I dislike that interchange assemblies need to embed a link to the assembly in each part, which complicates things when an item might be part of more than one group or that some item might have been released and now needs to be revised prior to being included, even though zilch has changed except for the embeded reference.
I agree on the training part. I came across an assembly with a family table with about 130 columns of essentially gibberish. By adding named groups, this was reduced by nearly half by placing Y/N on the groups rather than Y/N on the individual components and features. How long would it last back in the original hands? Not 10 seconds. I exaggerate a bit.
One thing that would help is if PTC allowed for in-tree per-feature and per-component comments. Based on my view of parametric modeling as an interactive, GUI based software development environment for an intepreted program that happens to produce shaded images and drawings of parts, I have to ask - where can comments be placed? Other software places comments in-line with the code so one can see the comment and the code at the same time, but PTC doesn't, at least for parametric modeling. There are comments for some things, but they are hidden and so you can't know they are there unless you ask to see them. Mostly they aren't there, like for columns in family tables.
Dan, a family table is one of the better solutions. The only thing that occurs to me worth trying, and also just as much a start-over as the whole family table, is creating a prototypical clamp routing model (datum plane, points to guide the cable, maybe an axis) and use inheritance to duplicate it in clamp geometry models so that swapping unrelated will have the identical basis. That way, each clamp model can be managed independently, but still offer interchangeability.
Now to stop users from polluting the pool with free STEP models that are built badly and have no redeeming characteristics, but can arrive with hundreds of layers. Yay.
