In-Depth
Insights from Tableau's User Conference
The recent Tableau user conference sparkled with case studies and insights
- By Ted Cuzzillo
- 08/20/2008
When you walk into your kitchen, the first thing you're likely to spot is the spider in the corner. It's the one thing that may signal danger. That's how our brains are made, said Tableau CEO Christian Chabot in his keynote at the company's mid-July user conference.
"Too many people think that analysis is what happens in a series of reports," he said. Instead, analysis should be live and interactive, drawing from any data source on a whim. "That's where the richest insight occurs."
He made me think of Steve Jobs' talk about Mac OS. Jobs and Chabot are good showmen, though Tableau's overflow crowd left little room for Chabot to pace. More important, both base their products on simplicity without simplemindedness. Both products try to just get out of the user's way.
During two days of rare Seattle sunshine at a hotel overlooking Puget Sound, the company's visual analysis tools -- especially the just-released Tableau 4.0 -- made data seem to sparkle no less than the glistening water outside. The charts looked good, as usual, but the insights I heard users describe in case after case were even better.
For the director of data administration and reporting at Cornell University College of Arts and Sciences, adoption of Tableau was risky. By the time she discovered the tool, Cindy Sedlacek and her team had spent eight frustrating months trying to implement an enterprise-wide system from Hyperion.
"We spent so many hours and meetings mostly trying to learn tool functions [that] we couldn't concentrate on the data," she said. The workload was 90 percent about the tool and 10 percent about the data." Sedlacek said that when someone would ask how she got a certain number, she'd reply, "Oh, please just take it."
When she quietly raised the possibility of a switch to Tableau, writing off eight months' investment, she was told by peers, "'You could be fired.'" Her team discussed the risks of switching. Tableau Software's relatively small size raised concerns about the company's ability to deliver.
Over Christmas break, shortly after implementation of Tableau Server, Sedlacek dived into an analysis of faculty salaries for the dean. She drew data covering the last 20 years from 20 FileMaker Pro databases kept by human resources and other departments.
With the previous tool the college had tried to use, she said, the task would have been more difficult. "I don't even want to think about it. It would have taken a month or more. Then to get it into a user-friendly report for the dean? Forget it."
She finished in time and presented the data to the dean. He saw the evidence: the trend lines showed a crisis forming. Faculty salaries at the low end were edging upward while on the high end salaries had stagnated. Some assistant professors, for example, had come too close to older, full professors. The dean soon met with the provost, who quickly approved significant new funding.
Other Tableau users, such as government anti-fraud analysts, care only about anomalies. Earlier this year, one analyst -- who asked us to protect his identity, the name of his agency, and almost all other details -- spotted inconsistencies in a series of fabric shipments. The shipper always declared a much lower value at the originating country than the one declared at the receiving country. Container weights were also too low. The ensuing lawsuit may come to trial. The analyst notes that with other tools, those shipments probably would have slipped by undetected.
Tableau's filters were the key. In the agency's system, each shipment was scored for probability of legitimacy. The analyst could use the filters to eliminate from view all but the most questionable units. For points of entry that have no suspicious cargo, for example, his display filters out so much there appears to be no traffic at all. Though the same formulas could be used in Excel, the patterns are much easier to see in Tableau.
At Coca-Cola Enterprises, the manager of forecasting and planning is concerned with improving routines. Andy Kriebel explains that every improvement in operations -- such as improved sales -- contributes directly to the year's bottom line.
"You could never see sales goals before," he said. Tableau has made them visible in reports he and others issue, derived from data from Microsoft Analysis Services and SQLServer. Kriebel's supervisor was skeptical of Tableau at first but, said Kriebel, changed his mind within the first week. The supervisor can now see right away who's above and below the goal. "He knows who to chat with or pat on the back."
Kriebel later helped to demonstrate that top-notch charts are not so easy to make. He bravely submitted samples of charts for review in a session named "Extreme Makeover." Tableau director of visual analysis Jock Mackinlay frowned at numbers showing across a bar graph, but at least Kriebel had an answer: Coca-Cola executives demanded them.
Mackinlay, Tableau lead technical writer Erin Easter, and Stephen Few (a leading authority on visualization) picked apart more samples, criticizing color choice, excess white space in a scatter chart, scale, and other fine points.
"Tools by themselves don't guarantee well-being," Few said earlier in his keynote address. He gave up his fire-and-brimstone preaching career long ago, but he can still breathe fire.
"The BI industry has failed to let us run business more intelligently," he said. "Only through data visualization can we uncover meaning and make the light bulb go on in our heads."
He also cautioned Tableau: it must stick to its vision of simplicity and not give into temptation. "Help them stay true," he told the audience, "and ask for the right things."
- - -
Ted Cuzzillo, CBIP is a freelance writer based in the San Francisco area. He can be contacted at [email protected].