[troupo]

1. Those extraordinary claims need some extraordinary evidence

In a comment elsewhere in the discussion: accuracy of targeting is worse than random sampling https://news.ycombinator.com/item?id=43719816

2. Ads have existed for as long as commerce existed. Google became a trillion dollar ad behemoth before it started collected everyone's data by simply offering contextual ads.

Literally nothing in the ads business requires you to collect and sell so much of user data that it would even make Stasi pause and re-think.

[LunaSea]

> In a comment elsewhere in the discussion: accuracy of targeting is worse than random sampling https://news.ycombinator.com/item?id=43719816

I could provide a long answer but the gist of it is that the study is flawed. Among the reasons, they don't differentiate desktop and mobile traffic which is a massive measurement problem. They also use Nielsen DAR which is in itself a heuristic method of determining what age and gender a user is and thus is not a great pick as an oracle.

The study also does not mention click and bounce rates which are good proxies for targeting success.

Beyond the performance, the marketing and sales aspect of targeted advertising is also a strong selling point, no matter the performance.

> Ads have existed for as long as commerce existed. Google became a trillion dollar ad behemoth before it started collected everyone's data by simply offering contextual ads.

No, it didn't.

> Literally nothing in the ads business requires you to collect and sell so much of user data that it would even make Stasi pause and re-think.

It does because contextual advertisement does not provide enough volumes and lower performance (lower click rate, higher bounce rates, lower conversion rates).

Example: If 1/100 people read hockey-related content and out of those people, 1/100 pages read is about hockey, it means that you're reaching about 1/10000 page views.

Now if you do implement user tracking, you're available inventory is 1/100 page views.

[troupo]

> I could provide a long answer but the gist of it is that the study is flawed.

Show me a non-flawed study that shows you need vast amounts of user data and tracking, for each user, throughout their lifetime to deliver ads

> No, it didn't.

Yes, yes it did. The skyrocketing revenue is attributable to increased internet usage across the globe, and Google outright owning a large chunk of it.

> It does because contextual advertisement does not provide enough volumes and lower performance

Extraordinary claims require extraordinary evidence.

> Example:

Example: if you collect and sell vast amounts of sensitive user data without user's consent, and the outcome is indistinguishable from random noise, are you more effective?

Example: if targeted ads are found to be somewhat more effective than contextual ads, is the lifelong invasive tracking of every user action a preferred tradeoff?

(It's quite telling how people defending targeted advertising never address the elephant in the room)

[LunaSea]

> Yes, yes it did. The skyrocketing revenue is attributable to increased internet usage across the globe, and Google outright owning a large chunk of it.

Extraordinary claims require extraordinary evidence.

> Example: if you collect and sell vast amounts of sensitive user data without user's consent

The users do give consent and this is handled by Consent Management Platforms and passed in the programmatic advertisement auction chain in the form of TC strings.

The fact that you don't know this is also quite telling.

> [...] and the outcome is indistinguishable from random noise, are you more effective?

But it isn't, and if you are making claims, please provided sources.

Why would brands and agencies pay additional fees for data if they would provide no uplift?

> Example: if targeted ads are found to be somewhat more effective than contextual ads, is the lifelong invasive tracking of every user action a preferred tradeoff?

Are users prepared to pay for the difference?

[porridgeraisin]

> Are users prepared to pay for the difference?

If a service cannot be offered at a certain scale without such practices, it should not be offered at that scale. Before you start talking about how this enabled google's innovations, remember that the path we have taken to our current innovations is not the only path that could have been taken. By correctly squashing out immoral avenues like today's ad tech, we lay the path for the same innovations to happen taking a different, more ethical path. Sure, it could be that that would take more time and certain innovations would be delayed by an entire era[1], but note that we could also be going 5x faster than today w.r.t TPUs or whatever if we enslaved and forced enough people to work for Google's ML infrastructure team and nobody/nothing else. But we don't do that, do we?

[1] on the flip side, certain innovations may also come an era early

[LunaSea]

> If a service cannot be offered at a certain scale without such practices, it should not be offered at that scale.

That sounds like an opinion and one tha isn't shared by the hundreds of millions of users Google has o it's services.

