| |
|
|
|
|
|
|
|
| |
Financial & Credit Services Companies |
|
| |
 |
Operational Data Store (ODS) Process
|
|
|
| |
| |
Projects
|
|
| |
Project Overview: Operational Data Store
|
|
| |
Client: A Major Financial Services Company
|
|
| |
Challenge
Due to the expansion of various business intelligence initiatives, application teams of different operational systems were overloaded with many similar operational data requests to feed the new analytic systems and enterprise reporting systems. The data channels established between data publishers (operational systems) and data subscribers affected both the system performance and the network performance of the mission critical operational systems. The problem was that they were still unable to deliver just-in-time transactional data to support ‘decision critical’ analytic business intelligence systems.
|
|
| |
Solution
Client management recognized the direct exchange of data between systems must be changed to a Hub-and-Spoke approach. AD/S worked with the client to establish a central operational data store and a management portal known as the Data Hub Dashboard to realize the benefits of a Hub-and-Spoke architecture. Raw data is acquired into the Data Hub and it is then cleaned and transformed into the Operational Data Store. The subscription data is then built and fanned out to analytic systems. This fan-in and fan-out approach reduces network traffic and leverages efficiency by having a team of specialists to manage this transformation hub. In order to avoid the pitfall of performance bottleneck and single point of failure of a central repository, a high performance parallel tool is selected to ensure transformation efficiency, with clustered hardware being used for its scalability and its failover capability.
|
| |
Tools
Ab Initio
IBM Sequent NUMAQ
Storage Area Network
XML and XSLT
Java J2EE
|
| |
Results
- This scalable solution acquires and transforms over 300 GB of data daily.
- Reduces the cost and CPU cycle from IBM mainframe host
- Network bandwidth is saved because data is sent from the source system once and transformed output is stored into storage area network.
- The Data Hub Dashboard portal allows users to monitor and track transformation process and view metadata defined for data elements.
- With the added step for data cleaning and data validation performed before loading data into the Operational Data Store, data quality is better controlled.
- Redundant data archives are eliminated and historical data is managed reliably and consistently.
|
| |
Project Overview: Customer Data Warehouse
|
|
| |
Client: A Major Credit Services Company
|
|
| |
Challenge
In order to better profile customers and generate reliable customer risk scores, the client decided to build one of the largest customer data warehouses in the nation. This data warehouse holds over 60 million customer’s information and over 6 million daily transactions. Third party data sources, such as credit bureau data, was purchased and linked to provide a comprehensive view of each customer for even better data mining and other cross-sell opportunities.
|
|
| |
Solution
AD/S worked with the client’s staff, software vendors and the hardware vendor to design and establish the hardware and software infrastructure for this customer data warehouse. In order to achieve the highest scalability, an IBM UDB database was chosen to run on massive parallel processors so that data can be partitioned across multiple node server clusters. AD/S assisted the client to build a robust and high performance ETL environment to acquire facts and dimensions data daily from the operational systems. AD/S also designed and integrated a checkpoint restart procedure throughout the extraction, transformation, and load stages. In each checkpoint, data is validated against the financial system to ensure warehouse data does not deviate from the financial system more than the defined error threshold percentages. Once data is cleansed and validated, it is staged and scheduled to be loaded weekly into the data warehouse. This solution supports several hundred concurrent users accessing data using the OLAP reporting tool.
|
| |
Tools
IBM DB2 Universal Database, Extended Enterprise Edition
IBM SP2 MPP Multi-Node Server Clusters
Customer Relationship Management Software
Micro Strategy 7i
Java J2EE
|
| |
Results
- The system represents the core of the
organization's customer acquisition strategy.
- The scalable solution supports hundreds of
concurrent users running complex queries against a multi-terabyte
data warehouse.
- The customer data warehouse is one of the
largest customer acquisition solution in operation within the
financial services industry.
- The checkpoint restart procedure
demonstrated the data quality to client management from the first
day the warehouse was deployed.
- This customer data warehouse provides a
solid platform to launch the Customer Relationship Management system
and provide data for risk score building and analysis.
|
