We've looked at data like this. I'm trying to get examples from the community to supplement it. Looking at aggregate statistics isn't a great way to measure what's bothering power users.
I've saved a few searches where autocorrect gets too aggressive. I've run in to more than this but I don't save all since usually I just want results for my search immediately and don't feel like recording them:
I needed `+pyinotify` on this one to get the results I requested. I've come across this problem with quite a few other python packages but I can't recall them at the moment.:
Search: pyinotify thread
Result: Aggressive correction; wrong
`Showing results for inotify thread. Search instead for pyinotify thread`
Search: win 7 processor not +downclocking -gpu
Result: bad autocorrect
"""
Showing results for win 7 processor not +downclocking -cpu.
Search instead for win 7 processor not +downclocking -gpu.
"""
I have the impression that Google likes to collect data. If you are indeed interested in fixing the recent quality dip of your search product, example queries where plus (now quotes) are necessary can be automatically collected by inspecting all searches where quotes are included.
If these are too many, filter by those that had a recent similar search without quotes. I doubt this is necessary, though. Who puts quotes in their search string the first time?
If I worked at Google, I'd be interested in the growth of use of the plus-operator lately. It has become a necessity for nearly all technical searches I do. Make a chart of it. Show it to the people in charge of this recent trend. Google search has seriously gotten worse. Forcing quotes instead of plus highlights this problem because it's twice the work.