I work in the field and have seen companies who spend several million a month, agree with your key points.
For companies of that size, data is everything, if an ad is underperforming it stops running, if an ad is bringing you low quality clicks (for example, like farms) it will stop running too.
It's really silly to look at someone saying they dumped $20k in FB ads and it didn't pan out without providing any kind of data at all, how many creatives/targeting permutations did they try? did they monitor their ads at all? did they try to negatively target those bad likes they were getting?
I've even seen "custom" lookalike audience implementations, to squeeze that extra cost/performance ratio out of it.
Last year, we had great success with buying likes -- but for it to work you need to execute very well in terms of regularly publishing highly engaging content to your page. I'd love to compare notes some time. Send me an email if you are interested. (Google my name.)
We spend way more than $46K/year.
I agree with Jacob here. If you are willing to work at it, facebook can give you a very healthy ROI. It is not nearly as easy as with Google though.
- Buying likes is basically a waste of time unless your goal is to stroke someones ego or are trying to generate some sort of "social proof"
- Re-targetting via Custom Audiences is awesome (you need a robust CRM program to take advantage of this, or at least an email list).
- Lookalike audiences are promising, but you have to work at them.
Edit: paragraphs