Hacker News new | past | comments | ask | show | jobs | submit login
After GDPR, The New York Times cut off ad exchanges and kept growing ad revenue (digiday.com)
572 points by chrisxcross 3 months ago | hide | past | web | favorite | 452 comments

I honestly would not be surprised if the difference in effectiveness of targeted vs untargeted advertisement turns out to be none, negligible or even unfavorable to targeted. Sure, NYT is not enough data to be significant, especially when compensating for my confirmation bias, but I would really like to see more sites stop tracking, at least from an ethical standpoint.

I would be absolutely astounded if targeted ads did not provide significant long-term advantage to a big player like Google. How can it be that having intimate knowledge of someone would not allow you to sell them more stuff? John has a 10th wedding anniversary coming up. Mary is single and goes to the gym every day. Steve just got a promotion and likes the BMW 3 series. Michael is overweight but has just gone on a diet. Are such details of no significance determining what ads to show these people?

Having said that, I'm constantly surprised by how bad Google is at targeted advertising. For example, today when I visit nytimes.com I see an ad from Google with the ad text in French. Hey Google, despite my recent visit to Paris I don't speak French - maybe your AI experts could analyse my 13 years of Gmail and search history to figure that out!

PS: I'm a Google shareholder, so my confirmation bias is in the other direction :)

> How can it be that having intimate knowledge of someone would not allow you to sell them more stuff? John has a 10th wedding anniversary coming up. Mary is single and goes to the gym every day. Steve just got a promotion and likes the BMW 3 series. Michael is overweight but has just gone on a diet. Are such details of no significance determining what ads to show these people?

It's possible that some combination of efficient markets, semi-inaccurate/incomplete tracking data and chaos theory combines to make it mostly irrelevant.

So Steve just got a promotion and likes the BMW 3 series, but advertising to him is useless because he has already decided to buy a BMW 3 series. Whereas Larry has had the same job for ten years, and the same Ford for ten years, but if you put a luxury car ad in front of him it may get him to take a test drive and actually create a new customer.

The data says Mary is single and goes to the gym every day, but it only thinks she's "single" because she's in a committed long-distance relationship and isn't interested in dating anyone else. And she's a fitness expert who is willing to spend her time researching fitness products, so she already knows everything there is to know about those products, already buys the ones she wants, and advertising them to her isn't going to create any new exposure. Whereas Jane never goes to the gym, so you might actually sell her a gym membership or a piece of fitness equipment because she doesn't already have one or know anything about it yet.

John has a 10th wedding anniversary coming up, but he has also known exactly what he's going to do for years. Whereas he has a third cousin whose wedding is coming up and has no idea what to get, so you should be showing him ads for toasters and flatware rather than jewelry and chocolate.

In general, you may do better to advertise your stuff to the people who aren't already interested in and knowledgeable about it. Which actually looks kind of a lot like random scattershot rather than targeting.

There’s a bigger effect: context.

I don’t want to buy beef jerky when I’m reading about a military strike in the Middle East, no matter how much my profile indicates I like beef jerky — but I might be open to a book on politics.

Because reading about content is self-selected disclosure of interest, the NYT already has all the information they need to target me — they know I read politics, and when I’m in that context.

The only thing Google can provide is slight refinements on which political book to suggest — which isn’t far enough above the noise floor to matter. Anything else is just them giving the NYT statistical fuzz to pretend carrying their ads on beef jerky isn’t an all-around negative so they can fleece advertisers.

Targeted advertising isn’t about efficiency, it’s about raising the number of places they can (uselessly) place beef jerky ads so as to increase their cash flow.

>Because reading about content is self-selected disclosure of interest, the NYT already has all the information they need to target me — they know I read politics, and when I’m in that context.

This seems to me what "targeted advertising" should mean - NYT says "for all routes of nytimes.com/politics, show political ads. If it's local news, open it up to classifieds for local business." Etc.

If someone's on the page, is that not enough information?

To show the ad to the consumer? I'd say yes. To sell the ad, when google is selling hyper informative profiles down the road. Maybe not. Not that people who buy ads or keywords or whatever have the best idea of who they should sell it to

But also, if it's general interest that makes sense as well. Say an ad for BestBuy while browsing the NYT. That's one of the benefits of well trafficked sites - you can reach a pretty wide swath of the population.

In this sense I would like it if site took advertisement more seriously. Asking feedback, allowing personalization, selecting a few "endorsed" ads (as a quick and cheap Boolean review).

For Google and Facebook it doesn't make sense, but for a newspaper i would expect some selection on the ads they show.

BTW Stackoverflow does many of these things with their ads.

In other words: targeted advertising is about making ads cheeper for advertisers, whereas the traditional assumption is that the real estate is valuable and desirable. e.g. companies want their ads in the New York Times or the Superbowl. Targeted advertising is a way of spreading ad dollars out more, which ultimately is detrimental to large publishers as well as being unethical on the data harvesting side. (I choose to adhere to the ACM's guideline that collecting personal information without informed consent is unethical.)

It’s actually about raising the price for advertisers:

Instead of only political books being able to advertise on that page, they’re bidding against beef jerky and cars and so on. This raises the price for the ad slot by creating (false) demand.

Without those out-of-place, targeted ads there would be less demand for every ad slot, and people could bid less for them.

In theory it can go either way. If you can target ads better to get more return per money spent, then it is worthwhile to spend more on advertising. In the end more on advertising is going to result in more money to publishers.

In other words, you're better off with affiliate links and content marketing than you are with algorithmic banner ads.

Heck, even Google AdSense scans your site to match ads with the content (source: I've tried using AdSense). They just don't rely exclusively on the content for ad targetting.

for actual comparisons: https://thomashunter.name/posts/2019-01-09-generic-banner-ad...

For sure, the type of person who reads the NYT politics section in English is _already_ a fairly niche audience compared to everyone who uses the internet.

I think you came really close to the salient point, but then missed it: Targeted advertising is, as you said, really not that much more effective as context-based advertising, unless you're an ad-tracking behemoth trying to convince companies they need your product.

The whole surveillance capitalism nonsense we live under is a way for ad networks to justify their existence. If you're a company that has data on everyone, you need to convince people they need that data to accurately sell ads, and targeted advertising exists solely for that purpose. The amount of cross-site tracking that these companies are able to do outscales what any company could manage to accomplish themselves, or even for a smaller competitor to step in, so as long as targeting users across the entire Internet is believed to be a must-have, these ad behemoths stay on top.

This reminds me of a meme I saw once upon a time:

> Dear Amazon, I bought a toilet seat because I needed one. Necessity, not desire. I do not collect them. I am not a toilet seat addict. No matter how temptingly you email me, I'm not going to think, oh go on then, just one more toilet seat, I'll treat myself.

