Case Study

Cornell Takes Visual Approach to Data Analysis

• By Linda L. Briggs
• 11/13/08

One of the challenges with business intelligence (BI) software, as many campus IT departments can attest, is the difficulty of implementing and using it. While powerful, BI tools can also be a challenge to master, especially for the non-technical business users who typically need the tools' analytical capabilities most.

Cornell University learned that lesson the hard way when it installed an enterprise BI system from a large, top-shelf vendor that the university prefers not to name. What Cindy Sedlacek, director of data administration for the College of Arts and Sciences will say, however, is that the installation, attempted back in 2007, was less than successful. Implementation was difficult, and, after eight months, the university had little to show for its time and money.

A year ago, Cornell turned to a different sort of BI tool, one that generates visual representations of data and analyses and has proved both powerful and extremely easy for users. The product, from Seattle-based Tableau Software, is helping the university's 11 colleges understand, graph, and compare a wide range of essential metrics. According to Sedlacek, Tableau just might be "the best thing that ever happened to us."

As a business intelligence tool, Tableau presents data and analysis visually, which can make it easier for business users to understand and manipulate information because it appears in colorful maps, charts, and three-dimensional diagrams. The live content in Tableau can be shared with Web applications, including intranets, and embedded in documents.

Cornell is using Tableau as part of a decades-old Key Performance Indicators (KPI) initiative, in which the various colleges within the university track and analyze key metrics to discern trends and set future direction. Sedlacek describes the initiative as the colleges' highest-priority IT project on an ongoing basis.

More and more, Sedlacek said, top business officers from the 11 colleges--who function as CEOs of each college, essentially--were asking for better access to critical data they needed to make decisions. "They wanted metrics at their fingertips," Sedlacek said, such as faculty and student retention numbers, enrollment rates, and expenditures. "A lot of this is very hard information to get. It takes a lot of time." Often, by the time the data can be collected and verified, she said, the numbers are out of date.

In the spring of 2007, the KPI committee decided to address the need by purchasing an enterprise BI tool. But after spending eight months implementing the software, training the KPI technical team, and holding various meetings, "we didn't have much to show," Sedlacek said. In the end, she said, "we'd spent lots of money on software and training.... We learned a lot, but with little result."

Before Tableau, 90 percent of the team's time had been spent on the BI tool itself, she estimated, and just 10 percent on the data. That changed when, as the technical lead for the KPI project, Sedlacek discovered Tableau. She proposed it to the project chair, arranged a demo, and saw impressive results almost immediately.

Tableau, with its far more intuitive interface and quick implementation, quickly changed the team's focus to the data, rather than the tool being used to manipulate the data. "We've flipped the ratio with Tableau--we've spent 90 percent of our time on the data [and 10 percent on the tool]," Sedlacek estimated. Tableau comes with intelligent filters, she said, which allow users to perform some of their own data analysis. Instead of the five-and-a-half full-time IT employees Sedlacek had estimated would be needed to maintain the project, just two spend time on Tableau today. That savings of three employees, Sedlacek said, means that Cornell probably recouped its Tableau investment within a matter of months, probably fewer than six: "We had a significant ROI."

Data at Cornell is stored and manipulated in a wide range of systems, including a mainframe, Oracle PeopleSoft for human resources and student modules, Microsoft SQL Server and Excel, and more. Cornell also has a large deployment of Hyperion Solutions' Brio, now part of Oracle, which continues to be used separately as a high-end BI tool for faculty and staff. Tableau, Sedlacek explained, "is meant for decision-makers."

About a hundred users take advantage of Tableau now, nearly all high-level decision makers, such as vice presidents, deans, directors, and the like. Training for Tableau has been minimal. "We had to create a system that didn't require users to attend trainings.... Tableau helped us meet that requirement," Sedlacek said.

After purchasing a server license and 20 desktop licenses originally, Cornell recently purchased another 100 desktop licenses. Sedlacek is now holding training sessions for a range of groups across the university to introduce other university users to Tableau. Those users include the alumni marketing group and the diagnostic center at Cornell's animal hospital. "I always want a BI tool that works for me, not against me," Sedlacek summarized. "You have to be careful with these enterprise BI tools--they can be complicated.  Yes, there are payoffs, but at what expense?" But with Tableau, "they really have built a tool from the end-user's perspective--truly the first intelligent tool for managing your business."