Gini coefficient measures the proportion of bad customers accepted in relation to the proportion of good customers accepted. The Gini can range from 0 to 100 (100 is optimum, though unachievable in practice).

•  KS Statistic – The KS (Kolmogolov – Smirnov) statistic measures the maximum difference between the cumulative percentage of good customers and bad customers. The KS statistic can range from 0 to 100 (100 is optimum, though is unachievable in practice).

•  Bad Rates – The bad rate at various cut-off scores is noted.

Separate databases are then created for each of the splits and the criteria within each database are re-looked, to optimise the strength of the data for that particular sub-population.

Specific scorecards are built for each of the splits. Scorecard strength statistics are produced for each of the split scorecards.

The scorecard strength statistics, resulting bad rate and starter rate for the overall and split scorecards are then compared. Then the final scorecards are chosen.

Read next months Credinews for more information on building and testing a scorecard.

Written by Ben Maseko, Scoring Analyst, Experian Decision Analytics.

Applying the principles of Business Intelligence to improve revenue collections performance

	Online sales have low overheads and are a cheap way to acquire a customer. As there is no commission to pay to a third party, there is a quick return on the handset investment and profit is seen considerably earlier than through indirect acquisition channels.

Therefore, there is an obvious appetite to grow online sales in the current marketplace of falling margins. Coupled with this, customers now demand an instant decision when applying online – they no longer accept that web orders may be batched up and processed downstream. To remain competitive we need to provide a real-time decision. This was achieved during 2006.

Challenges

Distance selling, together with a real-time automated credit decision, creates issues in the removal of any form of human interaction between vendor and customer. This results in obvious fraud patterns not being picked up as quickly as they may have been with individuals manually processing orders. Therefore fraud applications can slip through the net if efficient controls are not in place.

Almost as soon as the credit decision became real-time, the fraud community became aware of this and attacked it. These are very well organised fraud rings which always attack the ‘weakest link' in the chain and are constantly evolving and changing tactics in order to bypass the corrective actions taken. In addition to the increase in volume of fraud we also noticed a shift in the type of fraud. Traditionally, it had tended to be subscription fraud, but we now saw the latest ‘desirable' handsets being targeted for export to Eastern Europe or Africa, which was largely through impersonation fraud, mirroring the trend seen throughout the credit industry over the last 18-24 months.

Impersonation is so much more difficult to detect at the point of application. Whereas previous scoring routines were able to indicate propensity for subscription fraud, ID theft is harder to determine without a resource intensive manual review – something which is neither practical nor acceptable to apply to all applications.

Mitigation

At Orange Credit Risk Management, we use a range of fraud prevention measures from Experian Decision Analytics, comprising both generic products and more bespoke solutions developed in association, specifically for our requirements.

The primary fraud detection tool we use is Detect. We worked with Experian to develop and enhance the detection rules for the online channel, which proved to be more predictive and produce fewer false-positive matches. The system incorporates the Detect Credit Score, which provides a risk assessment of the applicant as well as indicating the likelihood of the application being fraudulent. This enables the automation of decisions and reduces the resources required within the underwriting department.

At Orange, we also use a bureau-based scoring service, which gives a confidence level of the applicant's identity being valid. The tool has proved invaluable in highlighting high risk applications and protects the business from identity fraud.

Analysis demonstrated that a large proportion of impersonation fraud was committed as company directors rather than individual identities (as company information is held in the public domain and is easily accessible). Experian's systems accessed the Directors Database as part of our process flow, and provided a flag to state that an applicant through the web was also a company director, allowing for a manual referral policy rule to be created around this.

All of these products are deployed within the LinkSM Orange Strategy Management Generation 3 application processing system. Using this for front-end credit decisions has been extremely powerful – it brings all the above elements together and is very flexible, which is vital as the rules for fraud detection have to be continually updated as fraud trends evolve. As each new measure is implemented, it has an immediate impact on the fraud rate which is gradually eroded as the fraudsters change strategy. The flexibility of the solution means that, typically, changes can be made within 72 hours from concept to implementation, allowing us to react quickly to a dynamic fraud threat and counter new threats. The result is that the use of Experian Decision Analytics products has helped reduce fraud losses by 88%, saving millions of pounds.

The Future

In addition to our internal Orange database of known frauds called by the Orange LinkSM system, we are currently working with other UK networks and credit bureaux to establish a Telco industry fraud sharing database.

The power of sharing data comes from the increased level of information available for fraud checking, and the sharing of fraud information, means that the fraudster has nowhere to go, protecting the whole industry from this crime.

Melvyn Prescott – Senior Credit Analyst, Acquisition Management, Orange UK

The following career opportunities are available at Experian: Development Manager - The successful applicant will need to manage and lead a team of developers and analysts focused on the design, planning and implementation of software information systems to support the vision and strategy of Experian. Ensure maintenance of existing software is performed in a timely manner to ensure maximum uptime and return on investment. Click here for more information.

Legal Compliance Officer - The purpose of this job is to provide an administrative and legal and compliance service to both the Experian Decision Analytics and Experian Credit Services businesses under the direction of the Commercial Manager.Click here for more information.

Key Account Manager - The Key Account Manager is responsible for managing the existing client relationship and must have a sound understanding of the developments within the client and ensure timeous delivery of initiatives by Experian which are well within budget and in line with client expectations. Click here for more information.

Management Accountant - The purpose of this job is to provide an accounting and analytical function to the Experian Decision Analytics (EDA) line of business revenue and all related matters. Click here for more information.