I've wondered about this, if it is kinda like coke commercials or car commercials. These ads aren't actually trying to sell you anything, at least directly.

It is easiest to understand it with coke. It is so prolific that people say "coke" instead of "soda". That we have "is Pepsi okay?" as a joke. Instead what these ads to is make you feel good for your choice in coke. Not to convince you to buy it, per say, but so that when you do buy it you get an extra kick from those sweet endorphins.

So is Amazon trying something similar? Perhaps the opposite? As in "Hummm... maybe I could have gotten a better one", causing you to return and replace (or just flat out replace). But that might be a weird strategy.

Show of hands. How many people have clicked on an ad and bought something directly related to the ad?

I was researching cameras. I saw very little camera ads in my research. After I bought a camera? Ads for the camera I just bought everywhere, for months afterwards. I have not once clicked on a ad and bought something. I maybe clicked on 10 ads in my lifetime and they were 75% stuff I THOUGHT I had no intrest in. I think targeted ads is the biggest con no of our age.

I wonder how much of the purported "success" of retargeting is just mis-assigned credit? You research a camera, but it takes you some time to make a decision. In the meantime, you get blasted with camera ads, which you ignore. Then you actually make your purchase. Even though the ads had no effect on you, there's a fairly high chance you still purchased an advertised model at a vendor that targeted you with an ad for the camera you bought (there are only so many of each). The advertisers may claim credit for that sale, even though in reality they deserve done.

Even more egregiously, maybe sometimes the advertiser's metrics have some reverse-causality: you buy the camera, then they show you the ads for it; but something in their metrics isn't properly modeling the sequence of events (e.g. they correlate retargeted camera ads shown per week with camera sales per week). Then they mistakenly take credit for a sale that happened before the ad was shown.

That's not to say such mistakes always happen, but I wouldn't be surprised they didn't happen fairly frequently. Given that measurement is hard and both adtech and advertisers are motivated to present success stories, and may not be too motivated to second-guess positive-sounding numbers.

Google should know when I've bought the camera. They can read my email, as I use Inbox!

> Google should know when I've bought the camera. They can read my email, as I use Inbox!

I can't even imagine the outrage that would happen if it emerged that Google was using its access to people's emails to measure ad effectiveness.

I have literally never done so. I have, however, mentally blacklisted brands that had obnoxious ads, and I have always had a really low threshold for that determination.

I hate ads (particularly the ones interrupting TV programming) and try to avoid products/services that are heavily advertised. Imo it is the hygienic thing to do but also with a spin it is also smart business if one is price sensitive, why pay the for the obscene marketing budget factored in?

Amazon affiliate links for books are the sole times I've ever clicked on an ad with the intention to buy the product behind it. And I can count the number of times I've done that on one hand.

Incidentally, those weren't targeted ads. Those were ads hand-picked by the person producing the content about those books. From that very small sample I conclude that ads relevant to the context in which they're placed are more effective than targeted ads.

Whenever I get hyper-repetitive ads (usually GM, Dodge, or financial products - seriously, is that all of the ad inventory in the network on some days?), it usually has the opposite effect; I just assume those companies' products and processes are as inefficient and wasteful as their advertising.

I will say that I'm thankful that Squarespace, Audible, and Skillshare support independent video and podcast creators. I click their referral links sometimes to check out new features etc, but would only ever sign up if I needed that particular service. Still, I think that form of repetition helps their mindshare in a positive way.

I did buy the Glif directly from an ad on Daring Fireball, albeit months later, and in combination with brand awareness from their free time-lapse / speed ramping app.

Haha, similar. I was shopping around for a Sony mirrorless camera a year ago. Ended up with the model I actually had in mind in the first place (A7S2 for its supreme low light capability), but shit am I hounded on the Internet either for Sony cameras or accessories. And there's no way I know of to get rid of it! And for what its worth even Facebook and Instagram (at least the latter, I only use on my tablet in the app!) showed me camera ads.

This is annoying and, when one thinks about the implication that everyone and their dog knows who you are, scary.

Would an ad blocker not get rid of most of it? I rarely see any ads these days with Ghostery running.

I still occasionally get ads for laser eye surgery, nine years after I had it done.

I have purchased products/services off of ads, but only because I was directly searching to buy those things and they came up as sponsored results. I'm not sure if that counts.

I have never bought off of ads otherwise and find that most of the time they are not relevant to my interests. I also started using an adblocker because of other reasons (mal-ads/autoplay) even though I previously did not do so because I felt text/banner ads were quite acceptable in order to help keep the lights on.

I've installed grammarly after seeing an ad for it >100 times on youtube. Did even briefly use it.

I do all the time!

Mostly clothing...things I didn't even think did existed

Even though Steve has already decided to buy a BMW 3 series, he most probably hasn't decided which dealership to buy it from. So a dealership advertising has a good chance of a look in.Same case for Mary, she knows she wants to buy a fitbit , but where to buy from can still be influenced by an ad.

Then you're sniping existing customers from other retailers, which is zero-sum when everybody does it, and it's possible for your competitors to tell when you're doing it and retaliate by doing the same. At which point you're both paying for advertising that only cancels out the competitor's advertising. A smart retailer is not going to be the first one to take a turn down that road. It's essentially iterated prisoner's dilemma.

And it may even be profitable to be the first to turn away from it when everybody else is on it, and instead use more of your ad spend for non-zero-sum customer generation. Then you get all the new customers to yourself while everyone else burns their margins fighting over the existing ones.

That may be the genius of targeted ads. Once one vendor starts doing it, the other vendors have no choice but to bid on the same target.

Winner: ad platform.

> That may be the genius of targeted ads. Once one vendor starts doing it, the other vendors have no choice but to bid on the same target.

Except that they do have a choice.

Suppose there are five competitors. Everybody knows Steve is going to buy a BMW 3 Series. The margin is $5000. Assuming everyone else is going to do what you're going to do, how much does it make sense to bid to advertise to Steve? If you bid $1000, you have a one in five chance of making a $5000 margin. If you bid $5000, you have a one in five chance of making $5000.

So the optimal amount to bid is $0, because you have the same one in five chance at $5000 if everyone bids $0 than if everyone bids $1000 or $5000, but then you have a $1000 expected value rather than a $0 or $-4000 one.

Now suppose the others aren't using the same utility function as you and so may bid a different amount and you get a bidding war. So the first thing that happens is somebody bids $100 expecting to make $5000 -- bully for them, until the second one bids $200. Everyone raises their bid to $1000 because that's what a 20% chance at $5000 is worth. If one of the competitors drops out, the others will raise their bid to $1250 because now it's a 25% chance.

