[mochomocha]

> All of this money and effort to create something that has never scientifically been proven to work: Super-targeted ads, instead of contextual ads.

Have you worked in the field? Because I have, and I can tell you that ads targeting works. I've built some of these systems that people love to hate on HN. Hundreds of PhDs at Yahoo Labs, Google, FB etc have worked on this for decades and run thousands of A/B experiments. Are you saying that all these people are fraudulent / incompetent and that somehow the whole market cap of Google and FB combined (above 1 trillion dollars) is just a complete fraud?

Contextual advertising works, but much less than behavioral targeting. Anyone who has seen and worked on the data knows that.

Knowing that you just visited Best Buy website 10 minutes ago and searched for a camera is _much_ more relevant to figure out which ad to show you on nytimes.com right now than the content of the article you're reading on nytimes.com

[cgiles]

Speaking as a PhD, I can assure you that just because PhDs research ways to make a thing work does not mean it does. Nor are they, or those hiring them, immune to presenting their results in the most optimistic possible way.

I am personally acquainted with several active fields that have been trying to make things work for 20+ years with very moderate success. They present their results which amount to "barely better than nothing" in order to keep funded. They also make the same arguments you do, "it stands to reason that it should work", to keep the money flowing.

There is the drug industry, which is chock full of new drugs that are barely better than placebo or generics, if they are at all. The results are hyped because it keeps the cash flowing. It would not be entirely true, but nor would it be too far off the mark, to say that almost the entire drug industry is based on fraud and exaggeration. And that industry is far more transparent WRT data and superficially altruistic than the advertising industry. This example is perfectly parallel because no one denies advertising works, just like no one denies antibiotics work, but the difference between new and old drugs, just like new and old advertising methods, seems to be greatly exaggerated.

Google built its entire empire based on contextual advertising, not behavioral targeting. With Facebook you may have an argument, but Facebook has a very special dataset not available anywhere else, and I would also add that Facebook makes the same amount of money whether behavioral targeting really works, or if they've just convinced their advertisers that it does.

I am neutral on the subject of whether it works because I have never looked into it, but "lots of self-interested people say that it does" is not convincing, and the fact that such an argument is so frequently made makes me think there is no actual proof.

[mochomocha]

> Google built its entire empire based on contextual advertising, not behavioral targeting.

True. But why do you think Google goes to a great length to track you all over the web? They literally have thousands of engineers doing just that. If there was 0 value for Google in behavioral targeting, and given their monopoly on search and the great value contextual advertising already brings for search, they surely wouldn't bother with tracking.

[cgiles]

Assuming that behavioral targeting does not work, which I do not know, several possible reasons:

1. They started tracking when behavioral targeting seemed like a reasonable hypothesis, and keep doing it now to build a dataset in case it might work in the future.

2. Google does lots of things that are speculative and generate no revenue, and the general attitude in the industry is "why not collect all data we can because storage is cheap in case we can somehow use or sell it later". If we applied your overall logic to every part of Google, they would probably have 500 employees.

3. If Google can persuade advertisers that it works, they still make more money from them even if Google knows that it does not. Thus BT can generate value for Google without generating any for advertisers.

4. If everyone else in the industry also makes these claims about behavioral targeting working, any company not claiming to do it too would be at a competitive disadvantage.

Also, it is possible that BT "works", say, 1% better than CA. In that case it technically "works and generates value" but most reasonable people would say in that case that it does not justify the privacy tradeoffs or general hype.

[mr_crankypants]

Google built their empire on contextual advertising, but it's not the only kind of advertising they do these days.

Also, Google's revenue isn't necessarily derived from what ads provide the highest ROI; it's derived from what ads people want to buy. Regardless of what anyone at Google thinks about contextual vs. behavioral ads, it's in Google's interest to go nuts with behavioral ads simply so that they aren't letting all the other adtech companies keep all the behavioral ad spend uncontested.

[eitland]

> But why do you think Google goes to a great length to track you all over the web? They literally have thousands of engineers doing just that.

As far as I understand even as the British empire became smaller the number of people employed to oversee the colonies went up.

Bureacracy will find a way to generate work for itself and it would be a shame to have all those data scientist wasting their time on fixing actual problems instead of making reports about how smart the current system is ;-)

[skybrian]

Knowing that the user is not a bot is pretty important to avoid ad fraud.

