Quelle: CIO Asia
In 2000, when David Malligan joined Bank Rakyat Indonesia (BRI), Indonesia's largest bank, as its CIO, the Australian native found himself needing to make some major changes. For years, BRI's core banking system had been decentralised: each of its 4,500-plus branches had its own server, did individual batch processing and kept separate databases.
Malligan, whose career in the Indonesian banking industry spanned 10 years, was hardly surprised at this nightmarish degree of decentralisation: after all, BRI was Asia's biggest bank outside of China and India, with 40 million accounts and a network of 12 regional offices, 323 branches, 64 sub-branches, 3,916 BRI Units and 238 Village Service Posts. Since its inception in 1895, the bank has gained profitability by providing loans and financial services to micro, small and medium enterprises in Indonesia. In 2002, loans to this market made up 90 percent of business.
The mechanics of how these loans are meted out does not worry Malligan: the data behind them is what has been keeping him busy. The bank recorded tremendous amounts of data locked up in separate databases from Progress Software Corp., running on Intel-based Unix operating systems. A sample of this data, which cover BRI's business lines and markets, include reports on weekly foreign currency positions, due dates of term deposits and funds available in the bank. They are used by divisions to present the bank's consolidated position and analyse its profitability. "The fragmented nature of our available data made it difficult to analyse the profitability of our business lines and verify if targets, say in lending, have been met. Having consolidated data is key to managing the bank's total portfolio effectively and efficiently. Better quality data would also improve reporting and our understanding of the bank's position," he says. The prior database systems churned out paper reports that left details out and could not be read online. User divisions that wanted reports had to go through long processes to acquire them because the branches operated on different systems than theirs.
In 2000, therefore, the bank set out to change all that. They consolidated their databases into a centralised, core banking system from Malaysian-based software firm Silverlake System Sdn Bhd, running on IBM's AS/400 and DB2 database platforms, with the data warehousing application residing on Microsoft's SQL database. The project was executed in two phases. The first was the implementation of the core banking system, which was up and running at the first BRI branch in January 2001. Currently, all branches have rolled it out. Phase two, which started in 2002, involved taking data from the Silverlake system and depositing it in the Microsoft data warehouse. By December 2002, 100 reports were in production, and now, all reports are accessible to divisions online.
Malligan says that centralisation has changed data gathering processes. "This project created a single source of data so that everyone looks at the same figures. We have achieved statistical consistency, gotten a better understanding of the profit-and-loss picture, and can more confidently manage the information we have. Before, business divisions had different data sources, and branches were making their own Excel reports and sending them to the divisions--so each one ended up having different versions of the same data."
Fixin' it
BRI hired Silverlake, whose core banking software is used in about 100 banks in Southeast Asia, not only to provide software but also suggest how BRI might go about implementing its data warehouse. Half a dozen Silverlake data warehouse experts worked onsite at BRI to confirm business requirements and processes with users, develop software, test it and get it up and running. Earlier in the year, BRI had hired Accenture to draw up a blueprint for management reporting within the bank, and the new data warehouse system had to conform to this plan. Consultants from Microsoft took care of designing the data warehousing cube.
Over the next few months, consultants worked to extract data from the old system, clean it and move it to a HP clustered 8-processor server running on Microsoft Windows Server 2000 32-bit operating system and SQL 7 database. The backend is a 33-TB EMC Symmetrix enterprise storage system.
Recently, BRI also hired Microsoft Services to improve the performance of the warehousing component of its solution, as it had begun to fall short of their standards. Explains Irwan Tirtariyadi, director, Microsoft Indonesia: "Their pain point was that users--business managers accessing the data warehouse for ad hoc reports--had to wait about one to 10 minutes to access certain reports. That was really slowing productivity down." After understanding the pain points raised by BRI, Microsoft, which recently launched its 64-bit Windows 2003 server, made the recommendation that the 64-bit environment be used for the data warehousing engine, which was subsequently ported to Windows 2003 64-bit and SQL 64-bit. "The previous 32-bit Microsoft platform was sufficient when BRI first implemented its data warehousing system, but the company had been growing very fast--their latest data store stood at 2 terabytes--and is growing in the order of 300-400 gigabytes a month," says Tirtariyadi. "And with this increasing data size and complexity of environment, they needed to move to 64-bit."
The benefits of the 64-bit environment have emerged even during the testing and development phase. "We have seen a 50 percent improvement in transaction time for end-of-month processing and SQL queries," says Malligan. In the 32-bit environment, business managers had to wait one or two minutes to access certain reports; that has come down to seconds. Before, it took eight hours to consolidate monthly reports; with 64-bit, it takes three hours.
Because 64-bit is still new technology, challenges were faced in porting the data warehousing cube from the 32- to 64-bit environment. Tirtariyadi: "Multi-vendor coordination was critical as a number of hardware vendors were involved: Intel and HP--because BRI's latest hardware is based on Itanium 2--and EMC--because BRI recently employed their storage area network."
BRI is also using Microsoft Services to provide facility management, outsourced services, HR and capacity to run its data operations. Within the umbrella of facility management, Microsoft Services is providing BRI 'break-and-fix' services. Value-added services include checks on the health of BRI's IT environment, training for staff and updates on the latest Microsoft technologies and their road map to allow BRI to align adoption with business need. Never resting on its laurels, BRI has also signed up Microsoft to perform a technical review of the performance of its data warehousing software to examine how it can run more efficiently. The full-scale review will extend beyond technical porting to architecture and will involve revisiting cube design.
