Path to EnlightenmentDirect business intelligence can help you avoid data warehousing headaches. But without the proper planning, it offers plenty of its own
By Cindi Howson Continued from Page 1 Both Brio and Business Objects use microcube technology to support OLAP. As the results of a query are sent back to the midtier server in a Business Objects Web deployment, or the desktop in a Windows or Brio deployment, the software dynamically creates a microcube. Although not a microcube, Arcplan caches query results on its application server. These microcubes/caches aren't as big, powerful, or complex as the heavyweight cubes built by Hyperion Essbase, Microsoft Analysis Services, or Cognos Powerplay; in fact, all provide access to leading OLAP data sources. However, the microcube approach supports simple drill-downs and can offer greater flexibility, as cubes don't need to be built in advance with precisely defined information requirements. Because Cognos would prefer that its customers use Powerplay for full-blown multidimensional analysis, it doesn't provide drill-down capability in Impromptu. However, Impromptu has a unique feature called linked reports that can simulate drill-down capabilities. In this approach, the report author defines a link on a summary report, and then builds a separate query to present the details when a user double-clicks on the link. This may require slightly more administration and potentially slower response time than the microcube approach, but it is a cost-effective alternative. Rich, prebuilt functions. If you're working in a source system, you don't have the benefit of star schemas in the data warehouse to facilitate this-year and last-year calculations. Nor do you have the benefit of a robust OLAP tool to do multidimensional calculations such as gross margin and market share. Thus, the BI tools must make up for the limitations of SQL and lack of an optimized data model. The leading tools do, and all those listed in Table 1 (PDF, 20K) support the most critical needs such as cross-tabs, top-N reporting, running totals, and market share. (The latest version of Crystal Decisions also includes functions to calculate key performance indicators such as Inventory Turnover and Cash Flow vs. Current Debt.) Keep in mind, however, that most of the work is done on the client or midtier server. Thus, more data may be sent across the network to accomplish these calculations than if the calculations were done in a data mart or warehouse. Data from different sources. Without a data mart or data warehouse, you also may not have the luxury of consistent reference data or hierarchies such as customer and product groupings. However, with direct BI, you could define hierarchies in Microsoft Access, Excel, or a separate database. The BI tools then allow these hierarchies to be joined on the client so that they appear as one source in the report. Brio and Business Objects accomplish this goal in the microcube, whereas Impromptu uses "hot" files. However, matching data among different sources can be tedious and requires complex SQL. If this procedure has to be repeated for different applications, it eventually becomes more cost-effective to maintain via a data mart or data warehouse. Administering Direct BIDirect BI must be carefully managed to limit the impact on the source system and understand how people are really using the tools. Two ways you can improve the success of a direct BI implementation are to present users with a business view of the ERP system (item 17) and monitor usage (item 18). In the first approach, ad hoc users shouldn't need to know the entity relationship model of the source system. Instead, they should be presented with business terms such as "customer ID" rather than a meaningless field name such as "L33_No." Nor should they have to define joins; a business view of the source system hides the complexities of the entity/relationship model from the user. (Brio and Crystal Decisions call these views "dictionaries," Business Objects calls them "universes," and Cognos calls them "catalogs.") Crystal Decisions is the only tool that doesn't require the dictionary, but such an approach goes back to the model in which IT creates the reports, not the users. Arcplan assumes that companies will want to develop full applications, so it doesn't provide a business view for ad hoc reporting. Business Objects and Impromptu leverage the business views by providing prebuilt universes/catalogs for popular ERP systems, called Rapid Deployment Templates (RDTs) or Headstarts, respectively. (These are provided free of charge.) However, don't expect complete solutions; RDTs/Headstarts may not include all the information you need or they may not be formatted correctly for your source system. For example, it may be necessary to modify the RDT or Headstart to accommodate nuances in your OLTP system implementation. Such complex SQL and transformations will slow any query and increase administrative costs. When you find that much of your time is spent coding complex SQL in the business view (universe, catalog, or dictionary), you need to revisit implementing a data mart or data warehouse in which transformations are coded just once. With direct BI, an audit feature (item 19) is critical in determining which reports people are running and when. Not only does this help assess system degradation, it can provide useful insight for which subject areas would be most used in a data warehouse. Audit capabilities were completely absent in earlier versions of BI tools, yet are now available in most tools with varying degrees of functionality. Crystal Decisions and Arcplan don't directly support auditing, but claim that it's programmable. Risk vs. RewardDirect BI sounds risky, and it is. Yet the prospect of a quick payback and access to previously hidden information (without all the politics and complexities of a full-blown data warehouse) make it worth the effort in today's economic climate. Users will quickly discover the power in the latest BI tools, which can help provide support and vision for a data mart or data warehouse. The key to success is managing expectations and recognizing when to migrate away from direct BI to an infrastructure optimized for reporting and analysis. If you do that carefully, you can reuse much of the work from implementing realtime BI. Cindi Howson [cindihowson@askcindi.com] is an independent consultant specializing in BI. She previously led the analytics practice for Deloitte & Touche in Houston and was the BI standards leader for a Fortune 500 company. RESOURCESBrio Software: www.brio.com Business Objects: www.businessobjects.com Cognos: www.cognos.com Crystal Decisions: www.crystaldecisions.com
|
 |
Most Popular This Week
IE Weekly Newsletter
Subscribe to the newsletter
| |
| ||||||||||||||||||||||||||||||||