Which means there's never any profit in it for the advertiser. If anybody plays then at least one competitor will retaliate and everybody loses a total of $5000 to the ad platform. If nobody plays, everybody gets an expected value of $1000. Being the first to quit costs you nothing, but being the first to play costs you and each of your competitors $1000. Who is going to play this game?

You are assuming the ads don't have any impact on where Steve goes to buy.

In reality what happens is marketing runs an ad for $3000 offering a $1000 incentive, Steve makes the purchase, and making gets to (possibly correctly) claim they netted $1000 in sales.

Those marketing people want to keep their jobs, so they are going to work hard to justify their existence!

> You are assuming the ads don't have any impact on where Steve goes to buy.

No, I am assuming that advertising works the same for everyone. If there are five dealers and fifty customers and with no advertising each dealer gets ten customers, then with each dealer spending $50,000 on targeted advertising, each dealer still gets ten customers. When each dealer spends the same amount they each get the same benefit, so they still split the existing customers five ways.

What if two dealers spend a total of $20,000 and the other three don't spend anything?

Where are the customers gonna end up?

What you are describing sounds a whole lot like the Prisoner's Dilemma. For the highest personal success we need to think about maximizing group benefit, rather than being completely self-interested. However, this is rarely the case when competition is involved.

I don't know if this effect has a name, but this sort of thing happens in every industry. I noticed in on a popular tourist trap boardwalk where every restaurant had a "barker" out front trying to tell you to come in. It's basically added costs for all businesses with more or less a 0 increase in net demand for the set of businesses as a whole, it merely redistributes demand amongst them.

As he already said, it's called The Prisoner's Dilemma. The situation where everybody defects (everybody has a barker) is worse than the situation where nobody defects (barkers are unheard of). But the nobody-defects situation is unstable, because if Store A has a barker and Store B does not, then Store A will be swimming in money and Store B will go out of business (or, in reality, be forced to hire a barker themselves).

Probably the clearest example I can think of is doing a no-adblock Google Search for "Coca-Cola." The first result is an ad, put out by the Coca-Cola Company. This might not seem to make any sense, since the first non-ad result is an identical link to the same website, but imagine what Cott Corporation or PepsiCo would do if they could by the first result for their competitor...

And then there is me going into the only restaurant without a barker, because it was the only one I could check out the menu of without being hassled :)

Yeah, can get to a situation where an ad platform would be making more out of a market segment than the retailers in that segment.

If I have a $20 margin, I am better off giving Google $15 and getting an incremental $5 profit than not advertising. Even worse if my customers have a lifetime value of $50 it may make sense to give Google $45 and not see a return from my spend until years down the road while Google gets the $45 right now.

This is the main argument for a tax on advertising. Much of it is zero-sum. It adds real costs and runs up prices while providing no real benefit to anyone.

You've basically described the wireless carrier scene for the past decade. It is indeed zero sum, and the arms race in advertising resulted in increased costs for the end consumer.

He's probably going to buy the BMW from the only BMW dealership in his vicinity. They tend to try to keep from stepping on each other's sales territory.

And it's not just about selling BMWs. Maybe Mercedes would like to place an ad in front of Steve? Or Rolex?

Maybe a payday loans company finds that people like Steve are less likely to be interested short term loans - so they avoid advertising to him.

That's exactly how it works for me.

I have been running my online business for 5 years now. I learn (and keep learning) all I need to know from the same, high quality source.

I get targeted all the time by adds on landing pages, email funnels, Facebook ads, etc. Some of these are legit, but thanks to my experience I can see that most of them are of poor quality.

They will never sell me anything, while they might sell something to someone that hates his job and is looking for another way to make money.

Can you tell us your source?

Ramit Sethi, at growthlab.com and iwillteachyoutoberich.com

But if you determine that targeted ads are worst than scattershot, the correct conclusion isn't to do scattershot, it's to invert your target group.

The issue is that this "target group" is typically "people that visited your website and stayed there for some time".

Unless you have a promo, you really don't want to advertise to them... But that is exactly what you end up doing, because you can't target "people that visited my website 6+ months ago".

Why not?

Actually, you can. This is called "spam".

When someone stops visiting your site, make sure to send them a email. Ask them to return! You have their email address, right?

The real issue imho is that there is no shared sales information. So you advertise BMWs to people who may have already purchased a car. I think that is a bigger issue.

This happened to me both times I've purchased a Jeep. I did a bunch of research on models and packages, price shopped a few dealers in my area, and then purchased a vehicle all over the course of a Saturday. For the next 3 months I get jeep ads following me all around the internet. I wonder how much revenue those completely irrelevant ads generated for google.

Don't you give Jeep your name and address when you buy the vehicle? Seems like they are missing an opportunity to improve their advertising.

Also Google could place a survey question instead of an ad, asking "Which of these products have you purchased in the past month?"

Serious question: is a person always the same on every site? Can they be reached with the same message in different contexts? The argument for targeting says yes, but then why is porn ignored? Real people, doing, ahem, real things for significant time, but no advertiser (as a rounding error on total advertising spend worldwide) wants to advertise on porn sites.

I think a message's context matters - and this is almost self evident from examples like ads for airline tickets being shown on plane crash articles. I think targeting is great, I really do, but I think the value of trusted brands is likely just as, if not more, strong.

If context matters, and I think that argument was the entrenched idea pre-internet and has not been disproven, then where is the positioning of targeting? My hypothesis would be that targeting is likely best utilised when people stray from self selecting brands, e.g. NYTimes needs no targeting, but a smaller publication likely does, as people aren't there for the brand's known positioning, but because of their targetable interests, e.g. they ended up on a site about coding, or computer games, or knitting or whatever.

> In general, you may do better to advertise your stuff to the people who aren't already interested.

If true, that would be easy for an ad network to detect. If the click through rate for ads on a topic not matching the users interests (as determined be browsing history) are higher, then your theory is correct.

It's basic stuff, and I'm sure they already do that.

The ads in NYT can still be targeted to the article content. Surely it may be less efficient then personalized ads but then the newspaper does not pay ad broker.

Hmm, I guess you can A/B test it and see if there is any statistically significant lift with the targeted strategy.

I believe your counter examples show the need to pick good targets instead of the invalidity of targeting.

this is fantastic analogy. but, Steve just got a promotion and likes the BMW 3 series. But, advertising a Tesla 3 can be another worthy option you are trying to advertise. I think behavioral and randomness should go hand in hand.

It does make sense the way you describe it.

But the ads I see are way off target.

Hell Google's news ... AI or whatever they're doing has been absolutely convinced I'm a Nebraska Cornhuskers fan, probabbly because my team plays them and I google something about them once in a while.... but in the past even when I told Google directly "I don't want to see stories about Nebraska Cornhuskers" ... after a while it again becomes convinced I'm a huge Cornhuskers fan.