Planning the birth
Before creating their data warehouse, Malligan's team was particular about addressing users' concerns--by involving them in the project. User requirements, going down to the desired layout and data structure and fields for reports, were confirmed, documented and signed off. Users hail from every division of the bank--treasury, credit, business, retail, micro, middle-market, agribusiness, corporate banking, audit and compliance. "We had a project steering committee--with representation from two directors of the bank--and project working teams. Each division assigned representatives to work on the programme. We had regular meetings to make sure work was on track and requirements were met."
BRI put its finance and accounting division in charge of the programme. They had to coordinate with 28 divisions to find out their detailed requirements. "Out of that, we found that we needed 500 different reports. Upon reviewing the requirements, however, we realised that a lot of divisions were asking for the same data but in different formats, so we were able to reduce the reports from 500 to135." The second step was training users to access the SQL database. "The advantage of that is to make them more comfortable with cube and SQL technology and take on the responsibility of generating reports in different formats instead of relying on IT to do so."
A data warehouse is born
Users have been unanimous that the system has made their jobs easier. For instance, now that it has a data warehouse, the bank's administration and credit division can look at daily online reports showing, say, the top 200 debtors, or customers with overdue or non-performing loans. "We can look at data online, which is great. But the greater benefit is the improved quality and detail of the data--which allows us to make more informed decisions," says Malligan.
He adds: "This effort has enhanced the perception of IS. We wanted IS to get away from the tradition of producing management reports and owning data, which should not be part of our business or operations. IS's job is to provide the infrastructure and bandwidth connecting the systems, not the reports. Now that data from the Silverlake system [that runs on AS/400] is uploaded automatically to the data warehouse [that runs on Intel servers], no intervention is needed from IT to tweak any data."
Stopping the pain
The pain felt by BRI caused by data silos was a familiar one to Indonesia's fifth largest insurer, PT Asuransi Allianz Utama Indonesia (Allianz Indonesia). A joint venture with German insurance giant Allianz Aktiengesellschaft, the company, which provides a range of personal and corporate insurance products, operates seven branches and employs 200 staff.
In the late 90s, when Indonesian banks that faced difficulty in expanding started diversifying into insurance and brought in new products and new approaches, Allianz knew it was time to consider how to ensure its continued success in the face of increasing competition in an already dynamic industry. Recalls CIO Erwin Purwadarma: "It was difficult to innovate with insurance products. Take away price as a differentiator and you are left with customer experience as a critical way to stand out. If we could use the information we hold about customers' policies and their needs, we could improve our service to them."
Over the years, Allianz had collected loads of information about its customers but lacked the tools to use them to make better decisions. Purwadarma adds that an unexpected event like a large claim, or a request for an unusual type of insurance from a broker, could send these critical customers in search of a speedier response.
Unusual requests also exposed it to risks. It lacked the ability to quickly assess clients' past insurance histories and decide on suitable premiums as the information delivered to its customer service staff was not accurate or up-to-date. Since it did not know its customers' claim ratio, for example, it could be worried that by trying to serve them well it would be exposing itself to bad risks.
Too much data
To overcome these problems Allianz knew it needed to implement business intelligence (BI). Purwadarma was given a team of three people and a very short deadline. "I was told we needed to see results in one month." So he pioneered with Microsoft Consulting Services (MCS) and Microsoft Premier Support Services (PSS) to build a strategic prototype of BI using Microsoft technology from design up to implementation.
The project was started in early 1997 and finalised in end 1998, with continuous upgrades and major improvements in 2000, when the solution was migrated from Microsoft SQL 6.5 to Microsoft SQL 2000, combined with the Analysis Server 2000 and integrated with Microsoft SharePoint Portal 2000, to provide BI in real-time mode. It cost less thanUS$10,000 and took two weeks, to implement BI.
Allianz had already adopted the Microsoft Windows 2000 platform for mission-critical operations and was satisfied it could meet the needs of its BI project. Purwadarma and his team used Microsoft Data Analyzer 3.5 to analyse data sources required to improve customer services, and then drew them into a single SQL Server 2000 database. Data sources include agent and mobile applications, branch and broker connectivity, as well as front-end and back-end Oracle-based applications. With the data collected in the new, single database, OLAP cubes were created to perform advanced analysis of the contents.
To put the information extracted through this process in the hands of the Allianz Indonesia team and customers, the company also implemented SharePoint Portal Server. This takes information from the SQL server and publishes it to an intranet for Allianz Indonesia staff, and to a secure extranet for brokers and agents.
Making it worthwhile
"We used to have time pressures and difficulties gathering all the information on a particular client, but now we have much more information throughout the company and can move faster. We have also transformed our customer service team from normal into knowledge workers. Now that we have BI accessible across all our departments, our people can instantly view a detailed profile of our clients. This helps us to make better decisions, which is good for our business and also meets client needs," says Purwadarma.
He adds: "We can now group customers by which ones are profitable and which are not, by location, or by the products they use. This information is being used by our marketing and claim departments and also the accounts team to change the products we offer, the premiums we charge and they way we conduct our business," he adds.
In 2003, Allianz Utama Indonesia was evaluated the best general insurance company in Indonesia, for 2003, by Bisnis Indonesia magazine.
The new platform has also been faster, reducing processing time from 8to 10 hours, down to only less than an hour in most batch processes, to produce OLAP cubes.
Allianz Indonesia plans to make its BI systems a critical strategic component of its business planning. "Today we produce and distribute first-class clean data. Tomorrow we will base our strategies on the first-class knowledge our BI solution produces," says Purwadarma.