5 Steps to Big Data Success
Big data is big right now, and for good reason. It can help organizations make informed decisions, cut waste, uncover fraud and solve problems. Government agencies of all sizes are catching on, using it for a variety of projects. Indeed, figuring out to how to handle data and mine it for usable information is top-of-mind for most agencies and organizations.
It’s especially useful for managers looking to reduce energy use in their data centers, said Alex Rossino, principal research analyst at Deltek.
“The Department of Defense, for instance, is looking at big data as a potential requirement for its Defense Enterprise Computing Centers,” he said.
However, despite the hype and press surrounding the topic, it’s still early days for most public organizations implementing big data. There are few if any road maps for chief information officers who are looking to create a big data strategy from scratch, especially around data center resources and management. Here are five steps you can take to propel your big data strategy forward.
1) Define the mission objective and scope
Determine what’s going to be accomplished and whether the project will be a one-time program or the foundation for something that’s going to be expanded and used on an ongoing basis. Collaborate with line-of-business stakeholders about what they want to accomplish. If your organization is looking to use big data to reduce data center energy costs, will you open up your tools and data warehouses to other departments?
“You can’t even decide if you are going to get analytical capabilities from the cloud or if you’re going to create on-premises data warehouses using your own storage, software and tools” without a game plan, Rossino said.
Also look at your agency’s data management policy. Ask what kind of access it gives vendors, what you’re willing to share, how you’re willing to share it and how data is secured, Rossino said.
“Data management policies are areas where agencies have fallen behind,” he said.
2) Develop the business case
To secure funding, provide the return on investment for the project. This could mean how much money you expect to save or how much waste, abuse or fraud you expect to eliminate. In the data center, you’ll need to gather statistics on current and expected energy use and management costs, and extrapolate how much you can save by becoming as lean as possible.
“ROI can also be squishy, too,” Rossino said. “For example, we need this cancer data analysis tool in order to find a new cure for cancer. What’s the ROI for that?”
3) Outline the technology requirements
The amount of data stored is doubling every 18 months, according to a recent report by the Aberdeen Group. To handle such growth, you need to know what storage and computing resources you have and what data mining and analysis tools are already in place. In addition, understand what kind of data is already being used and what will be analyzed. For instance, are you mining existing information or dealing with unstructured data?
One technology element that is often overlooked during the evaluation is a data quality tool, which allows organizations to ferret out data mistakes and inaccuracies and make data easier to analyze and use. Companies that use data quality tools find that their data accuracy is 20 percent higher and the time spent looking for data is 33 percent shorter, according to another Aberdeen report.
Extract, transform and load data tools are another technology to consider. These tools extract data from their sources into a single data warehouse. Organizations that use ETL tools see a 33 percent faster integration from new data sources and save up to two and a half years on integration, according to Aberdeen.
Finally, information technology groups must consider data life cycle management and archiving. The more distributed the data is, with only the most important data on the more expensive, highly available storage devices, the faster access and recovery are. The average time to recover archived data goes from 44 hours without life cycle management to 20 minutes with it, according to Aberdeen.
4) Solicit information from vendors and providers
Issue a request for information on the General Services Administration’s Federal Business Opportunities website, FedBizOpps.gov, detailing what your agency is looking to accomplish, a statement of work and delivery dates, Rossino said.
“You’ll need to say not only what you want to do, but also what you want it to cost,” he said. “There are dozens and dozens of big data vendors out there and even traditional vendors are adding big data capabilities, too, so you’ll need to be very clear in the market research phase.”
5) Evaluate your options.
The final step before implementation is circling back to the mission objectives. Do the responses meet the criteria of the mission? At this point it can be helpful to go back to the line of business and have them look at what’s being proposed because one of the worst things that can happen is handing data scientists, users and analysts a tool that they can’t or won’t use.
Explains Rossino: “If 70 percent of [government] IT budgets are being used to maintain legacy systems and you use precious funding to support a project that fails you’re going to have a problem.”