I'm just not sure these systems are ... that effective.

Not to say it is easy, the systems have to be really complex / interesting.

> But the ads I see are way off target.

The two sides see "off target" quite differently. If the baseline is 0.1% accuracy, and an ad platform offers 1% accuracy, then you see relevant ads showing up at 1 in a 100 rates rather than 1 in a 1,000 rates. You still feel like you're being spammed by useless nonsense. As an advertiser, my cost to reach my audience just dropped by 90%.

Google News for the longest time kept showing me updates for the Fremantle Dockers, an Australian football team (I'm Swiss, and never set foot on Australian soil). I wonder why...

Do you use / search for information about Docker (the container platform)?

Yeah I've gotten some ultra disconnected bits too.

I'm really curious how those happen.

well in this case I assume that my frequent searches for Docker (the technology) led Google News astray :)

Wow, that's super random. As an Australian, hope you enjoy Aussie Rules!

Maemo user? :D

I can’t help but think the only thing most people buying ads go for is demographics (age gender race income) rather than more personal interest targeting, since that’s been the primary driver of TV ads for the past many decades. Interest targeting seems like it doesn’t work, or if it does the interest based ad buys are few and far between.

That's an interesting point. I wonder if ultra granularity ... isn't really the current state...

At one point it was sure that I was female (am not, have never been) too.... to their credit they figured it out after a year or so, but man the ads. But that was at least a limited run of confusion.

Or "let’s sprinkle some noise in here so people think we’re less efficient than we truly are, so that we appear a bit dumber and less creepy."

But probably the truth is that these systems aren’t that good :)

Oh they definitely do this. At least the traditional loyalty-card/coupon-enabled physical targeted ads do add enough random stuff that you find it non-creepy, after they had a few high-profile cases back in the day where they started targeted advertising of baby products to teenage girls, and parents got mad. The kicker was they turned out to frequently be real cases of teen pregnancies.

I remember that, it was Target wasn't it and they inferred her pregnancy status because she started buying unscented deodorants and stopped buying sanitary products.

Funnily enough when I'm discussing the dangerous nature of large datasets in the hands of companies/bad actors (sometimes the one and the same...) and the "creepy" effects this is one of the examples I use, I've found that if you know a little about the person and couch the argument in a way they can personally respond to, they understand even if they don't know the first thing about technology generally.

EDIT: https://www.forbes.com/sites/kashmirhill/2012/02/16/how-targ... for anyone who hasn't seen it.

The have to put in some out of context ads sometimes, just so they can train their system.

For example they may show you an ads about hairdryers after you searched for computer keyboards. There is absolutely no relationship between the two, but Google doesn't know it yet, so they try. After all, there may be a correlation. In the likely event where there isn't, Google will simply stop showing hairdryer ads and show you computer mice instead (a known correlation) or running shoes (another experiment).

You'd think they'd sprinkle in something less ... annoyingly skewed that obscure what I actually want to see.

My experience of targeted advertising is Amazon trying to sell me things I’ve already bought ...

This is a point I bring up when casually talking with non-tech minded friends when they start talking about machine learning taking over everything...

I bought a Dyson last summer. Amazon is still suggesting that I buy more vacuum cleaners.

"We showed him an ad. He bought the vacuum cleaner."

'In that order?'

"Shhhh, just give us money."

Half-joking there -- in a previous discussion, the claim was that the ad-buyers don't care enough to check for whether these product-buyers were ones that had already seen an ad, or it was after the fact.

This makes sense. While the "you might need a second vacuum cleaner" argument maaybe checks out, the "you might need a second sous vide machine/espresso bean grinder/welding machine" definitely doesn't make sense.

There are some people who end up buying all the best coffee bean grinders in search of perfection; there are people who can't resist a classic MIG welder. But those people were going to buy them anyway, and represent too few sales to justify any expenditure.

What you need are complementary goods. These people bought coffee bean grinders? Show them ads for travel mugs, for high-end coffee subscriptions, and fancy grinder-cleaning brushes.

Agreed; just to be clear, when I've had this happen to me it's showing an ad e.g. for the exact same model coffee grinder from the exact same store that I already bought one from. Clearly no intelligence involved there.

Dyson users probably disproportionally have second homes, holiday homes, etc.. You're more likely to recommend to a friend of you see a deal "hey, I bought this awesome vacuum, you should get one, I saw a deal" ...

It's not as silly as it seems.

I buy repeat items as gifts (I like this, X will like this), or for my parents (home stuff), or when I bought something for one child I know and it turned out to be well made and now another child needs/wants one; or tech stuff that turned out well when someone else needs the same repair (eg new HDD).

It's not as silly as it seems.

I'm not sure I agree. I mean, I see the theory you're getting at, but the Dyson example is built on a few assumptions we'd need more data on -- starting with the notion that Dyson owners are disproportionately more likely than owners of other vacuum cleaners to have second homes. And even if they do, that's, well, one more sale at most, and one that's more likely to be influenced by their experience with the first Dyson rather than ongoing advertising. When you say that you buy repeat items as gifts because they turned out well, you're tacitly confirming that further advertisements for that product aren't necessary to reach you. At best, the ads can sway you if they happen to be running an unusually good deal on the product you've already decided you want to buy.

The big problem targeted ads have now is, as other people have pointed out, that they seem to be targeted with knowledge of what you've recently been looking to buy, but not knowledge of what you've recently bought. If I search for polo shirts, I'm in the market for polo shirts, but once I buy polo shirts, I'm probably not going to be in the market for them for a few months. Once I buy a car, or a television, I'm probably not going to be in the market for another one for years.

>further advertisements for that product aren't necessary to reach you //

I'm one if those "advertising doesn't affect me" people. Except, then I realised that advertising acts subconsciously and uses human psychology against us, and I'm not as immune as I thought.

I buy a Dyson, every ad I see for vacuums is a Dyson, it confirms I made the right choice, everyone is buying them, they're everywhere, etc.. Everytime I see a shiny new vacuum it's a chance for my brain to compare it with the old rubbish one; why are all those people "enjoying" vacuuming when I have to suffer it.

With polo shirts it's like "these new ones look smart/fashionable/etc." vs the old one.

Yes there is likely a lag in "this person was looking for ..." signals; but I still don't buy (heh!) that as _entirely_ silly.

>Once I buy a car, or a television, I'm probably not going to be in the market for another one for years. //

Car, probably, depends on the person's wealth; TV, I've v heard people say "we liked it so much we got another one for our bedroom" or whatever.

FWIW I'd never buy a Dyson ...

