Getting the right information to the right people at the right time to allow them to make informed, intelligent decisions quickly—broadly defined as operational BI—is a panacea for data-focused organizations. Although many enterprises are attempting to do this, few seem to be doing so with the combination of technological sophistication and data savvy that such a task demands.

One of the key problems: unstructured information, or data that exists in email, call center reports, on paper, and elsewhere, but often not in companies' data warehouses. Karen Parrish, vice president of business intelligence solutions at IBM, estimates that as much as 85 percent of enterprise information is unstructured, thereby hobbling efforts to make decisions based on the most complete set of data. "The way for people to move forward [with operational BI] is to marry the structured with the unstructured so that the analytics provide the full picture," she says.

Doing so, however, is an arduous process. Beyond the obvious—that any BI project will fail without the usual upper-level executive buy-in—experts recommend that organizations start by ensuring that their technology infrastructure can support the accumulation and analysis. Unfortunately, many IT folks attempt to forge or retrofit a BI solution within existing systems, rather than figuring out what they need from a strategic or practical standpoint and working from there. Working with the IT staff to correct any such issues often proves a difficult task, both practically and politically.

"They go off somewhere and build something that doesn't make anybody happy, believing that users will eventually come around. They think, 'You just watch, they'll see how it works,'" says Lisa Dreyer, director of product marketing for Sybase's worldwide marketing organization. "Sometimes, you just have to start from scratch."

Given that the IT arm generally doesn't like to be told how to conduct its business, and that even the biggest BI booster might be awed by the price tag of the new technology needed to fuel the company's efforts, Gartner analyst Betsy Burton recommends a test-drive of sorts. "You involve both the IT and business sides in a proof of concept. You take a real-life situation and run through it using three years' worth of data." Adds Dreyer: "Once you do something like that, the IT side maybe has a better idea where the [technology] blind spots are and they realize that they're going to be highly involved in the process. That tends to placate everybody somewhat."

Still, where most organizations tend to have trouble is in the process: coordinating the combination of structured and unstructured data, rather than the technology that facilitates it. The problem? Different constituencies within a company have different ideas as to what types of unstructured information will prove useful. "It's a unique situation," says Todd Price, managing vice president of business intelligence and the performance management national practice at Hitachi Consulting. "People don't stop to think that what's in their email can be used for real-time analysis." Again, the proof-of-concept exercise outlined above tends to convince the skeptics.

Once an organization has considered the strategic, people, and process issues, the goal becomes bringing BI to the shop floor, so to speak, in a manner that won't clog resources. "To a lot of people, 'change' is a dirty word," explains Pat Morrissey, senior vice president of marketing at Savvion. "If you generate a quick success on a small project, people won't be asking, 'Why are you making me do this?'"

A possible example: Rather than attacking every individual's inbox and attempting to merge all information contained therein into the existing data warehouse, an organization might instead start with information in call center reports. Such information, which ostensibly already exists in a semi-structured format, could help give the people behind the BI push that first win they need.

"There's so much information they're using now that they couldn't use before," Parrish says. "For HDFC, wouldn't it be great if the data warehouse 'knew' that a customer just bought a house, so that they could adjust the offers they're sending that customer? Now that they're putting the structured with the unstructured, they can do things like that."