How to get information to end users at lightning speed is what Gary Moore of Baltimore-based Aegon USA Inc. is
hoping to learn at The Data Warehousing Institute (TDWI)'s 2003 World Conference this week in Boston.
Moore, a team leader for data warehouse investments at Aegon, a parent company of Trans America Insurance Co., said he is hoping to discover ways to shift his company's focus to the end user.
"We need to deliver more information in real time, and we're not quite there yet," Moore said. "We came to the conference to make sure we get a leg up in this area."
Monday's conference keynote speaker was Ralph Kimball, author of several books on data warehousing and co-inventor of the Xerox Star workstation. He has served as the lead architect for 20 data warehouse systems.
Kimball told a room filled with hundreds of data warehouse experts that they should focus their work on bringing easily digestible information to decision makers in real-time speeds. Database administrators need to think like media publishers if they want to continue to play a key role in making business decisions easier for end users, he said.
Kimball offered conference attendees a list of 10 problems often encountered when planning and building a data warehouse, along with tips on how to avoid them.
Necessary data is unavailable. Data warehousing pros should regularly canvass decision makers to understand their current data requirements.
Lack of collaboration with end users. IT pros often blame end users for complaining when they haven't read the documentation. IT staffers should work in end-user environments, where they can learn more about the end-user experience.
Lack of explicit end-user cognitive and conceptual models. IT designers too often assume that their end users are like computers. To alleviate this, data warehouse builders should construct conceptual models of how a user performs a task and makes a decision. Customized user interfaces and canned reports may have to be built as a result.Data needed for decisions is delayed. With the demand for real-time information rising, sometimes data delivery is too fast for a DBA's extract-and-load system. DBAs should create a seamless connection of fact-table partitions to the static data warehouse.
Dimensions and facts not conformed. Often, labels are incompatible, and this results in confusion and wasted time. Use a data warehouse bus matrix to drive the organization to define conformed dimensions and conformed facts before data marts are implemented.
Insufficiently verbose data. The more data DBAs have, the more verbal they need to make it for end users. Turn codes into verbose text. DBAs need to go overboard on this. Troll for auxiliary data sources that provide more descriptive text.
Data in awkward formats. Any commercial data source has thousands of panels. Users can be confused and intimidated. There is a lot of evidence that it's a good idea to present data in a dimensional format.
Sluggish delivery of data. Instantaneous performance is the end goal. Choose good DBMS software that has clever algorithms and powerful indexing. Maintain a lot of memory and beware of shared nothing architectures that may choke on large joins.
Data locked in a report or dashboard. In an effort to make a sexy format or innovative design, data sometimes gets locked in a tool that can't be transferred. Choose a tool that allows one-click copying of data in tabular format to the clipboard from the table, chart or fuel gauge.
Prematurely aggregated data. Data marts must sit on a foundation of aggregated data, and DBAs want them to drill down gracefully and silently with aggregated data navigators.
FOR MORE INFORMATION
Check out a Featured Topic on Data warehousing.
To provide your feedback on this article, contact Robert Westervelt.