No one has the data for both sides of the equation. Google dominates search advertising, Amazon dominates online buying. If those companies merged, one company would have all the data needed to use deep learning to take all our money. They'd be able to distinguish products that, once purchased, means the buyer won't buy again for years, and items that are regularly bought, and they'd have the timing between seeing ads and buying. It wouldn't be perfect, because some people would buy elsewhere, but it might be frighteningly effective.

If Amazon of all companies, who regularly ship things to my home, can't figure out that I only have one home then maybe all this targeting stuff is hopeless after all.

And selling Hawaii vacation packages to people after they have been there. And I live on the east coast.

This also puzzles me. If I buy a household appliance that's not a consumable product (most shouldn't be), why does it recommend more of the exact same? Why doesn't it instead recommend other types of appliances?

The key take-away should be that you're comfortable with buying household appliances online, usually not that you need another vacuum cleaner within the next year.

It's data, and a lack thereof. It's easy to know you're browsing a site. It's almost impossible to know you bought something.

Banks and merchants and publishers and advertisers and all the layers in between create dozens of data silos and nobody is interested in sharing. Add to that the weird expectation that people don't want to be tracked but yet want adtech to know when you've bought something.

Well surely Amazon knows what I bought on Amazon...

Tbf Amazon has the least experience with ML of the FAANGs.

It seems that it would pay out a lot for them to do a few simple heuristics by hand. For example: if just bought something do not show a thing from the same category. At most show a thing from accessories category, which I think they do.

Up until recently i used to work at a major Indian eCommerce and can confirm that most companies use a mix of the simple heuristics you suggest together with some data mining.

So, they would take your recent purchases, clicks, searches and wish listed products and recommended products from some accessory category or related products. ie. Traditional mining combined with heuristics.

It also helps to have separate widgets for similar products and for cross-sell products.

I always wondered why that happens at all but then I thought maybe it is because of their historic origins starting out as book seller - where showing another book after you bought a book makes absolute sense. While for refridgerators it makes no sense, they should put stuff in categories to distinguish that but then they are a super big company and I am single selftaught developer trying to make my living out of it.

Maybe that just goes to show just how much money they're making that they don't really need to feel bothered!

Nah. That's not AI enough!

It depends on how you measure that. They certainly were the first to make serious money by using simple regression to suggest 'people also bought'.

In general I think that all of the personal surveillance / categorization will end up being more truly profitable (as opposed to 'twinkie calorie' profits of ad networks) as they are used to suggest price points for products and services rather than to direct attention to products.

And Booking pushing advertisements for the exact same hotel I've just checked on their service. I can't understand how that would seem like a good idea to anyone.

And Booking pushing advertisements for the exact same hotel I've just checked on their service. I can't understand how that would seem like a good idea to anyone.

This happened to me just last week. Unfortunately, it's the least of Booking's user-hostile traits.

Like sending me e-mails pestering me to review a hotel I checked into less than an hour before.

Or flashing "Hurry! Only 1 room left!" when I know for a fact that the hotel I'm inquiring about is 90% empty.

Or flashing "23 people booked this hotel in the last 24 hours!" when I know the inn only has 4 rooms and is in probably the lest popular destination in North America.

Unfortunately, there's a particular place I have to travel to which only has one motel within 70 miles, and the only way to book a room in advance is through Booking.

A hotel can be 90% empty and Booking could have only 1 left in their inventory to sell.

A hotel can be 90% empty and Booking could have only 1 left in their inventory to sell.

There's a difference between "Only one room left!" and "We only have one room left!"

It's called lying.

The opposite works as well.

On one occasion, I've tried booking straight from the hotel, it told me no rooms available for the selected dates. I've tried Booking, and sure enough, that same room I had my eyes on was available.

Larger hotels have blocks of rooms reserved for the different booking services.

The opposite can also happen, Booking shows no rooms, and calling the hotel results in a room. It just means Booking's block of rooms is all rented out, not that the hotel itself is all rented out.

These are so obvious that I find them mostly hilarious, even if they are clearly dark patterns. I can see how less experienced hotel bookers can fall to this trap though. I do end up booking with them, as for some reason they usually are cheaper and have more stock listed in Europe than the main competitor Hotels.

All those little tricks you described are part of a strategy usually called Conversion Rate Optimization. They do this because it works. What you described was likely A/B tested. Some visitors to the site saw the nagging message, while others didn't.

Turns out you can grease your numbers by manipulating user sentiment.

Who cares if you need to lie a little bit, right?

Actually... That one works, if certain conditions are met.

They also don't update that fast.

You looked at a hotel in London and a certain hotel is the best revenue generator in the area - for a little bit you're going to see that ad... or maybe you're going to see that ad for a long time, if there are no better options to advertise.

I've thought the same (and heard the same booking "blunder"), until I visited a city I had visited before, and went for the exact same hotel...

But did you need an ad to get you to do that? Seems like that's the one ad you don't need to see.

That depends on where you see that ad. If you search for that particular hotel on Google, it's going to have an ad from N online travel agencies(OTA).

If you do a broad search on OTA website, then you're likely to book the same hotel that you stayed last time... Thus pushing it up is a great way of getting you to book... But that depends on how much the hotel is contributing in marketing

Surely that would make sense the next time you visit the city, not for the same visit.

There’s some marketing theory about brand reinforcement near a purchase creating a more likely customer later. I’m not a marketer though.

The easiest way to get "brand reinforcement" is to deliver a good product. I will remember them. Right now it seems they want to get "brand reinforcement" through constant nagging which already causes me to dislike the brand.

This seems to be ingrained deeper than targeted ads. Orders give me the option to "buy it again" for things that I only ever need one of.

Or, sometimes in my experience, something I was selling.

I browsed retailers to copy the specs into my ad.

In fairness, this is an understandable mistake for any automated system to make, even more so when compared to Amazon thinking everything is a consumable I repurchase every month.

Hint to Amazon: not everything is a consumable ;)

Me too - but I do wonder sometimes if there's a long game involved, where it's more likely to make you recommend it, or give it as a gift. I can't quite see how that would work in practice, but it just seems so utterly stupid otherwise.

> Me too - but I do wonder sometimes if there's a long game involved, where it's more likely to make you recommend it, or give it as a gift. I can't quite see how that would work in practice, but it just seems so utterly stupid otherwise.

I'm going to go with Occam's razor here: the advertisers are not scarily competent but instead typically bumbling and stupid.

Or the slightly more prosaic British version: cock-up, not conspiracy. Me too, on balance. But it seems like it should be easy to mark entire categories - particularly large purchases like vacuum cleaners - as "recently bought one of those, try something else."

I've heard this is done sometimes to make you feel good about your purchase and the decision you made by showing you inferior options.

can you elaborate more on this or perhaps drop a link?

You be surprised how much less sophisticated Googles and Facebooks ad targeting is!

We are far away from a machine classifying you as being single and going to the gym every day or liking BMW and having gotten a promotion.

