As far as revenue goes, KU is obviously better for longer books. My novellas earn pennies in KU compared to what I earn for a full sale. In the next couple of weeks, I plan to crunch the data and write up a more detailed analysis on this KU experiment for my 2021 business plan.
It's also very genre dependent. From what I have heard, space opera is very much a KU genre.
Last point first, yes. Space Opera is a KU genre. So is LitRPG and GameLit.
About sales verses reads though, while KU does cannibalize sales, you can tap into a whole KU section of people who were never going to buy your book anyway. With these people, KU is a bonus income you were not getting with just sales alone.
The calculation is if the reduction in sales income is offset by the increase in read income.
And I'm using income deliberately. The people who use KU to binge read will pick up on whatever book is visible, read it, like it, then binge on your entire catalogue. This happens in sales, but not as often because of the high cost of buying.
It's the binge readers using KU you need to reach. Once you do, KU works.
On a book by book basis, the stats might say KU is not working. But on a catalogue level, the stats may be totally different. The only book I pay any individual attention to is the current one. Otherwise it's the daily totals which I track.