Advertising junk food and pickup trucks to someone that "can't think of two products I care about less" isn't an error. It's an attempt to make you care (if possible), and reinforce the brand so their name is familiar= if opinions change about junk food or pickup trucks in the future.
This sounds just like what people in the ad industry would tell worried clients. The actual reason is likely to be more mundane: the guy is, according to the tracking database, in an age and income group with high statistical correlation with interest in junk food and pickup trucks. Not all targeting is as detailed and personal as everyone thinks after reading scary articles.
Because the platform knows ads can be creepy, and will result in user disengagement.
Facebook often accurately targets people, and the result is terrified users.
So, the answer is other platforms have taken note, and are providing cover for users, to retain them.
This augments the return on ad spend, but only slightly, since ads actually aren’t very effective in practice. Ads are simply signage, nothing more. They don’t actually change behavior, but only reorganize behavior that usually would have happened anyway, so that the existing outcomes possibly land in the finite set of buckets differently.
Thus ad platforms only need to target so vigorously, and with that, it becomes obvious that targeting can be made less acurate, and buyers of ad space would never know.
So you can quickly see that Facebook’s game is absolute overkill, and other players likely underkill by a margin, with much improvement to their platform reputation.
But even if, doesn't this just bring you back to the original point of the blog post? Just advertise to everyone, no data mining and ML required. Best case, you do indeed change my mind about your brand, and the stereotypical junk food consumer sees the ad, suddenly feels hungry so hops into their pickup truck heading to the nearest McDonald's.
My interests, according to Facebook, include:
- "product (business"),
- "Price",
- "Providences and Territories of Canada",
- "Livestock", "experiment",
- "Lizard" and "Shark" (as a hobbies!),
- "USB".
As far as I know, I neither own nor operate a battery-powered reptile farm. Honestly, reptiles kinda creep me out.They do get a few things correct, but none of these seem "inferred"--it's mostly organizations that I've explicitly liked and...substrings of their names.
I was reading a dark fantasy book series and all the similar items were romance novels…
It doesn't matter if their targeting algorithm is wrong. It just has to be 1) better than random and 2) not significantly worse than anyone else's.
In fact, even that isn't necessary. It just has to be perceived as those two things by those who are willing to give Google or Amazon money to try to improve their chances of selling you stuff.
If I want to target 18-25 year old males with make-up products then no algorithm is going to save me from myself.
I'm sure if your Twitter feed was full of automotive posts you'd see car ads. It's probably full of stuff no advertiser cares about, like your love of old British railway cars, so instead you get generic ads about spicy nacho cheese.
As a railway enthusiast myself, I would say that's actually an opportunity for very effective targeted marketing:
* There are loads of specialized reference and historic books and video on the subject.
* The crossover interest to model railways, a hobby involving many high-margin purchases, is very high.
* Many enthusiasts literally plan their vacations around preservation railways and museums. Everyone from the tourist-railpass sellers to full-fledged package tour operators should be targeting them.
If that interest is not well advertised to, it's either because they're doing a poor job identifying the interest and routing it to advertisers, or they've done a terrible job of fostering ad buyers in those verticals.
But even on this platform I've spent almost 10 years curating the perfect feed on, the one that should know absolutely everything about me, I still get served endless garbage ads for junk food and pickup trucks. I can't think of two products I care about less.
Amazon is a similar story, they seem convinced that I'm a woman despite having some pretty detailed purchase history going back to 1998 suggesting otherwise.
The targeting algorithm is often wrong, they just don't seem to know it be willing to admit it.