Not that I don’t think we couldn’t get there...but today these system are far away from it!

Just go to your Facebook ad settings to understand how primitive the information is they have about you.

But I agree that targeted ads (as primitive as they are) are more efficient then I untargeted ads

> You be surprised how much less sophisticated Googles and Facebooks ad targeting is!

If Google's ability to discern my interests for ads is as bad as is their ability to discern my interests for news, then I would not be surprised.

I have never been aware of listening to any music by the band "Foo Fighters". I've almost certainly heard their songs in passing, but never in a context that gave me the name of the band.

A friend on Facebook posted a link to a YouTube video called "Dave Grohl brings kid on stage in Kansas City to rock out", in which Grohl, who is apparently the lead of the Foo Fighters, invited a kid from the audience at a Foo Fighters concert up to play a song, and I watched this video. (The kid asked to play a Metallica song, which Grohl and the Foo Fighters knew and played, so I still have not knowingly listened to any Foo Fighters music!)

Google has latched onto this and decided that I am interested in Dave Grohl news. Every time I go to Google News, for the last several weeks, the top of the "recommended by your interests" section is a story about Grohl or the Foo Fighters. Even though I hit the "show fewer stories like this" link for most of them, still they come.

Furthermore, they are almost all negative stories. The one thing they have of me showing any interest in Grohl and/or Foo Fighters, that video of the kid on stage, was a positive thing. But what they are giving me is a parade of stories about Grohl screwing over current or former bandmates, Grohl misbehaving while drunk, etc., from what appear to be trashy gossip publications. Yet a bit of research shows that Grohl is apparently actually a nice guy, well regarded. If I click the link to tell Google to not show me anything else from one of these trash publications...it obeys and just turns to other one to find Grohl gossip for me.

What the heck, Gooogle!?

I find that after a few weeks of training, Apple News does a much better job of picking stories I'm interested in than Google News. With two exceptions:

There doesn't appear to be a way to filter out sports I'm not interested in, or even all sports.

Apple News has a hard time telling the difference between Paris, France and Paris Hilton.

To be fair... Paris Hilton can be a Hilton hotel in Paris. So... Even an proficient English speaker can be mislead

Even an proficient English speaker can be mislead

I don't think so.

A human would understand context. AI isn't supposed to be simple string matching.

Really? Take anyone that has no clue who Paris Hilton is and without implying what you mean by "Paris Hilton" ask them something about Paris Hilton.

There's a difference between "Paris Hilton" and "The Paris Hilton." "Paris" isn't a completely unheardof first name. Again, a human capable of working with context will understand that, as should a proper AI. Simple string matching won't.

Also, there is no "Paris Hilton" hotel. There is Hilton Paris Opera, Hilton Paris La Defense, and Hilton Paris De Gaulle.

I hadn't looked at google news for a while, so went to see what headlines they'd offer compared to the Apple news app. I wasn't looking for it, or expecting it, but, second headline under "For you":

"Letter from Foo Fighters frontman to Cornwall Council goes viral again"

Oddly enough, I recently have been seeing an onslaught of Foo Fighters stories on Google news. I am in fact a music fan and musician who knows their songs but I have no special interest in them. I wonder if something about them triggers some runway like function in Google’s algorithm.

Obviously Google treats "foo" as a placeholder and returns these stories whenever it thinks the user might be interested in a story about a conflict.

More effective: unsubscribe from any sources that report on the crap. Don't just say you're not interested in the story, say you're not interested in the source at all. After a few weeks of doing this daily, you should have mostly purged it.

Perhaps there's some kind of Baader-Meinhof phenomena going on here.

When a targeted ad is creepily accurate and on point, people flip and think that the machines have figured us out. But the ads that are irrelevant just fly by us without a hit. The one-out-of-a-hundred ads that get it right, likely due to an ad targeter's lucky strike, are the ones that get under our skin, and the only ones we really notice.

I read somewhere (can’t remember the book name unfortunately) that Tesco in the UK faced this problem when they started sending out personalised coupon booklets based on loyalty card purchases. People were freaking out. Then they started adding coupons to the booklet designed to be irrelevant - say if they figured out that someone was newly pregnant, they’d send a few pages of baby stuff, but also inject lawn mower coupons and stuff only men would care about. That achieved the whole targetted effect and avoided looking creepy for just a few pennies more.

> I read somewhere (can’t remember the book name unfortunately) that Tesco in the UK faced this problem when they started sending out personalised coupon booklets based on loyalty card purchases. People were freaking out. Then they started adding coupons to the booklet designed to be irrelevant - say if they figured out that someone was newly pregnant, they’d send a few pages of baby stuff, but also inject lawn mower coupons and stuff only men would care about. That achieved the whole targetted effect and avoided looking creepy for just a few pennies more.

It was Target in the US:


> As Pole’s computers crawled through the data, he was able to identify about 25 products that, when analyzed together, allowed him to assign each shopper a “pregnancy prediction” score. More important, he could also estimate her due date to within a small window, so Target could send coupons timed to very specific stages of her pregnancy.

> ...

> “With the pregnancy products, though, we learned that some women react badly,” the executive said. “Then we started mixing in all these ads for things we knew pregnant women would never buy, so the baby ads looked random. We’d put an ad for a lawn mower next to diapers. We’d put a coupon for wineglasses next to infant clothes. That way, it looked like all the products were chosen by chance.

I remember during college I was working at Tesco, and a customer way buying a lot of baby food, and she had received some Clubcard coupons for them. I asked her if she had grandchildren, she said no, but her husband really likes baby food.

A niche yet to be exploited?

You still get that if you shop online, everything you ever bought suggested as an add on to your basket.

I can see how this could cause embarrassment and, maybe, awkward questions if one's partner also uses the same account.

But do people generally think that targeted ads are especially accurate? If anything, I’d think people are bothered by all that privacy invasion and still getting completely irrelevant ads!

God forbid you accidentally click on the wrong product in a re-targetting carousel. Now "the advertising complex" thinks I'm super into motorbikes. I don't even have a license.

IMO many people managing ads do not seem to use it properly. I worked for a marketing agency once and we got pretty good results, but we tried to understand our target, use negative keywords, use different ads for different keywords and so on. It was actually a lot of work. I guess some people wants to cut corners and just gets the client money and puts it down into adwords and fb ads, write some reports and pass it to client. In my experience clients most of the time get confused with so much jargon, and they are unable to link sales with marketing expending correctly, the only certain thing they know is that (for most of them) if they don't do marketing sales will go down over time.

I'm always surprised when I see a company wasting money on adwords when they are also ranking many #1 keywords organically, specially on brand keywords. I may be outdated but I doesn't make any sense to me.

