Every system is unique, so citing statistics for one won't apply very well to another. Finding the statistics in the first place is going to be a problem as well.
It's a bit of a catch-22. Discussing the effectiveness of one strategy versus another could easily lead to bad actors changing their communication strategies to avoid what is currently effective. It is, in that regard, not much different than the reason why we don't compare the effectiveness of various secret agents and spies and their techniques in public forums; if we know, so too do others. The very effectiveness relies, in part, in not being part of public knowledge.
Finally, I think one point (perhaps the most important in this regard) that you didn't address is that these tools aren't necessarily allowing anyone to do anything new, assuming they had the data in the first place. They've taken what intel analysts were previously doing by hand for many years and automating gathering the data and generating views into it. Both of those save a significant amount of time, but I have yet to see a system that generates something that a human:
a) didn't need to verify first by inspecting the source data to confirm, and
b) couldn't have come up with given time to do the job by hand
The products themselves are only novel in their packaging. The amount of analyses that can be performed has also increased due to the broader array of data that can be ingested, but again, that in itself is only a matter of scale, not novelty.