I too found this thread at the perfect time.   We're rapidly approaching go-live for our initial release (small - incident & knowledge).   We are moving from Remedy and very poorly implemented Op Cats and Prod Cats with very very poor CMDB.

When implementing Remedy, I wrestled with the same basic issue... there are no hard and fast rules for CATEGORIZATION.   It is somewhat bounded technically, but really is an area of configuration that can be used flexibly, based on the requirements of your organization.   THAT is where the difficulty creeps in.

I find myself trying to get key users within the IT community to agree on what they want from incident management in terms of data about the work we deliver to keep the organization running.   The opinions vary from IT discipline to IT discipline... the infrastructure guys have one, fairly simple idea... while the SAP team would like cat/sub-cat to have provide a combination for 1000's of scenarios.   One for every conceivable issue they might have.

I'm prepared to make a call, as unpopular as it might be.   I've been coming to an approach very similar to what Ken advocates.   I've been trying to think of cat/sub-cat from the point of view of employees (callers) also... how do they think about issues they report.   This yields a relatively simple group of combinations - such as cats:   Accounts & Access, Collaboration & Communication, Application Issues, etc.   Sub cats within the Accounts & Access category might be password resets, password unlocks, etc.   Or issues reported for applications (cat), might have sub-cats as logon issues, data issues, transactions not working, etc.

Then - we use business services and ultimately CI's to provide a pretty robust set of data.   Couple that with closure codes on incidents (maybe we tweak those a bit) and you have a pretty good idea of the broad areas (types) of incidents you're dealing with, the specific business service and info about the resolution outcome.

My thinking is start more simply, make it easy (and quick) for the service desk agent to categorize and look at the data we have after 4-6 months.   Then we adjust from there.   I'm also looking ahead to tweaking the descriptions on Categories when we implement self-service and allow the employee (caller) to select the category at least.   (EXAMPLE:   "internal" category = Customer Hardware while the "What are you having problems with drop down = "My computer".   My computer maps to Customer hardware category and we go from there with the SD agent ultimately selecting the sub-cat)

To wrap:   Excellent discussion and info here... perfect timing.   Planning to read through and hit the links contained here again a second time to get it all to sink in.

Thanks to all.