[roland00]

Bingo. It is not exactly better targeting ads. You can sell the same ad (from the same supplier) to the same customer (the person viewing the ad) and the supplier will pay more if they are convinced the ad actually has a higher return rate.

What is the quote again from John Wanamaker "Half the money I spend on advertising is wasted; the trouble is, I don't know which half."

Thus knowing more about their users and what they buy and the ad is successful 1 day, 1 month, 1 year in the future allows Facebook to get paid more.

[roland00]

Note this is a first order effect, what I explained above.

The 2nd order effect is that by making more money for the same ad, while your competitors would make less money from the same ad...you can either sit on that profit.

Or you can use your market share and guaranteed profit to make ads less profitable to competitors and this creates a positive feedback cycle where Facebook Ads get more and more market share but other competitors and platforms of Ads gets less and less, which in turn actually allows Facebook to charge more per Ad in the long run (but in the short run Facebook may want to decrease the price per Ad.)

[mikorym]

I think the key here as you mention is the "pull through". Previously they could tell which ads generate buzz. The more important metric is which ads generate sales.