This might be enough incentive to do some user tracking, all by itself?

[Reedx]

> ...track you all over the web? They literally have thousands of engineers doing just that.

Seriously? If that's true, most of them must be hanging out on the roof.

[Cpoll]

> Knowing that you just visited Best Buy website 10 minutes ago and searched for a camera is _much_ more relevant to figure out which ad to show you on nytimes.com right now than the content of the article you're reading on nytimes.com

The concern here is that you're just selling a camera that they were already going to buy. So the ad agency wins, Best Buy _thinks_ they win because they register a conversion, but you didn't actually create any value.

In my experience when the PhDs say "this doesn't work," the PMs say "that's fine, because we still get to say we have machine learning [insert other buzzword] and the customer thinks it's delivering value."

> Contextual advertising works, but much less than behavioral targeting. Anyone who has seen and worked on the data knows that.

I admit this is possible, and my gut feeling is that properly implemented targetted ads should be immensely effective, but theory isn't implementation, and I'm taking your word on it either way.

[hombre_fatal]

Driving the user back to BestBuy.com to convert into a concrete sale seems much more valuable than "well, they searched for cameras so they might come back one day and pull the trigger. Fingers crossed!"

Why wouldn't Best Buy pay for that?

I search things on Amazon all the time without checking out. Those aren't locked in as eventual purchases at all. There are even things in my Amazon cart as we speak that I probably won't buy. I'm often a mere teeter from pulling the trigger. Coming home drunk or being reminded at the right moment sometimes push me over the edge.

There's obvious value in giving me the right shove.

[mochomocha]

> The concern here is that you're just selling a camera that they were already going to buy. So the ad agency wins, Best Buy _thinks_ they win because they register a conversion, but you didn't actually create any value.

I agree. The industry is (too slowly) moving towards measuring the actual causal effect of ads to remove the correlation/causation leap that has unfortunately been the norm. But it's much harder to implement given the very fragmented ecosystem, and big players in a monopolistic situation at this point have little incentive to do so.

In reality, it's not uncommon for the causal ad effect to be 10x smaller than the claimed correlational effect.

[dwild]

> The concern here is that you're just selling a camera that they were already going to buy.

Are you sure of that? Why would he not buy it on Amazon instead of Best Buy? Why would he choose that Sony camera instead of that Nikon ones?

Ads made him choose that model, at that price, at this specific shop. All theses variables could have changed and he would have still bought a camera, but nothing guarantee that theses variables would have been the same and Best Buy wouldn't have that sale.

Best Buy did win, because someone else didn't.

That's all considering he was already going to buy that camera and it wasn't a poor impulsive choice, which sadly happens too often in this world.

[cwyers]

I mean, building a panopticon to show me things I've already looked at may be effective but it seems like it's not the most worthwhile use of hundreds of PhDs for several decades? I also appreciate how Google continues to pay to show me Pixel 3A adds days after I purchased it.

[ohum]

Perhaps the advertisements you see after a purchase are intentionally delivered. Maybe testing has found a consumer is more likely to value and keep (vs return) a purchase, or recommend it to their friends, when related ads continue to be served for $time.

[disgruntledphd2]

A corrolory to Hanlon's razor suggests that never attribute to intelligence that which can adaquetely be explained by stupidity.

He's probably in an audience for targeting, and the audience is presumably only updated every so often.

[user17843]

I don't work in the field. I think part of the reason why my perspective diverges from yours is probably that your profession is extremely intransparent about what actually happens behind the curtain. Obviously people will speculate and err on the side of caution.

I should probably clarify that I did not mean to say that it doesn't work. What I meant is that it doesn't really work out for the publishers. The parties who profit are just first and foremost the platforms, then come the advertisers, and the publishers come last.

That explains the 1 trillion market cap. And also why Criteo didn't lose money after Apple blocked them from following their users.

I wonder if you can provide me to a study about the difference between Apple/Safari users and Google/Chrome users when it comes to advertising effectiveness, because since Safari users can not be tracked, that means according to your statement the revenue from Safari users would be way less.

[soared]

What? Criteo is worth 40% of what they were prior to the safari change.

