Wednesday, July 06, 2005

Data Modeling the FBI

Without knowing anything about the FBI, I've speculated on why it might have had problems with its big IT project. Part of the problem, I guessed, was a culture that gave priority to the special agent and not the administrative side.

One of the fun things about my time in USDA was learning about "data modeling", which is the process of analyzing processes to pick out the key data that you need to capture in the computer. Now this was in the days of relational databases (i.e., almost 20 years ago) and I'm sure the process has progressed since. But, that won't stop me from opining about things I know nothing about; I am a blogger after all.

I picked up Ronald Kessler's book, "The Bureau" which is a history of the agency, based on his past reporting for the Post and other books on FBI issues. There I find that Hoover specified the filing system for the agency early on: case files, with the designation the reference of the law violated, the abbreviation of the office, and a sequential number (not clear whether the number is within the office or within the law and office). That gives me just enough information to go to lots of speculation.

Observations: It makes sense to file by case--after all FBI's raison d'etre was to enforce federal law, so a violation is a logical starting point. The coding makes it easy to report to the main office--file case folders by status (under investigation, referred for prosecution, closed, etc.) and just do counts by law and status--so many bank robberies, so many Mann Acts, etc. etc.

But, in a modern context it wouldn't work at all well for intelligence investigations. (The FBI may have changed their filing system, but I'll assume not.) Note that the system assumed the field office is the alpha and omega, so it is awkward to transfer a case to another office or the main office. I think starting with the violation would make it hard to investigate organized crime, drugs, or any network.

Note--plan to do some more research and post more on the subject.

No comments: