Empowering the new problem solvers
Without business intelligence applications, most agencies wouldn’t be able to keep up with the internal, regulatory and congressional demands for detailed oversight and analysis of their programs and operations — whether that means tracking agency spending, the default rates for home loans or wildlife populations.
That insight comes with a catch, however, because those reports take time to produce and can tie up administrative and IT department resources. In addition, a growing trend toward real-time analysis of performance, rather than more historical-focused data warehouse applications, is pushing agencies to consider so-called self-service business intelligence (BI) applications to get around the bottlenecks of traditional BI.
In the old days, agency users had to send their report requirements to employees trained in BI, who would set up the queries and generate the reports. Self-service BI tools allow end users — from frontline workers to the most senior agency executives — to produce their own reports with relatively simple Web-based database querying and reporting tools.
Theoretically, that means agencies are more tuned into their performance and responsive to changing mission needs because subject-matter experts can produce more tailored and meaningful analyses than the mostly canned reports for which traditional BI is used.
However, despite the friendly sounding moniker, self-service BI comes with a few do’s and don’ts agencies should follow if they want to keep it from becoming another well-intentioned but ultimately wasteful effort.
DO tailor the system to different types of users
Self-service BI doesn’t equate to simple BI, so don’t assume that one user interface will suffice. What will satisfy a general information consumer might not be enough for a departmental power user or a high-level executive who needs a bird’s-eye view.
“It’s a matter of classifying the users and providing the right interface, whether that’s an ad hoc query tool, a dashboard, a search interface or whatever,” said Cindi Howson, founder of BI Scorecard, which measures the effectiveness of BI solutions. “That makes sure the user can interact with the tool in a way that’s appropriate with their job level and their data and computer literacy level.”
For example, you need to consider whether you want to empower all users to be able to answer any question or just a small subset of questions, said Bob Gourley, founder and chief technology officer at Crucial Point, a technology research and advisory firm. And do you want them to be able to navigate the agency’s entire index of BI data or do you want to limit access based on individual jobs?
“Those are major architectural questions,” said Gourley, who introduced self-service BI to the Defense Intelligence Agency when he served as DIA’s CTO. “You need to do it with an end state in mind and architect self-service BI so it’s focused on your mission.”
As with just about every technology, security is a big issue with self-service BI. Gourley said agencies should wrap good auditing and log management around its self-service BI tools. That way they’ll know exactly what their employees are doing, which will make it easier to serve their needs and intercept any questionable behavior.
DON’T push the IT department aside
It’s a tough temptation to resist, given that the IT department is often seen as a bottleneck when it comes to traditional BI. IT employees are the ones who usually field the requests, formulate the database queries and then send the reports back to the users. So if users can handle that process themselves, why bother with IT any more?
“IT’s role is still a critical one,” said Boris Evelson, a vice president and principal analyst at Forrester Research. “IT maintains the infrastructure, and IT is the one that will procure the self-service BI tools and maintain them.”
The IT department will also be the party that makes sure the data that self-service BI users depend on is clean, integrated and reconciled — in other words, fit for duty. That work is typically 80 percent of any BI initiative, Evelson said.
Self-service BI can disrupt the workflows that were established for traditional data mining, in which IT plays such a critical role. Therefore, it’s important to be sensitive to potential opposition to self-service BI among the affected IT employees.
“Some people [in IT] won’t like it because they’ll feel their jobs are being threatened, but if you do it right, most people will love it,” Gourley said.
DO offer education and real-world examples
It’s a mistake to assume that self-service BI will sell itself or be so intuitive you only have to build it and users will flock to it. You’ll need a comprehensive plan to convince people of its worth, and you’ll need to follow up with training to turn that interest into action.
Through measures such as weekly seminars and slideshows, you can demonstrate how easy it is to produce visualizations of data using self-service BI tools, said Brand Niemann, former senior enterprise architect and data scientist at the Environmental Protection Agency.
“Then you show people how easy it is to manipulate things to show different aspects of the data and, importantly, how to produce a data story,” said Brand, who is now senior data scientist at Semanticommunity.net. “Then people get hooked because it’s a natural thing to want to tell a story, and that’s something traditional BI totally misses.”
You also need people who will champion the system. If you leave it to the inherent intuitiveness of self-service BI to sell itself, the effort will fail.
Howson’s BI Scorecard runs surveys every couple of years on the success factors involved with such efforts. “One of the critical ones is how much there’s a hybrid business/IT leader championing the effort,” she said. “You have to promote it, talk about it, evangelize it and have incentives in place.”
DON’T expect a single tool to be enough
As the self-service BI push builds, you’ll hear many vendors claim that they offer the be-all, end-all tool. Don’t believe them — or, at least, not everything they say.
“None of the leading BI vendors have all the range of self-service capabilities, which is why we see more organizations using two or even three BI tools at a time,” Evelson said.
Agencies have lots of options. BI vendors such as Endeca, IBM’s Cognos, Information Builders, QlikTech, SAP, Tableau Software, TIBCO Software and others have introduced self-service BI tools, either as stand-alone products or as Web-based front ends to traditional BI solutions.
However, you’ll need to make sure that the tools you pick will work with your agency’s existing software and data. As Gourley pointed out, data comes in multiple types, so the self-service BI tool you choose should be able to index any data you have and deliver it to users via a friendly interface.
Niemann advises agencies to start off with one core tool that handles most requirements, then look to other tools to provide any additional functionality that might be needed. But be careful.
“The problem with using more than one tool is interoperability,” he said. “You need to have enough commonality between the tools you choose.”
The keys to any successful deployment of self-service BI are scalability, reliability, ease of access to data and, above all, ease of use. It’s true that nontechnical users will have to learn to develop and design their own reports rather than simply printing out a canned report prepared by an IT department specialist. But the potential benefits outweigh that effort.
Checklist of self-service features
Even the leading business intelligence products don’t deliver every self-service feature that an organization might find useful, market analysts say.
Here are the self-service features you can expect in most products and others for which you might have to shop around.
- Customized metrics.
- A portal that allows centralized access to tools and facilitates user collaboration.
- Templates for reports and dashboards.
- Wizards for reporting, querying and building dashboards.
- The ability to customize reports with prompts, sorts, filters and ranks.
- A semantic, metadata layer to shield casual users from database complexities.
- A search interface for finding relevant data and reports.
- The ability to drill anywhere for access to undefined data.
- A post-discovery process for analyses based on raw data using little upfront modeling.
- Prompts for columns of attributes, such as names and locations, not prebuilt into a report.
- The ability to update the underlying database in real time.
Source: Forrester Research