That is correct - My previous employer spends $30k/day on programmatic ads, and they spend $0 on safari. Worse targeting, worse performing.

[epiphanitus]

What's the case for building these kinds of systems in 2019? I would love to hear your perceptive on this.

If I were to play devil's advocate to this article, (and this isn't necessarily my perspective) I would say that FB/Google have already carved up the ad market, and even their businesses are under pressure from native ads on Amazon.com, which is where a huge portion of e-commerce takes place.

The twist though is that this is the Washington Post we're talking about, so this plan may have benefited from the business mind of Jeff Bezos himself who knows these businesses very well.

[mochomocha]

> What's the case for building these kinds of systems in 2019? I would love to hear your perceptive on this.

The business case has become very tenuous at this point. Because of the dominance of Google and FB + the increased pressure from regulators on privacy, most third party companies are getting crushed. GDPR increased the market position of Google which also gives them less incentive to do behavioral targeting as it's legally riskier, and they don't need it to beat the competition on ad spend as they control publishers inventory. So they'll probably push more in the direction of AMP and controlling content.

What I expect in the coming decade is a regression to "the old world" of TV advertising: ads are going to become spammier and spammier as it's going to become increasingly difficult to collect and use data to make them relevant. And big players won't have an incentive to do so because they'll control content even more.

So I expect even more HN hate for ads :)

[dontbenebby]

>Are you saying that all these people are fraudulent / incompetent

This sound vaguely like an appeal to authority - or perhaps I'm not using the correct phrasing, but the underlying logic of "X number of scientists can't be wrong" isn't a good look. While I hate when people say things like "evolution is just a theory", it's not a good idea to use # of scientists who hold a view as a proxy for if it's valid if the idea is much less tested than things like evolution or climate change.

[mochomocha]

Well the original claim was extremely grand. I was trying to bring some nuance to the typical HN comment on ads which usually goes along the lines of "oh I saw a terrible ad yesterday, Google is so dumb and evil and ads don't work based on my own anecdote".

What I'm saying is that maybe it's not unreasonable to believe that some of the most successful tech companies of these last decades didn't completely build their empires on sand and that maybe what they sell is not 100% BS.

[reitanqild]

> Knowing that you just visited Best Buy website 10 minutes ago and searched for a camera is _much_ more relevant to figure out which ad to show you on nytimes.com right now than the content of the article you're reading on nytimes.com

This sounds reasonable. But there is quite a way from that to the current implementation which seems to be along the lines of:

  if(male) { 
    showDumbDatingSiteAds();
  }

It doesn't take very many PhDs to come up with that, does it?

[JohnFen]

> Contextual advertising works, but much less than behavioral targeting.

There is credible evidence on both sides of this particular debate. But to me, that's not a relevant issue. The issue is what is right vs what is wrong, and behavioral targeting is just wrong.

[reitanqild]

> Are you saying that all these people are fraudulent / incompetent and that somehow the whole market cap of Google and FB combined (above 1 trillion dollars) is just a complete fraud?

TLDR: despite your models the result I see as a male in Norway are ridiculous and laughable.

Longer version: All I can say is that the last 10 or so years the ads I have gotten have served no other cause than to make me despise the advertisers and distrust any claim about working AI.

Maybe I'm extremely unusual, but I would think this

- pushing mostly crappy dating site ads

- for 10+ years

- to a married man

- who is generally very happy

- and has small children

- while not pushing ads for major infosec events

- and not for family holidays

can hardly be considered very smart unless you are paid handsomely to show those ads regardless of if they work or not, especially as I've repeatingly clicked x -> not interested -> irrelevant.

Seing how Google is botching search accuracy, Google+, Reader etc etc I feel I have good reason to think the company in sum is a lot dumber than the people who work there.

And yes, I think so highly of most men in my exact position that thet won't fall for the crap I've seen.

And, based on my talking with others it seems not to be a fluke: if you are man this seems to be their best (and often their only) idea of what ads might be useful.

Or maybe it works on enough other men to pay off.

But personally I've now concluded that my best explanation is that Google is dumb or deliverately fleecing their advertisers.

PS: Facebook ads have been somewhat better for me. I actually bought something from one once.