Why Fidelity Investment execs use IRI’s Express OLAP software to view financial data from many angles
By Gael Core, Open Computing
Gaining fresh insights about timely investment data can mean the difference between making or losing money. And streamlining senior managers’ access to that data translates into faster decision making . That’s why Fidelity Investments, a company with $506 billion in customer assets, selected a multidimensional database to crank up its lagging data analysis process. The company’s Boston-based IS group knew that traditional relational database management systems would be inefficient at running complex data analyses. Fidelity’s IS department wanted to provide its executives with quick and easy access to multidimensional data from more than 100 sources of the company’s data distributed all over the world.
In the early part of this year the group considered online analytical processing (OLAP) as a solution to the problem. OLAP databases, the much-touted new generation of technology heralded by data warehousing, excel in analysis capabilities, an area in which relational database management systems (RDBMS) are often weak. Specifically engineered for flexible analytical queries, OLAP databases use a spreadsheet approach for managing and storing the analyzed and summarized business information.
Fidelity has a customer base of more than 9 million investors. As the nation’s largest mutual-fund company, it controls and operates 223 mutual funds; it is also the nation’s second-largest discount brokerage firm. To handle all of its data and give managers what they require, the company’s IS department selected Information Resources Inc.’s (a subsidiary of Oracle Corp.) Express OLAP to create its new executive information system (EIS).
“We aren’t just using [Express] as an analysis tool, we are using it to collect a bunch of data and throw it on a screen in several different ways,” says Steve Rubinow, a vice president of corporate management information systems at Fidelity. “[Senior managers] are just interested in having the information they want at their fingertips and not all of them want to see it in exactly the same way.”
Prior to implementing IRI Software’s Express data engine for its EIS, Fidelity had a similar system that used Microsoft Corp.’s Access database . Although it worked fine in a small setting, the group determined that as the data and applications grew, both in the number of data sources and in historical data, Access wasn’t going to work, explains Rubinow.
That’s the reason Fidelity chose Express. It not only scaled well, it also supported users working from remote locations. Express gave users traveling anywhere in the world access to data in the corporate database, which is updated five times per day. The new data engine also had to be extremely easy to use, because, in Rubinow’s words, “the people using it are not analysts and wouldn’t necessarily know what to do with a screen full of tool bars and push buttons.”
The first step in implementing Express was determining a way for the managers to dial in. The other products that Fidelity evaluated were solely client-server based, so it would have been difficult to satisfy the additional requirement of having a remote access connection. Express got a leg up on the competition because it works bo th in a client-server environment and as a standalone.
Fidelity’s selection of OLAP technology was heavily influenced by the track record of the vendor. Also, when it came to searching for tools to build the EIS, Fidelity’s IS managers had to look no further than other companies within the Fidelity group that were already relying on such tools. Among the OLAP tools considered were Pilot Software Inc.’s (a subsidiary of Dun & Bradstreet) OLAP Server, SAS Institute Corp.’s OLAP ++, Arbor Software Corp.’s Essbase, and Comshare Inc.’s Commander OLAP Server.
Although Fidelity’s IS department built EIS using the Express data engine, it did not use the product’s front end because Fidelity had “very specific user requirements that negated the use of any tool that was on the market,” says Rubi-now. “The Express data engine was the most valuable piece for us.”
The group designed and wrote the front end of the EIS using Microsoft’s Visual Basic programming environment. With a combination of vendor support and internal developers, it took four months for his colleagues to write the front end. “That enabled us to give this sort of unique navigational style to facilitate the ease of reuse,” says Rubinow.
In the end, the group developed a powerful front end that required only one or two mouse clicks to gain access to a great deal of information. “It’s so intuitive that anybody without training can immediately understand how to use it,” says Rubinow. “If you understand the data–and most of these people do–they know what logical groupings of data there should be.”
Rubinow says Express has given his group “a platform that can grow with us as we add more types of data and additional requirements for different views of the data. The expansion possibilities are really limitless; we don’t have any walls that we expect to bump up against.”
At Fidelity, senior managers and their support staff use Intel-based 486 or Pentium machines to access data over a Novell Netware loca l-area network. Express runs on a Novell server that is linked to Unix servers running Sybase Inc.’s RDBMS.
Express is a groupware application and data server based on a multidimensional data model. Express View, its reporting tool, is a Microsoft Windows-based Object Linking and Embedding (OLE) 2.0- compliant product that includes tables, graphs, and data sets that can be linked to Express OLAP server.
Basically, OLAP databases allow users to pivot through data in much the same way they do when using a spreadsheet, so users can examine a product by region and distribution channel, or look at sales by distribution channel for both product and region. These multidimensional workhorses provide other analytical query functions that are difficult to perform with SQL. OLAP tools are optimized to extract data into multidimensional arrays for quick analysis and answers.
“Essentially these OLAP databases work by precalculating the interception between dimens ions in your data,” says Wayne Eckerson, editor of Open Infor-mation Systems and an analyst at the Patricia Seybold Group in Boston. “It’s like creating a big spreadsheet with more than two dimensions where you’ve calculated all the answers of any two dimensions that you want. The good part is that when you want to find out what the value is at the intersection of any two dimensions, the answer is already there and the performance is very, very fast.”
Express is specially designed for building enterprisewide business intelligence systems. It serves as the foundation for IRI Software’s OLAP applications, which include ExpressEIS, an application generator tool for building EIS systems; and ExpressFMS, a financial management system for building and consolidating financial information in an enterprise.
One problem Fidelity faced was getting people up to speed with Express. “There is not a whole army of people in the world that are familiar with the product,” says Rubi now. Second, he points out that Express “occasionally runs slowly, and that’s always a concern for us.”
Analysts see some potential shortcomings with OLAP: OLAP databases can mushroom in size quickly from 100 megabytes to 40 gigabytes, largely because vendors try to provide answers to almost any question; and there’s the double whammy of pushing your database administrators to their limit by having them support another database.
Chet Geschickter, vice president of research at the Hurwitz Consulting Group in Watertown, Mass., says IS organizations that are having difficulty supporting different database architectures will probably not want to support another database engine for OLAP. “For certain parts of the market, it is unpalatable to redo the database,” he says. That is leading database vendors to create OLAP schemas on their RDBMSs. Both Sybase Inc. and Oracle Corp. appear to be moving forward with products that will aim to satisfy these users.
“[OLAP databases] are not relational databases, ” says Eckerson. “To acquire one would require a company to either bring in someone with expertise in that database, or free up personnel with sufficient resources to become familiar with that database.”
In general terms, OLAP tools suffer from the same weakness that most new products do: administration and management. “There is also the issue about how you set up and design [an OLAP tool],” says Geschickter.
And there’s another significant problem with OLAP: a lack of standards. In fact, the Gartner Group recently reported that vendors have not yet been able to agree on standards that will be meaningful or bene- ficial to their customer bases. This despite the fact that an OLAP council has been formed to recommend an interoperability interface different from SQL that reflects the structure of the data.
Rubinow expects that the use of OLAP will grow within Fidelity. He also predicts that Fidelity will use other IRI products, such as ExpressView, but at this time the IS department doesn’t have an i mmediate need for them.
Today data warehouses appear to be on the minds of most companies’ IS departments. The reason is simple: They provide users across the enterprise with the ability to centralize repositories of corporate data. Data warehouse users can quickly perform complex analyses of data with OLAP. But some cooperation among vendors and standardization is important for the long-term viability of OLAP, especially in the areas of terminology, query language, data structures, and application programming interfaces. And considering the rapid expansion of databases within companies, OLAP databases will have to offer greater scalability beyond the current ceiling of 50 gigabytes.