Experts Urge Restraint When Rolling Out Enterprise Information Integration
Some experts worry that EII—which helps enable painless connectivity to disparate data sources—has the potential to be easily abused.
Enterprise information integration (EII) is a surprisingly divisive subject. Reasonable people concede that EII has indisputable value as a complementary data integration technology, but reasonable people also worry that EII—as a mechanism which enables painless connectivity to disparate data sources—has the potential to be easily abused.
Consider the views of one reader, a BI veteran with a prominent insurance and financial services company, who says he has current and deep experience with EII. This BI pro bemoans the lack of discipline which so often attends EII roll-outs, and says a laissez faire approach to EII could result in data management disaster for many organizations. What could happen? According to this BI stalwart, EII could undermine the evolution of a well planned data architecture, unless a high degree of rigor, governance and restriction is enforced; and, furthermore, that if such a degree of rigor is pursued, the maintenance overhead will severely limit any ROI.
What’s the attraction of EII? For one thing, it promises quick access to more timely data than the traditional data warehousing model. Philip Russom, senior manager of research and services with TDWI, has argued that frequently refreshed reports—exposed, for example, in the form of dashboard or operational reporting solutions—might eventually comprise a killer app of sorts for EII. Frequently refreshed reporting can help bolster time-sensitive applications—such as just-in-time inventory management systems, or yield-intense manufacturing efforts—Russom argues, and might also “bolster the data access and integration capability of BI platforms that refresh reports directly from app databases.” In this model, Russom suggests, “EII is not married to ETL, although it's an alternative to ETL in a very short list of situations.”
Russom emphasizes rigor and restraint when it comes to EII adoption. So does TDWI research and services director Wayne Eckerson, who sees the potential for a Trojan Spreadmart of sorts in widespread—and undisciplined—EII adoption. “I'm beginning to wonder if there is such a thing as EII-spreadmarts. Like spreadsheets, a tool with universal connectivity makes it easy to rapidly build applications that probably circumvent architectural standards in an effort to meet business needs,” he comments. “In organizations already afflicted with spreadmarts, EII might be like adding gasoline to a fire: it could cause spreadmarts to rage totally out of control and burn everything of value in sight.”
That’s a perspective echoed by Mike Schiff, a principal with data warehousing consultancy MAS Strategies and a member of TDWI’s extended research collaborative. He cautions against the use of EII “straight from the tap,” so to speak—or in the absence of rigorous data management and data cleansing policies. “In theory, if you’re going to do EII you’ve got to solve the problems of data cleansing and master data management. You need some mechanism to make sure that you are indeed adding apples to apples and oranges to oranges,” he points out. “That’s the whole purpose of a data warehouse model.”
Schiff, like Eckerson and Russom, emphasizes that EII is a valuable technology. He cautions, however, that it can be easily abused. So does Mark Madsen, a data warehousing consultant and—like Schiff—a member of TDWI’s extended research collaborative. EII could potentially be “the final item needed to bring Excel to BI dominance,” Madsen suggests. “If EII isn't under the control of the data warehousing [or] BI team, then there's a good chance something like this could happen.”
In some organizations, Madsen points out, the data warehousing and BI teams are already effectively out of the loop. “I've noticed that DBAs in a lot of large organizations are charged with the care and feeding of replication and such. They tend to provide database views and data replication to different development groups where sometimes they shouldn't, architecturally speaking,” he concludes. “EII is one of those things that is probably going to land with the DBAs, but could just as easily land in the ETL group or the BI group.”
Stephen Swoyer is a Nashville, TN-based freelance journalist who writes about technology.