When it Comes to Data, Image is Everything

• 04/26/2000

Cognos Corp.'s Visualizer 1.0 is a product that leads the way to better decision making via visual representations of data. It takes a data source, evaluates groups of information, then relays the information to the senses in a mode we all can grasp: a picture. The product helps with the business intelligence decisions process -- which may be more valuable than the project itself -- before actual implementation.

The data can come from many different database sources: PowerCubes, OLAP storage resources, ODBC repositories, or plain old Microsoft Excel spreadsheets. Visualizer comes with numerous sample files, and helps users get on their feet with the product in short order. Our tests were conducted using an Excel 97 spreadsheet of a customer’s field sites from a past project. Despite completing the project on time, under budget, and meeting the customer’s expectations for the project deliverables, Visualizer's representation of the spreadsheet data -- which showed the sites by state and hours consumed at each site -- made us aware of the project's shortcomings too late, even with a good closeout.

This review was performed using two servers on a Cisco Systems -- routed network using switched 100 Mbps Ethernet LAN. The first server was a Microsoft Windows NT Server 4.0 running Internet Information Server 3.0, patched to the latest revisions. This server was a dual-processor Pentium 233MMX with 256 MB of memory, 20 GB of storage, and a Diamond Viper V550 video card to develop the visualizations. The second server was intentionally built using an older Pentium 200-MHz Web server with 96 MB of memory and a single Enhanced IDE 8.4 GB drive.

We wanted to test Web performance on a slow server, then use our dual Pentium-II 400-MHz, load-balanced Web server for higher performance testing as a comparison of real world Web servers. Also, we tested Visualizer on a Windows 95 client machine, which required the use of the DCOM client software supplied on the product’s CD.

The installation process was a bit awkward, but successful. We first installed the Web edition of the product so we had a place to distribute the client application, which was archived at 8.4 MB in size. You can share out the archive on the local network, or make it available via the Web server.

The Web server was initially the slow Pentium 200-MHz server. This distribution point could easily serve 35 to 50 users without much of a problem. As always, the more powerful the machine and the more powerful the network, the better you’ll be able to serve larger clientele. The Web edition has an administrative function, which we installed. Noticeable with the product was Netscape’s LDAP for authentication of users. The supplied product was version 3.1, despite 4.1 being the current revision. The provided 3.1 was very light on memory and server resources, and performed flawlessly.

Next, we installed Visualizer on the workstation. It can be put on a laptop for travelers. Weighing in at only a couple of megabytes of memory, the disk requirements are equally slim until you start creating the actual product. Our test visualization came in at 1.94 MB with a medium size spreadsheet. We didn’t have any problems installing it on any desktop workstation, some of them being Pentium 166 MHz clients with 64 MB of memory.

We bypassed the choice of ODBC data sources since we didn’t have the needed add-on products, but this wasn’t a problem. We opened up a few of the provided samples and viewed the handiwork of the developers. This gave us a feel of what we should expect from the forthcoming test. We were pleasantly surprised to find that these samples were relatively fast and easy to edit and update. It also allowed us to add our own styles.

We created a brand new blank visualization. Using the wizard, we told Visualizer to import the test spreadsheet. Composed of four spreadsheets within the workbook, we picked out the one spreadsheet that listed all of the customer’s sites scattered across the US. We chose an image of the US mainland for Visualizer's backdrop. It was interesting to note that this product is internationalized out of the box, containing many international images.

Visualizer took a few minutes to crunch the data, and gave us the first glimpse of our past project from a perspective that we’d not thought of before. This view of the project clearly demonstrated that we’d deployed in the wrong order of sites. Despite completing the project under budget, we could have saved an additional 18 percent of travel costs by changing the approach to a dozen sites in the Midwest.

For that past project, Atlanta was the kickoff site. The sites immediately to the west of Atlanta were assigned to our offices in Texas, but were executed during a time of higher airfares in that region, where the airlines were known to have seasonal rate changes. The people making travel arrangements weren’t very good with spreadsheets, but anyone can understand a picture. Presented with this view of the field sites, they at once recognized changes that could’ve been done to review some sites later than others.

We made some changes to the depiction, added labels, and used impressive 3-D presentations, as well. We used the 3-D pictures by adding pivot points to the data so that we could perform a visual analysis in forms that interested the different parties of our MIS group. Once done, we published the visualizations to the Web server so people could access them across the enterprise. The Web page had links to the files and to the client application if the users didn’t have that product to view the files.

Not everyone is fluent with the many different data repositories out there, but anyone can understand a picture. That’s the gist of what Visualizer does, and it does so very effectively.

[Infobox] Visualizer 1.0

Cognos Corp., Burlington, Mass.
(800) 426-4667
www.cognos.com

Price:
Cognos Visualizer Web edition begins at $50,000 for an enterprise server license with 100 Web users. Cognos Visualizer author edition begins at $695, with volume discounts available.

Pros/Cons:
+ Excellent use of source data.

+ Versatile editing functions.

+ Easy to create new visualizations.

- Installation is awkward.

- Processing visualization can be demanding.