To add to Van's comment, a counter intuitive aspect of the intelligence business is that, in general, no collection or analysis is done unless there is a requirement to produce that intelligence. Collection and analysis resources are always constrained, so what gets collected and what gets analyzed is subject to those requirements and their relative priority. The requirements and their relative priority are set by various levels of the national security bureaucracy.

Now, there are two basic kinds of requirements (and this is simplifying things a bit) - standing requirements and ad hoc requirements. Standing requirements are anything that needs to be regularly updated on a periodic basis, which can be anywhere from continuously to every few years depending on the topic and situation. Some examples are military capabilities, orders of battle, indications and warning, etc.

Ad hoc requirements are one-time or limited duration deals. In a crisis, for example, you'll see a ton of high priority ad hoc requirements (and periodicity of standing requirements will be almost continuuous). When the crisis abates, those requirements tend to go away and return to "normal."

To add to the complexity, there can be standing collection requirements, but no standing requirement to put that collected information into a comprehensive, all-source product. Things get even more complex when one factors in the organizational and technical differences between the various collection disciplines (IMINT, SIGINT, ect.) and the agencies that support them. Now add in the various levels of classification and special access programs and caveats that restrict information dissemination.

That complex reality, combined with legacy stovepiping (still around, but somewhat better than it used to be) and completely inadequate tools for finding existing information on our networks, make intel an often frustrating business, particularly at the low end of the intel food chain. For those at the pointy end of the spear, who require information to do their mission, and who often don't have dedicated intel support that can translate their needs into requests, it's even more frustrating. On top of all that is the chain-of-command, which tends to severely frown on any attempts to get information from the source or bypass needless layers of "management."

See what a confusing mess the intel world can be?

Van mentioned that it's important to check to see if requirements have already been answered. That's true, but the sad reality is that in too many cases it's actually quicker and easier to simply task an asset to collect the information again.