Integrating cloud applications with remaining on-premise BI systems isn’t easy, either. CIOs who dawdle when weighing these concerns will likely lose control. Yet the IT group was heavily involved in just 38 percent of off-premise BI purchasing decisions, compared to 57 percent of traditional BI projects, according to a recent survey of 400 business intelligence users by Aberdeen Group. “CIOs have to get in front of it,” says Mark Popolano, a senior advisor at Kurt Salmon Associates. Popolano was formerly global CIO at AIG and global CTO at Reed Elsevier.
Spelling out service levels and contingency plans can be more difficult with analytics than with, say, storage, according to Kathleen Barret, president and CEO of the International Institute of Business Analysis, a professional association for business analysts. Analytics systems are inherently speculative. A marketing manager may not know what data is most useful until she is deep into her queries. With an on-premise system, she can ask IT for ad-hoc access to something new. But in the cloud, says Barret, if the data isn’t covered by the contract, new terms may have to be negotiated.
Data security must be defined for an outside provider, then monitored, Popolano says. He suggests one possible approach: Hire cloud providers for periodic projects where extra processing power is needed. Then stipulate in your contract that the vendor delete the data and related queries after the analysis is complete.
Integrating on-premise and cloud-based analytics will be a big project this year for Colin Zvaniga, CIO at the Ontario Association of Community Care Access Centers. Last year, the organization deployed a private cloud to provide a common electronic medical records system for its 14 members. Each member still does its own analytics on its own data, using Cognos tools. Zvaniga plans to deploy Microsoft SharePoint in a private cloud so members, whose combined revenue totals $1.8 billion, can also query their collective data.
“If one center wants to understand something about their business, 95 percent of the time, it’s applicable across the board,” Zvaniga says. For example, they might develop more effective wound-care plans by studying geographic, demographic and outcome data from all the centers.
The hard part will be melding the analyses produced in the cloud by the new tools with those from the local Cognos systems, he says. “We haven’t figured out how much of that [integration] can be automated and how much will be up to the on-site experts to interpret for themselves.”