Comments on: FBI abandons old computer system, tries again

After pulling the plug on its failed Virtual Case File system, the FBI tries again with a new project called Sentinel.

[Mendz]

We learn from mistakes...

Especially if it costs millions of dollars... :-)

[Len Bullard]

Not Necessarily

If you learn but the decision maker doesn't, the system is no smarter.

Preparation of paper-based forms from digital sources (going from a QBE form to a report) is the dominant interface metaphor and real output for almost every case management system sold in public safety. Why?

The systems are designed to support the preparation of administrative statistics (say NIBRS/UCR), not the gathering and dissemination of case facts and conditions. In short, the systems are very good at getting information in and terrible at getting it out. In most cases, the facts needed to put the dots together are in the unstructured and opaque supplemental reports attached to the case records. Some advice to procurement officials: don't approve the purchase of any records management system for public safety that doesn't include at least a rudimentary version of these features:

1. Ad hoc cross table querying capable of creating RSS or HTML formatted reports. Prefer RSS/Atom.

2. Unstructured search capabilities (see the UIMA architecture from IBM (No I don't work for IBM)). You need this to get into the supplements and other attached BLOBs.

3. Geolocation-based aggregation. Think pin maps with user defined queries, but for your own protection, don't a) assume you can do this with google maps and b) do try for real time 3D.

4. Work with the local, state and tribal agencies to get messaging interop. You have some of that now, but it isn't consistent and the cruiser installations aren't as good as they can be. Start eliminating the middle vendors.

No administrative management system or dispatch system ever solved a rape case or detected a terrorist in real time. Pre-real time analysis is hard but if you are to better focus your surveillance systems, this is a must have.

[Mendz]

We learn from mistakes...

Especially if it costs millions of dollars... :-)

[Len Bullard]

Not Necessarily

If you learn but the decision maker doesn't, the system is no smarter.

Preparation of paper-based forms from digital sources (going from a QBE form to a report) is the dominant interface metaphor and real output for almost every case management system sold in public safety. Why?

The systems are designed to support the preparation of administrative statistics (say NIBRS/UCR), not the gathering and dissemination of case facts and conditions. In short, the systems are very good at getting information in and terrible at getting it out. In most cases, the facts needed to put the dots together are in the unstructured and opaque supplemental reports attached to the case records. Some advice to procurement officials: don't approve the purchase of any records management system for public safety that doesn't include at least a rudimentary version of these features:

1. Ad hoc cross table querying capable of creating RSS or HTML formatted reports. Prefer RSS/Atom.

2. Unstructured search capabilities (see the UIMA architecture from IBM (No I don't work for IBM)). You need this to get into the supplements and other attached BLOBs.

3. Geolocation-based aggregation. Think pin maps with user defined queries, but for your own protection, don't a) assume you can do this with google maps and b) do try for real time 3D.

4. Work with the local, state and tribal agencies to get messaging interop. You have some of that now, but it isn't consistent and the cruiser installations aren't as good as they can be. Start eliminating the middle vendors.

No administrative management system or dispatch system ever solved a rape case or detected a terrorist in real time. Pre-real time analysis is hard but if you are to better focus your surveillance systems, this is a must have.