Credit Card

Problem:

he global financial crises that hit the world in 2008 halted the market especially the credit companies. It was expected that these companies would become more cautious to take risks in a year of uncertainties.   In the other hand, in case no action taken the Marketing and Sales certainly would be seriously affected.

Strategy and Methodology:

In the late of 2008, when the trends analysis started to indicate difficulties in reaching the goals established for the loans sales, the teams of our customer, a credit bureau and a technology provider company joined in the search of an alternative for the problem.

By using the bureau historical databases enriched with the client’s information, the bureau models development area designed custom-made models to predict not only the risk but also the propensity to consume through the combination of clustering and regression techniques. Several models were designed and exhaustively tested with the back test procedures developed by the technology provider company for the choice of the most efficient model in terms of profitability.

Performance was essential to that situation, and in February 2009 all the sales process had been modified into the new technology in order to minimize the impacts expected therefrom.

Results:

The annual sales goal was accomplished, and the total sales amounted to approximately R$ 7 million less than the total cost that would be incurred to sell the same amount of product without using the model. This result was calculated from the control groups used in each action.

The return rate was increased by around 300% in comparison to the return rate of the control population.

The first results allowed the continuation  of the level of sales of personal loans throughout the year, in contrast to some competitors who had to slow down the sales in the period.

Credit Services

Problem:
The Finance Market has been suffering great losses resulted from counterfeits mainly for Fraud Application, that is, frauds using fake or third parties documents. All prevention process has a high cost of checking since the process was manual and slow.

Strategy:
Perform a large number of validations of consistency and veracity of information when filling in the proposal in association with the validation of Federal Revenue Office’s information and external bureau's by making the process safe and fast.

Methodology:
Using the filling in system of existing proposals, the store employee accesses an online system comprising:
1)A Windows / Intel-platform databases including all information from customers and proposals history.
2)Smart comparison of names and addresses using proximity percentages between the values;
3)Mechanisms allowing consultations to several internal or external databases, and to Internet sites;
4)A decision engine configurable by the fraud prevention area to apply the fraud and validation rules.
The employee gets the proposal from the system after 2 to 5 seconds accompanying by the results of consistencies and information validations in external bureaus. Problematic proposals are taken out from normal processing.

Results were obtained both from frauds reduction and quality of information provided. When the system was started up, it was observed that 20% of proposals were rejected due to the employee's misspelling. An effort for the employees' awareness led to a substantial improvement in the quality of information filled in. A significative quantity of frauds is avoided by using the system, 12%, and in 35% of proposals the submittal of home address and telephone proofs are no longer required. The client was allowed to reduce the number of analysis and prevention positions, in addition to the savings from frauds being avoided.

Credit Bureau

Problem:

Bureaus information’s are of paramount importance to help award the credit. However, instantaneous portraits including restrictions only may be harmful for consumers once they cannot show their historical behavior, which leads to faulty assessment conclusions, thus limiting the business opportunities.

Strategy:

Organize the information collection on the consumer's behavior, thus allowing the construction of generic or custom-made indicators for those awarding credit, in order to prevent from losing this rich source of historical information.

Methodology:

Creation of a Windows/Intel-platform database including the qualified information on 100 million people. Generic credit scores representing the historical behavior of people were designed. Approximately 110 variables, which are provided for the design of custom-made scores by the credit award-companies were also designed. This has represented a market paradigm break considering that previously the bureaus did not use to provide the variables used in their scores. All these scores are recalculated on a daily basis, and made accessible for online consultations.

Our customer designed a new business using the processes based on this databases. This area has increasingly grown by starting with a first trial-based client and then leading to the current 30 clients in short time. Over 50 custom-made formulas have been developed with the use of combined variables and data of each institution. The first client states that by using this process, the sales goal of 2 million new customersclients was exceeded in 20%, and simultaneously reduce the credit losses in 50%.

Telecommunications Operator

To attend court orders, regulatory or customers complaints  for reissuance of old telephone bills is a complex process for telephone companies. Currently our customer has a database including over 60 billion registrations to be consolidated and combined for redesign the bills. In addition to the delay, the cost is also high: Storing files on backup tapes is expensive, and relies on the specific knowledge of the person in charge of storage for retrieval and use.

In searching for an alternative, StoneAge was contracted to develop a Storage and Recovery of Bills Project, a dataset structured with detailed information on telephone bills to attend not only the storage and recovery information needs, but also the need for more sophisticated data analysis.

Using extreme IQ, a quality and ETL tool designed by StoneAge, the files containing the lines of each telephone bill over the last 10 years were recovered. Even the registrations stored on tapes of unknown format were recovered and used...


Then, a dataset was developed by using the eXtreme database manager. These bases containing different visions of each line of the telephone bill over the last 10 years, allow the compact storage of information, and the efficient access, segmentation, filtering and recovery of any telephone bill in that period.
Finally, by using eXtreme OLAP the end users of information may not only recover old bills, but also analyze and carry out tasks of Business Intelligence on a dataset of information which had been considered lost, and difficult to work because of the volume of information.

Some benefits:

•   Compression of original volume of 100 Terabytes to 9 Terabytes.
•    Efficient storage of over 60 billion of registrations.
•    Savings of U$10 million in the hardware cost only.
•    Total elimination of delay in the information recoveries (data provided online).
•    Possible BI studies on the entire database.
•    Recovery of R$100 million resulted from the recovery process of paid taxes.

Return on investment of over 1,000%.