My Facebook/Instagram are stunningly on point. I've even bought a few things from them.

I want to echo this. I need a new bag and gym clothes. Instagram is hitting me hard with this. I haven't consciously shopped for them either, but I assume I've looked at enough content to get me in those buckets. I am considering getting one of the bags they've shown me.

Facebook is full of scam ads and spam. They are literally the worst ad marketplace...e

One thing I have wondered for a little while is how all these learning algorithms will handle change over time. A lot of my advertising is for things I may have liked perhaps 5 years ago, but no longer have anything to do with. It's also a bunch of super wrong things, like parenting, even though it also thinks I'm a single male who likes holidays, and skiing, even though it doesn't snow where I live and I've never even entertained the idea.

It's clearly confused, and so my wonder is how do they correct that over time? How do you un-scramble a scrambled machine learned profile? Can you even detect that it's gone awry?

Yes! Both Facebook and Google have ad preferences pages you can go to to see what they have inferred about you and delete any wrong things

The strongest signal to fix your targeting is to buy something you actually like from a ad!

The models react strongly to that...i went from years of poor ads to fantastic ads within 2-3 purchases

Worked there... Done that...

They are very efficient in telling people that are intending to buy your product that they are going to buy your product. Or even buy your advertised product on the website that you're advertising.

They are not efficient in acquiring new customers, as most(90%+++) of the ad impressions go to the people that are already aware about your product.

The only truly efficient ad targeting is restricted by geography, culture, age and timezone.

I think this is just it though. Everyone assumes targeted advertising works, because it seems self-evident that it should. And the only people likely to have significant data on whether it works have a vested interest in convincing everyone that it does work (regardless of whether that's true or not).

It may be that targeted ads are more effective, but some counters to your points:

> John has a 10th wedding anniversary coming up.

A wedding anniversary is a significant thing–some John's may be swayed by an advertisement, but many would consider it a bit of a cheat not to put more "novel" thought into it.

> Mary is single and goes to the gym every day.

Frankly, the former of those two data-points could mean anything. Much of what advertisers believe about consumer habits of single people could easily be correlation.

As for the latter, anyone with a daily gym routine is less likely to want to change it. Selling gym membership to someone speculatively hoping to start going to the gym is a better bet, but much harder to track.

> Steve just got a promotion and likes the BMW 3 series.

Going back to the anniversary gift—I seriously doubt many people are much swayed by one-off ads in buying something as significant as a car (unless Steve has 10+ cars, in which case the promotion is less relevant). This is case where untargeted ads definitely have a much bigger role (Steve sees many BMW ads passively over a number of years).

It may be there's some cases were targeted ads are extremely effective, and some where they're not. And—as you've pointed out—it may be that Google et al. aren't really incentivised to actually be good at targeting as long as their advertisers believe they're good at targeting (which goes hand in hand with market dominance).

In practice, I get the impression that the emperor isn't wearing a whole lot: It's really difficult to demonstrate that any particular ad campaign generates a positive return at all, let alone that any particular technique justifies its cost, or that something that worked well the first time someone tried it will continue working well once everyone is doing it. The whole space looks to me like a morass of wicked high variance, confounding variables for days, spillover effects, blah blah blah.

Combine this with marketers wanting to be more data-driven - and needing to report numbers to their own bosses - and you've got a situation ripe for exploitation. I expect the adtech companies could sell pretty much anything, as long as it smells like numbers.

You grossly misunderstand digital advertising or your bias has completed clouded your logic.

> A wedding anniversary is a significant thing–some John's may be swayed by an advertisement, but many would consider it a bit of a cheat not to put more "novel" thought into it.

No idea what you mean here by cheating or novel thought. We're talking about whether knowing that info allows you to have more successful advertising compared to random chance. Serving John ads for roses, chocolates, and vacations will be more effective than serving those ads to a random person.

>Frankly, the former of those two data-points could mean anything. Much of what advertisers believe about consumer habits of single people could easily be correlation.

Correlation is the whole point. If someone who is single and goes to the gym is correlated to certain purchases or behavior, you can advertise those. And being single, a woman, and a gym-goer isn't super valuable by itself but combined together you can advertiser women's athletic clothing that is functional but also attractive.

> As for the latter, anyone with a daily gym routine is less likely to want to change it. Selling gym membership to someone speculatively hoping to start going to the gym is a better bet, but much harder to track.

Again, you don't understand advertising. If I know someone goes to the gym every day I'm not going to advertise gyms to them. I'm going to advertise water bottles, protein, healthy meal kits, athletic clothing, etc.

> Going back to the anniversary gift—I seriously doubt many people are much swayed by one-off ads in buying something as significant as a car

We're not selling Steve a car with targeted ads. We're taking his promotion and propensity towards bmw to assume hes wealthy and likes luxury goods. You advertise more expensive goods to him, rather than cheap ones.

> Serving John ads for roses, chocolates, and vacations will be more effective than serving those ads to a random person.

That statement makes sense intuitively, but what we're discussing here is whether our intuition reflects reality. It may seem to self-evident to you, but do you have data?

All of your examples "make sense" from the same intuitive perspective, but all assume the subjects are positively influenced by the advertisements more than a randomly selected subject would be. That's an assumption.

Please don't condescendingly remark that I "don't understand advertising" when you've missed the point entirely. This isn't about how the ad industry works—ad companies are clearly economically successful—it's about whether (as another commenter eloquently put it) that emperor is wearing clothes.

> Correlation is the whole point. If someone who is single and goes to the gym is correlated to certain purchases or behavior, you can advertise those.

The point is that the correlation we're talking about here is a proxy. If a lot of people buy a product and are single, but most aren't buying that product because they're single, it means they're buying it for another reason, which may or may not change relative to their relationship status.

Marketing intuition is literally what you’re talking about. They teach it in schools. It’s existed for hundreds of years.

Your deep bias is breaking your logic. These are assumptions that marketers test. If they work, you keep doing it. If it doesn’t, you try something else. You clearly have no idea how the industry works, otherwise you’d understand that you’re arguments have been addressed.

Correlation is the entire point of that advertising. No marketer cares whether being single causes a purchase, simply that being single is correlated with a purchase more than a random amount.

You're probably right but ...

I don't have any data on my own actually behavior, only my feelings. If I search for "React" on Google and ads for react based services or other software services appear it doesn't bother me. If I'm on stackoverflow or jsfiddle and there are ads for software dev related products and services great! If I'm on Polygon and see ads for games, perfect!

