For some more direct evidence about how aggressively Facebook optimizes its content, we can also look at its content.
Here are a bunch of ads it showed me yesterday (prompted by me seeing a particularly compelling ad (#2) and thinking of this, then just noting the other ones starting at the top of the page):
- Gates foundation plans to improve math education: interesting, but not so much that I even read it enough to know exactly what it was saying. I’m not sure what it wanted me to do, and am probably not going to do it. Probably slightly changes my perception of GF.
- Something you put in your ears to make sounds less annoying: very well honed to my interests supposing it works, 50% chance I buy it in a few weeks, without checking if it works.
- A kind of exercise object that I searched for yesterday because I found one in my garage: maybe interesting if I didn’t apparently have one, may have still reminded me to use the one I have, which could lead to buying more things from them.
- 6 different ads for political candidates: I’m a foreigner who can’t vote, so no dice.
- a kind of non-coffee hot brown water with low-key drugs in it: a kind of thing I’m unusually interested in. Pretty weird because while I have recently renewed interest in such things, I think it has been entirely manifest in verbal discussion with housemates in the kitchen in terms like ‘brown water’, and in physically combining various roast grain substances that I found in my cupboard with hot water. I probably won’t buy it now, because I apparently have a lot of different brown waters in my cupboard, but decent chance I remember it and search for it in future.
- New Oculus Quest, a VR machine that my house just got so I just tried for the first time: Again, unlikely to buy since I sort of have one, but plausible that prompting me to remember that it exists or think its cool is still helpful to them. I don’t remember searching for it, but did search for whether The Witness could be played in VR recently.
- Another kind of obscure new drink containing mild drugs, this time alcohol-themed though probably not containing alcohol: Again, great try, though did not click.
- Humorous T-shirts: not funny to me, and hard to imagine nearby scenario where I buy humorous t-shirt
- Borat movie: advert suggests to me there is very little chance that I will like. Some chance I would so aggressively dislike that I am tempted to watch it to explore the nature of my dislike, but not that probable. Does increase my perception that movie is a big deal, so maybe that helps somehow.
- The National Academy of Sciences is looking for nominations for a prize, in an ad that I would have missed if I wasn’t looking for ads, and my brain still wants to miss having noticed it: I don’t really know what I’m being asked, and am not curious enough to find out.
- At home fitness program that I probably would have skimmed over, but I guess it looks maybe interesting, now that I am really attending to looking at ads: unlikely to click on or buy, but if I want such a thing in future, may remember that this one exists. Though not sure what it was called, so may have to search ‘home fitness videos sexy trainer and normal people exercising ad’, which doesn’t seem promising.
So we have:
- very well targeted and I don’t have: 2 (#2, #7)
- well to very well targeted but I have: 3 (#3, #5, #6)
- kind of interested to see, but not going to do anything: 3 (#1, #10, #11)
- not at all likely, but it’s subtle: 2 (#8, #9)
- basic demographic info would tell you there’s no chance: 6 (#4)
(As an aside, 3/16 seem weirdly well targeted to the point that I assume they are based on fairly specific surveillance, or would if I could imagine how (#3, #5, #6).)
This seems to me very high variance: a mixture of surprisingly well targeted and surprisingly poorly targeted things.
It also seems more successful overall than I would think, but probably I should exclude the one that prompted me to think of doing this, which was #2.
On the theory that most of the optimization is going into making me look at Facebook for longer, rather than the ads themselves, the ads shouldn’t necessarily be very optimized. But on a theory where the manipulation is broader, or where the ad-makers are also powerfully optimizing, or any theory where I am being strongly compelled to do the things ads want me to do, then they should be.
Probably this has been overall some evidence for the ‘advertising is well optimized’ hypothesis, though I’m not sure how good I expected it to be.
I do have the broad sense that Facebook ads are much better than ads in any other place. For instance, I remember buying the things they offer sometimes, whereas I don’t remember that for ads elsewhere except signs in the street, though it must have happened.