Database: What's In a Name?

Much more than bank identity management systems can handle

Its somewhat incredible that fifty years into the digital age, large data-intensive, data-driven organizationsfinancial institutionsare still confronted with questions about the integrity of basic customer records.

Ironically, the question is not whether the systems are corrupt or unstable. After all, enterprise core data systems are generally awesome in scale and efficiency. At issue is something more fundamental: knowing whats in there. What those records are and what do they say.

To hear independent analysts and software vendors (with a financial interest in saying so) tell it, the quality of data is getting worse, not better. They may have a point. Even at a time when Google makes scouring the universe happen at the touch of a button, its still a very tough job to isolate, clean up and keep accurate the duplicate or misstated customer records in financial databases which have been cobbled together, all using different standardsin many cases from the 1970s.

Today there is a rapid consolidation afoot among database, search and identity management-oriented software businesses. Perhaps this will spur program developers to someday build even better identity management systems with data-scrubbing components. But in the meantime, it pays to pay attention to basic search. This is not only a regulatory issue, with officials demanding that banks thoroughly know their customers, but its basic finance as wellhow much does it cost to send out duplicate marketing pitches to the same person?

Product names, addresses, descriptive textual informationits all very different from the other sorts of data that is stored, manipulated and processed in computer systems, says Ramesh Menon, head of North and South American operations for Greenwich-based Identity Systems, which is part of Nokia. Identity data cannot be easily validated. There is no spellchecker for names that can tell you Ramesh Menon is a valid name but Menon Ramesh is not.

As even Google showsits the best mass search tool in the world and a fairly fallible onenon-intuitive, letter-based data sorting is very hard to get done precisely. It takes repeated and diligent measures in order to get it done at all, much less done to a very high standard. Identity Systems experienceand theyve got 60 or 70 FIs as clients, including Goldman Sachs, Citi, Morgan Stanley and GE Capitalis that about four to six percent of customer records on an enterprise database are duplicates or symptoms of some other kind of error, files stranded there on server farms like dusty fossils in the desert.