The worst part is that I completely agree that advertisement and at the very least targeted advertisement, shouldn't exist.

Problem is that users prefer not paying in cash so companies find alternatives.

If advertisement or privacy really mattered for users, we would already have alternative Facebook, YouTube, etc, but it isn't and so we don't.

[porridgeraisin]

Well, I don't think most people really understand the way ad-tech supports these services. I wager if you asked a few lay men they wouldn't even know google is an ad company(apart from YouTube ads)

[troupo]

> Before you start talking about how this enabled google's innovations

To whit:

Most the big projects Google started or acquired, and that are still available today, Google started or acquired before targeted advertising. Google itself, Docs, Youtube, Cloud, Android...

Targeted advertising is not a requirement for innovation

[troupo]

> The users do give consent and this is handled by Consent Management Platforms

No. The users are tricked into giving "consent" through a plethora of dark patterns.

When dark patterns are removed, users refuse to give the information

> But it isn't, and if you are making claims, please provided sources.

Have you provided sources for any of your claims?

> Are users prepared to pay for the difference?

Still waiting for any source of your claims that:

- targeted ads are more effective

- ads require lifelong collection of any and all user data

- that without pervasive and invasive tracking ads are somehow prohibitively expensive

[LunaSea]

> No. The users are tricked into giving "consent" through a plethora of dark patterns.

Well you're free to bring this up to the various data privacy national organisations in the EU.

But I see that we're moving goalposts.

> Have you provided sources for any of your claims?

One of many available links on Google:

https://www.sciencedirect.com/science/article/pii/S016781162...

> ads require lifelong collection of any and all user data

Data collected has an expiration date. This is also part of GDPR compliance.

> that without pervasive and invasive tracking ads are somehow prohibitively expensive

We're now going in circles, but essentially you're ad performance (brand attribution, clicks and conversion) will determine your expected purchase price per thousand impressions (CPM).

If your targeting isn't competitive, your inventory will not be bought or at a lower price.

This already happens for Safari and Firefox impressions.

[troupo]

> Well you're free to bring this up to the various data privacy national organisations in the EU.

> But I see that we're moving goalposts.

We are not moving goalposts. You literally claimed that users give consent to data collection. No, they do not. They are tricked into giving this consent because that's the only way ad industry can obtain consent to livelong invasive tracking.

> Have you provided sources for any of your claims?

> One of many available links on Google:

> https://www.sciencedirect.com/science/article/pii/S016781162...

Let's see what the paper says:

--- start quote ---

Our simulation study reveals that more than 50% of audience segments (i.e., 23,407 out of 45,440 audience segments on the right-hand side of the dashed vertical line in Figure D2 in Online Appendix D) require a minimum increase in performance larger than 700% to be at least as profitable as no-targeting. ...we find that more than half of the audience segments require an increase in CTR0, CR0, and m0 larger than 100% to be at least as profitable as no-targeting.

Approximately half of the audience segments on Spotify require a higher increase in CTR0 for the advertiser, suggesting they might be less profitable than no-targeting.

--

By highlighting the questionable profitability of many audience segments, this paper aims to help advertisers decide whom to target.

We suggest using our model to calculate the break-even performance for many segments and then order these segments by break-even performance from smallest to largest.

... our findings also reveal that untargeted campaigns may yield higher profits.

Our proposed model has limitations, as it relies on inputs that may not be readily available to all advertisers

--- end quote ---

So. Questionable profitability, untargeted yield higher profits unless you ignificantly increase performance of targeted ads, a suggestion of an unproven model relying on data that may not be available to advertisers.

Yes, this truly is a paper that proves the amazing great efficiency of targeted advertisement, especially when offset against lifelong invasive tracking and wholesale trading of user data.

> Data collected has an expiration date. This is also part of GDPR compliance.

1. GDPR isn't available worldwide

2. Ad industry claims "legitimate uses" for a bunch of data and end up with "oh, we'll maintain your precise geolocation for 12 years" https://x.com/dmitriid/status/1817122117093056541

> If your targeting isn't competitive, your inventory will not be bought or at a lower price.

So what?

> This already happens for Safari and Firefox impressions.

Good.