Comment by zdragnar | Hacker News Reader

Comment by zdragnar | Hacker News Reader

Parent Full thread

1) The correlation between the content people peruse and the products people buy is pretty poor. Take a news site with an article about school shootings- what related content would they advertise? Guns? Back to school supplies? Without knowing why a person is on a page, it's difficult to understand what related products might appeal to them.

2) Tracking user behavior is fairly important to implement click fraud detection- obviously that doesn't justify shady behavior on the part of ad networks, nor does it necessitate the in-depth amount of personally identifying information they may or may not collect. However, some forms of that data- think IP addresses, click history- may be useful.

3) If you already have some data per point 2, why would you waste an opportunity to display something meaningful by putting up an ad that might be completely irrelevant given what you happen to know already?

None of this is meant to justify targeted ads, only to explain why they have better appeal to both content providers and ad networks.

My instinctive response is always: 1) people paid for untracked content-related ads in newspapers for what 200 years or more. Should be possible now too (it might support just a hundred publishers rather than ten million sites - but so be it).

2) in a vast global experiment where advertisers knew everything about everyone they still completely failed to show people relevant ads. Most notoriously even the giants who know what you ate yesterday fail to realize you probably don’t want ads for toilets for a year after you bought a toilet.

There's actually quite a bit of independent auditing involved in making sure that newspaper ads are actually shown to the number of people that the newspaper claims they are, especially for free papers where those still exist.

>it might support just a hundred publishers rather than ten million sites - but so be it.

You want millions of people to loose their lively hood over this issue? That sounds cruel and spiteful to me.

You also have to think of many small companies that make niche products that would no longer be able to find their customers and thus be forced to close.

> 2) in a vast global experiment where advertisers knew everything about everyone they still completely failed to show people relevant ads.

The premise of your statement is false, advertisers have never know every thing about anyone. That was just marking bullshit by ad-tech companies targeted at adversities and publishers, "We know everything about your users! So if you hire Acme Ad Tech your profits will go through the roof!" The fact that users hear the statements and understandably freaked out was collateral damage. The fact this statement was demonstrably false was irrelevant.

The fact is that the amount of data used to do targeted advertising is quite sparse in-fact, no-one wants to spend the effort to individually go through mountains of data and figure out the best advertisment. Really the only data used is

1. demographics (age, gender) 2. location (this is so a NYC based restaurant chain doesn't waste money advertising in California), 3. language (people actually hate seeing ads in a foreign language) 4. interest categories (like do you like Football, or Cars, or video games, this is how facebook advertises a lot they figure out your interests based on posts, then allow advertisers to advertise to you based on that category, (interestingly they don't actually sell the list of interests mostly because that would mean advertisers would not have to buy ads through Facebook and more)) 5. product targeting (this is you visited acme.com and put an acme roadrunner trap in your cart but never checked out so now it follows you around the internet.)

> Most notoriously even the giants who know what you ate yesterday fail to realize you probably don’t want ads for toilets for a year after you bought a toilet.

Yeah, that's my point, using more data (like the fact you bought the toilet) is more work and using the current amount of data works well enough for them so that is all they do.

I never even considered a million jobs as a factor in the matter of my privacy, and now that you mention it - I still don’t. My basic viewpoint is that if anyone needs your business you can monetize it tomorrow too, OR everyone is better of if you shut down.

> what related content would they advertise? Guns? Back to school supplies? Without knowing why a person is on a page, it's difficult to understand what related products might appeal to them.

Training for teachers? Security services? Access control systems? Yes, a lot of those reading are parents, not teachers, but what do you think they'll do if they are convinced something exists that will reduce the risk of school shootings?

Editing: ad "targeting" as it has existed so far has also been insultingly, embarrassingly bad IMO. Even Google tried for years to send me ads for Russian, Ukrainian, Philippine and Thai dating sites. Round robin. I was delighted when they started showing me Wordpress ads.

Those are not relevant either but now we are approaching something useful and not actively insulting (I'm very happily married and have small kids.)

I think you're right about the targeted ads. Many people have expressed similar sentiment.

I suspect that most advertising is actually driven by supply (that is, somebody wants to advertise X), not demand (somebody needing Y at the moment), and that's why targeting doesn't really work that well, and the disadvantage of untargeted ads (or targeted by content) is much smaller than advertising industry claims.

I think for a lot of businesses this isn't a huge deal because they're already targeting a specific market. If you know your market then you should know what will appeal to them and be able to advertise as such... High quality content plus high quality recommendations.

With regards to 2), there are specific exemptions in GDPR to allow this kind of data usage (specifically in 6(1)(f), this is the general interpretation of "legitimate interests", though the exact definition is up to member states' data protection agencies).

This was/is unclear, I actually remembering having an hour long discussions about this exact topic several months ago. This is the biggest problem with the GDPR it was hard to understand what could and could not be done.

The problem is, the "balancing test", a subjective measure of the users interest against the legitimate interest, may still cancel out the exemption.

It's completely subjective, which may or may not be a good thing depending on the mood of the regulator, court, and your opinion of rule by common law.

Reply on news.ycombinator.com