I wonder if your AMS success is based on KU, Bill? I've always wondered if my failure is due to me being wide and not offering subscription price.
Well, remember that I don't have a positive ROI, so "success" is a very relative term here.
AMS has always sold books for me, right from the very beginning (which was also roughly the same time KU started). The ROI has gotten worse as the field got more crowded. I don't think I ever had a positive ROI, but in the beginning, the loss was much smaller, and it was possible to write it off as an investment in the future.
Unfortunately, as I get older, my work day gets shorter. So, much as I'd like to try putting as much time into it as Anarchist does, if I did that, I wouldn't have any productive time left to write. I have been trying to monitor more carefully. Typically, I kill ads that don't perform. I add negative keywords when inappropriate search terms are leading to hits. I kill keywords or products that produce lots of hits but no few or no sales. Sometimes, that helps. Sometimes, it doesn't.
Books in smaller niches come closer to having a positive ROI. I assume that's because the genre-type searches lead more directly to interested buyers. In broad genres, you inevitably draw in more people. But even subgenre terms (like urban fantasy) are defined differently by different purchasers, and those are big umbrellas with lots of different approaches. Ask someone to distinguish among urban, contemporary, and paranormal, and you'll get lots of different responses. Ask someone what mythology is, and the definitions will be much closer. Hence, ads for my two retellings of myths consistently come closest to positive ROI, sometimes actually achieving it (fewer sales, but far fewer extraneous clicks). One of my education books used to do the same thing, but I notice it's not doing much of anything now. (I don't know if Amazon does this on purpose, but after a while, ads don't accumulate any impressions even if they're still listed as serving. Creating a more or less identical ad solves the problem.)
To your original question, KU is the gateway to making more on Amazon. It's consistently more than half my income, with a few recent exceptions. And the current calculations don't show projected royalties from it as part of ACOS calculations, which would look somewhat better with it included. But the raw numbers don't look that impressive. This month, as I mentioned, about 75% of my sales are from ads, but the ads only claim responsibility for about 10% of my KU pages read. I don't see any logical reason for that discrepancy and think that the KU part of the system may be undercounting how many pages come from ads, but that's just a guess.
I always gross much higher in KU than out of it. But as we well know, your actual mileage may vary.
The other puzzle is the way ad performances fluctuate for no particular reason. One month, an ad will pull in lots of sales. The next month, it will pull in almost none. Does the competition shift that much? Does the type of people who see the ad shift that much? I have no idea, but the fluctuations are wild. I guess if I had the time to do something like Anarchist does, I might find out.