If you ask any given group of executives what they’d really like in a business intelligence product, the wish list would look a lot like the description of Cosmic AC, the universe-spanning computer in Isaac Asimov’s classic short story, “The Last Question.” Cosmic AC contained all knowledge and could answer any question it was asked. After the first couple of releases, the IT department wasn’t involved.
And that is what the executives in many companies want-to have their business questions answered immediately, and without needing to involve the IT department to formulate the questions and provide the reports. In short, they want to be able to draw data from a wide variety of sources and use that data to discover relationships that were previously unsuspected, but which can impact their businesses, and to do it immediately by simply asking the right question.
Sadly for those of us forced to face the world without the services of an all-knowing, hyperspace-dwelling computer (only the NSA is allowed to have those), getting real BI can be a chore. In traditional settings, it means getting the IT department to set up specialized database queries that comb through your data warehouse and produce reports. But as useful as those reports are, they don’t reflect real-time information needs. For that, there needs to be a way for managers to find data whether it’s inside or outside the company and run their own queries. In short, they need to do it themselves.
“I think it’s where the whole industry is going,” explained James Kobielus, an analyst in Forrester Research’s IT Client Group. “Users want to do self-service BI.”
Kobielus said while the massive reports of traditional BI have their uses, they don’t lead to fast reactions. “It’s all about decision support, ad hoc queries and dashboards.” The ability to get immediate answers from diverse information results in businesspeople making better decisions, he argued. “They want to improve decision support more quickly. They want a personalized view of decision support, they want to grab the data they need and not be distracted by extraneous data. Self-service BI enables that.”
Kobielus also noted that moving to self-service BI lightens the load on the IT staff. “You’re taking a big workload off of the IT group. The big task is implementing the data warehouse. Building reports and report formats, maintaining them, [and] adding columns and fields,” he said.
Setting Up Self-Service BI
But implementing a self-service or do-it-yourself BI system is easier said than done. The reasons are obvious: BI doesn’t exist in a vacuum-it has to be collected, formatted and made available to the person who needs it. In addition, there needs to be some sort of query mechanism so that you can formulate the questions and send the queries somewhere to be answered. Finally, you need to format the answers in a way that makes sense, whether it’s in graphical form or as a set of tables. All of these activities require somebody in the IT department to set things up so that you can do your self-service BI. Depending on your data warehouse and your company policies, this may include preparing reports in advance that you can draw from, or it might mean providing access to data from outside organizations. But in whatever form, it doesn’t happen on its own.
“One of the frontiers of self-service BI goes beyond that to where the user can bring massive amounts of information into their own BI client so they can do complex visualizations within Excel and Web 2.0-enabled browsers. It’s ‘data mining light’ with complex visualizations and custom views,” Kobielus explained. “There’s more than that. There are new tools like QlikView, and Tibco with Spotfire. They offer in-memory BI clients that offer a very strong browser-based visualization tool. They bring data from your in-house data warehouse and external data. They have the ability to bring millions of rows of relational data without having to go back to the server. PowerPivot does much the same thing with Excel, SharePoint, Internet Explorer and SQL Server reporting services.”
Tony Murabito, CIO of Cubist Pharmaceuticals, makers of the antibiotic Cubicin, has found Tibco’s Spotfire to be invaluable in providing the fast-reaction BI that is critical to success in his industry. “We’ve started to leverage it across the business,” Murabito said. His executives use it to “track drug information and perform pharmaco vigilance and to give the staff the ability to drill into data, [and extend Spotfire] to our quality department to use with the CAPA system (corrective and preventative action), and basically we’re looking to use this across the organization.”
Murabito said he started seeing results almost from the first use of Spotfire in the company’s commercial area. “One of the earliest ‘ahas’ was when we went to analyze our speaker events. We wanted to understand the effectiveness of these events. We were able to determine that depending on the mixture of attendees we had a better result from those events. When we brought a mixture of attendees at events, it created a better dynamic and it improved sales.”
Cubist uses Spotfire to disclose data relationships following events such as mock drug recalls, and even to motivate the sales force. In addition, Murabito has formed a group of power users he calls the “Discovery Shop” that looks for new ways to use analyze data from a wide variety of biological agents.
But of course there’s more than one way to skin a data warehouse. Or in the case of Ayad Shammout-the lead database administrator for the CareGroup Health System, which is the network of teaching hospitals for the Harvard Medical School-a lot of data warehouses. “Initially we started implementing BI at the Deaconess Medical Center,” Shammout said. “We have many data warehouses. We had BI by using custom reports developed internally using excel and SQL analysis. A problem was how to bring the information to the desktop of the data analyst. They don’t really understand the technology.”
Other BI Approaches
For the IT staff, the shoe was on the other foot. “As IT staff we understand the technology but not the data. We have to work closely to develop BI,” he said. The solution? “Microsoft engaged us to work with PowerPivot. This allowed Excel users to use the technology without having to understand it. We’ve been amazed at the ability. It’s a managed do-it-yourself BI. We are managing SharePoint and SQL server.”
Unlike Spotfire, which is browser-based, PowerPivot uses the Excel interface, with which most people are at least slightly familiar. However, doing your BI this way isn’t for the timid. “They can draw from different data sources, including SQL, Oracle, Access and Excel,” Shammout said. He noted that IT needs to grant access to the databases before they can be used with PowerPivot.
“It’s targeted at Excel power users,” he said. “We have power users that need to do analysis, calculations and formulas. PowerPivot gives more data analysis. … They use it to compare years, events, etc. They can filter on specific data. There are complicated formulas they can add on.”
But PowerPivot, as useful as it is, shares a drawback of self-service BI in general: “It requires a basic understanding of Excel and data analysis,” Shammout said. “A few weeks of internal training are required for users who have never seen PowerPivot and never used Excel. Power users can use it in a few days.”
Once the CareGroup staff is past the learning curve, self-service BI does pay dividends, he said. “We have 2,000 users using the BI solution. They can view or slice and dice the data depending on what they need. We give more flexibility for using the same report without having to create different reports for each manager.”
As is the case with any new trend in IT, the definition of do-it-yourself or self-service BI depends a lot on who you’re talking to. Bhaskar Ballapragada, president of AdOn Network, a Web advertising service, uses his Sybase and MicroStrategy BI system to find ways to react quickly to advertising needs. But unlike with Spotfire or PowerPivot, he uses an overlay on his traditional BI system. To accomplish this, he said, his staff runs the BI reports it always has, and he has the ability to draw from those reports to get the data he needs.
The Sybase approach does have some limitations, however, since it requires the IT department to structure the queries, and it limits the use of outside data. “Once the user base has the report set up, the user has access to standard things,” Ballapragada said.
Right now, self-service BI is still in its formative stages, and some big changes are expected to come in the fairly near future. There’s a need for better predictive analytics, for example, and a need for a broader level of governance, Kobielus suggested. He said predictive modeling should be offered soon, allowing BI to attempt to look into the future rather than simply reporting on the past.
“The way BI is shaping up is to allow you to mine the deep past, deep future and the present,” Kobielus said.