Once the data scoping is underway, IT should proceed with highly targeted projects that can be used to showcase results as opposed to opting for a big-bang, big-data project. "You don't have to spend a few million dollars to start a project and see if it's worth it," Raden says.
Let business needs drive data dives
It sounds like a broken record, but the concept of IT/business alignment is absolutely critical to an initiative as big and varied as big data, IT analysts say.
Many of the initial big-data opportunities have been seeded in areas outside of IT, they say -- marketing, for example, has been early to tap into social media streams to gain better insights into customer requirements and buying trends.
While the business side may understand the opportunities, it is IT's responsibility to take charge of the data sharing and data federation concepts that are part and parcel of a big-data strategy.
"This is not something IT can go out and do on its own," says Dave Patton, principal of information management industries at PricewaterhouseCoopers LLP. "It will be hard to turn this into a story of success if [the initiative] is not aligned to business objectives."
Early in its big-data initiative, Catalina Marketing's Williams brought business managers together with its financial planning and analysis (FPA) group in a team effort to make a business case for information architecture investments.
The business side identified areas where new insights could deliver value -- for example, in determining subsequent purchases based on shopping cart items or through a next-buy analysis based on product offers -- and the FPA team ran the numbers to quantify what the results would mean in terms of enhanced productivity or increased sales.
Re-evaluate infrastructure and data architecture
Big data will require major changes in both server and storage infrastructure and information management architecture at most companies, Gartner's Beyer and other experts contend. IT managers need to be prepared to expand the IT platform to deal with the ever-expanding stores of both structured and unstructured data, they say.
That requires figuring out the best approach to making the platform both extensible and scalable and developing a roadmap for integrating all of the disparate systems that will be the feeders for the big-data analysis effort.
"Today, most enterprises have disparate, siloed systems for payroll, for customer management, for marketing," says Anjul Bhambhri, IBM's vice president of big-data products. "CIOs really need to have a strategy in place for bringing these disparate, siloed systems together and building a system of systems. You want to be asking questions that flow across all these systems to get answers."
To be sure, not every system will need to be integrated; approaches will vary depending on the size of company, the scope of the business problem, and the data requirements. But Bhambhri and others say the overarching goal should be to create an information management architecture that ensures data flow between systems. To create this foundation, companies will leverage technologies like middleware, service-oriented architecture, and business process integration, among others.
In the meantime, traditional data warehouse architectures are also under pressure. Gartner's Beyer says that 85 percent of currently deployed data warehouses will, in some respect, fail to address the new issues around extreme data management by 2015.
Even so, he says, "we don't want to give the idea that rip-and-replace is even on the table." Instead, existing repositories can be expanded and adapted to encompass built-in data processing capabilities.