So the heuristics you propose needs to be extended, somehow, or maybe replaced by a possibility for users to manually propose similar modules, ideally with a field to describe similarities and differences of the two modules (that would be really useful to make educated decisions before buying)
-- mcleinn
Yes, you're right - manual suggestions of similar modules would probably be far more accurate, given enough votes. If everyone's suggestions were recorded independently, too, then you'd have a measure of strength of similarity (number of suggestions).
I agree with your point about similar modules sometimes appearing in the same rack, however, I think given general scarcity (very few people have the resources to buy EVERY module they want, and so choices are going to be made between similar modules), it would still work to some degree. The question is just whether the signal would be stronger than the noise.
But yes, as I noted in the OP, it's definitely a partial solution, and I was hoping others would come up with additional constraints that might make it more useful (or even come up with entirely different and better options for an automated solution!).
Your particular example might be mitigated by putting higher weight on "similarities" found in smaller racks - if a rack is very large, then there's more chance that you might find a Maths and a Rampage. I doubt those two modules would ever appear together in a 1U 84HP skiff :)