Cindy Sedlacek, head of data administration and reporting at Cornell University’s school of arts and sciences, knew it was time to consider a new business intelligence (BI) platform when employees started
“We had people that wanted to quit -- literally,” Sedlacek said. “We had several people that were just exhausted.”
The university, located in Ithaca, N.Y., is made up of 11 colleges, all of which had for years run their BI and reporting programs separately -- if they even had BI, according to Sedlacek.
By the summer of 2007, the school of arts and sciences was using Hyperion’s Web Analysis interactive visualization application and Essbase multidimensional OLAP server to dig into admissions and financial data. The school migrated to the tools after many years using BI technology from Brio Software, a smaller BI vendor that Hyperion acquired in 2003. Hyperion was itself acquired by Oracle in March 2007.
While the platform’s front-end functionality was decent, the administration was a monster, Sedlacek said. Between developing data models and maintaining the underlying metadata layer, admins were putting in long hours just to keep the Oracle platform up and running, she said.
The platform also lacked sufficient self-service capabilities, meaning that Sedlacek and her team were spending lots of time creating reports and responding to end-user requests for new dashboards and visualizations.
Something had to give.
Not far to look for a more lightweight BI platform
When Sedlacek decided to look for a more lightweight alternative, she didn’t have far to go.
Hyperion and Tableau Software have been OEM partners since 2005. Hyperion uses the smaller vendor's well-known data visualization technology, calling it Hyperion Visual Explorer. As such, Sedlacek was familiar with Tableau’s visualization capabilities and considered turning to the vendor as the school’s standardized BI platform.
A demo sealed the deal, Sedlacek said.
But it wasn’t just Tableau’s data visualization technology that sold her. It was also the platform’s light footprint and self-service capabilities.
For one, it would require far fewer man-hours to administer than any other platform, she said.
“The other tools we looked at were going to require the same amount of effort [as Oracle] and pretty much be a beast,” Sedlacek said. In addition to Tableau, Sedlacek said she also evaluated BI software from Business Objects, Cognos and Microstrategy.
And Tableau’s Web-based platform allows casual business users to easily create interactive graphics like box plots and heat maps with drag-and-drop capabilities made possible by the vendor’s Visual Query Language, or VizQL. The school’s deans and business officers could, for example, quickly overlay and view expenditure data on top of budget data.
“You choose a bar chart option and then you go and say ‘layer.’ That’s it,” Sedlacek said.
The school signed on with Tableau soon after and has been creating new data visualizations with it ever since.
With the Tableau platform, school officials plot and analyze – on their own --admissions data on interactive maps of the U.S. They can adjust the filters to analyze who’s applying to Cornell by region, gender, ethnicity and more.
Another visualization allows them to analyze 27 years’ worth of faculty and staffing data via a moving time-series graphic.
And best of all, Sedlacek’s staffers are back to working regular hours.
Her team still creates KPI reports for the school’s faculty, but Tableau’s drag-and-drop capabilities made that job much easier. Creating the reports now takes half the time and requires half the resources compared with Oracle’s technology, she said.
Lack of data quality and cleansing capabilities
Ironically, though, the very thing Sedlacek loves about Tableau – the platform’s ease of administration – is what ultimately may undermine its appeal to others, especially large enterprises.
Tableau achieves this ease of administration by removing most data quality and data modeling capabilities, according to Forrester Research’s Boris Evelson. Instead, the Tableau platform ingests data directly from source systems rather than integrating and cleansing it in a data warehouse first.
Oracle and the other megavendors tightly integrate data quality and data modeling into their platforms with the goal of creating and ‘end-to-end’ BI stack. It is true that this means more admins and BI workers are needed to manage the platforms, but it also means that, if done properly, users can truly count on the integrity of the data.
Tableau takes another tack, assuming the data it ingests has already been run through the data quality and modeling paces, Evelson said. If that is the case, then Tableau could be a good fit, he said.
In fact, most organizations do not account for data quality at the point of collection, according to Henry Morris, an analyst with Framingham, Mass.-based IDC. That makes a BI platform’s data quality and modeling capabilities – or lack thereof – all the more important.
“The general rule of thumb is that 70% of the effort [in a BI program] is getting the data straightened out,” Morris said.
Tableau’s CEO, Christian Chabot, doesn’t see it that way, however. He said Tableau’s federated approach – integrating data from source systems without cleansing or modeling the data first – is what sets the vendor apart from its complex, inflexible and expensive competitors.
And he may be right. Tableau doubled its year-over-year sales in the first six months of 2010, according to the company, and now boasts more than 5,000 clients that include large enterprises like Bank of America, Eli Lilly, and Walmart.