It's when I see targeted ads unrelated to the activity that I feel angry, annoyed, upset. For example I viewed some apartments on an apartment site. Then I went and checked movie reviews on yahoo (japan) and every ad was for that exact apartment I looked at. It's not just creepy it's anger inducing. I'm not interesting in looking at apartments. I'm looking at movies. It's like the sales person from the last store I visited followed me into the movie theater and is badgering me to buy the clothing I looked at

So, I'd basically like to believe there's 3 tiers

1. 100% un-targeted, random ads

2. Content targeted. Video game ads on a video game site. Or BMW ads on an article about BMWs

3. User targeted. Being tracked all over the net to try to divine what I want and showing me ads for that.

It's easy to believe 2 and 3 are better than 1. I want to believe 2 is as good as 3, maybe better since it's not creepy.

> Having said that, I'm constantly surprised by how bad Google is at targeted advertising. For example, today when I visit nytimes.com I see an ad from Google with the ad text in French. Hey Google, despite my recent visit to Paris I don't speak French - maybe your AI experts could analyse my 13 years of Gmail and search history to figure that out!

No, they literally could not, because the data flow between Gmail and Ads has ben severed a while ago:


These kind of things are treated pretty seriously. There are teams whose job is literally making sure that your data does not leak from one Google product to another in a way that would violate your privacy. Source: I'm an engineer in Google, but in none of the mentioned teams.

Gmail is the exception, as google can harvest all the data from other Google services.

Virtually all google products are governed by a shared google privacy policy, which allows Google to combine data across products and services. This change was made years ago, despite some end user protest.

Most apps and products are also optimized to maximize data harvesting - ex. Android OS and Maps will nag if users don't give up data. So I don't really believe what you're saying.

I can't even tell what you mean by "data harvesting" in general. But the fundamental error you're making here is considering "Google" as a single entity. Frankly, with that kind of thinking you can't win, because either Google, Apple or Microsoft has all your data. No need of harvesting, you have literally gave it all to them and asked to keep it for you.

However, once you allow the idea that there are powerful people inside these companies, that genuinely don't want the company to create a panopticon, things can get interesting. I can't exactly prove that to you, but you've seen hints of that in the news over 2018. Or, you might have noticed e.g. that safe browsing only sends hashes of the URLs you type. Or, now this Gmail not telling Ads which languages you speak thing.

Now, again I can't offer more than my word for it, but the Gmail-Ads situation is the rule, not the exception. The only reason this made it to the news, is that the exception in place before has been lifted. But in Google, by default, if a product wants data you shared with another product, it has to make a very strong case that that's covered by intent of a permission you already gave. That, or ask you for permission, or ask you for the data again. If you ever wondered why Google keeps asking for the same basic immutable data, that might be the reason.

Funnily enough, this whole privacy thing is hard to tell from the outside. As you said, there is a single privacy policy, that governs the relationship between Google and yourself. All the internal data siloing is, well, internal.

Edit: I won't be continuing this thread. The cognitive load to not leak anything non-public is a bit unpleasant.

> How can it be that having intimate knowledge of someone would not allow you to sell them more stuff?

It allows you to sell more ads to the marketing people since it sounds very compelling. So there is your model.

The fact that I am still shown ads for items of the complete opposite (football) team that I am clearly (online visible, on FB, in my gmails, from my google search history, from my chrome browsing habits) fan of; show the targeting is still... moderate to say the least.

Google has probably put in you in the category interested in "sports", without giving it a second thought.

Does Steve need an ad to remind him that he likes the BMW 3 series? No doubt one of the first steps he took after his promotion was look up the prices, or finance offers on BMW 3 series, perhaps create a spec and then talk to his partner about it if he has o e.

The thing is if there is something I want to buy, or own at some point in my life, I don't tend to forget about it or need some reminder from a targeted ad. If it's something I really want I'll usually think about it daily.

On the other hand, there could be something that I need, and have never searched for, that I have forgotten about such as going out and buying more toilet roll.

I don't think the purpose is to remind him in the BMW case but to get him to purchase it at dealer X instead of dealer Y.

They would be the ones making the ad, targeting someone X miles from them, who is interested in BMW's.

Perhaps he could be reminded.

Car just won an award, which would direct him to look up more info about the car and why it's considered good.

Maybe they just made an update to the car model he likes, but is not aware of this change.

Maybe that'll give him a slight nudge toward buying the thing he was already going to buy, but how valuable is that nudge? BMW has a lot more to gain by providing that info to someone who isn't already into their cars.

You don't make these decisions in one go. One ad doesn't make someone who isn't into cars go out and buy a BMW. The money is in tipping someone who's been thinking about getting a BMW for a while over the edge.

I wasn't talking about people who aren't into cars at all. I was talking about people who aren't into BMWs in particular. If I'm about to spend a lot of money on a car, I'd think that BMW would want me to know about theirs.

I think it also depends on the quality of the targeting.

We tend to think that these companies know us very well, but (anecdotally) that's not my experience on Facebook.

There is a page where you can see a list of things Facebook thinks you like. In my case, it's full of things I don't care about and I would never buy. They are there because, according to that page, I clicked on some ad at some point.

Which I don't remember, so it might well be five years ago. So I get targeted by ads I am not interested in at all. A random ad has more chances to capture my attention than something targeted on a non-interest.

Maybe the cost for effective targeted advertising is too high. You have to collect a ton of data and analyze it which costs a lot of money. Maybe untargeted are more cost effective.

> How can it be that having intimate knowledge of someone would not allow you to sell them more stuff? John has a 10th wedding anniversary coming up. Mary is single and goes to the gym every day. Steve just got a promotion and likes the BMW 3 series. Michael is overweight but has just gone on a diet. Are such details of no significance determining what ads to show these people?

I think there are a few factors:

1. Targeted ads are more likely to show people things they're already familiar with. The ads are more "expected" so there's less novelty. Untargeted ads can inadvertently hit wants and needs that aren't part of the user's tracking profile.

Closely related to that: targeted ads can induce their own special kind of fatigue. If all you see are diet ads because your profile says you want to lose weight, you're going to get really sick of seeing diet ads. The targeting actually works against effectiveness.

2. Targeted ads are much more exploitable by scammers. See: https://www.theatlantic.com/technology/archive/2018/01/the-s.... This leads to a greater erosion of trust. The expense of traditional and untargeted advertising weeds out such scams.

3. Building a good user profile is hard. If you get it right, you have the problems I described in #1. If you get it wrong, you're blasting someone with ads that are completely off the mark (like weight loss ads for a very healthy and fit person).

4. If a user sees one targeted ad a week, that ad is likely to punch above its weight. If they see 50 targeted ads every day, the effect is much different. They're more likely to develop negative emotional reactions to the ads, as they gain awareness and get creeped out by the tracking and attempted manipulation.

Perhaps that's what google want you to believe, much like how Target scared people off with their targeted (heh) marketing that outed, among things, that a teenager was pregnant due to the things she started buying. After those events, they started inserting random ads to give a plausible chance that the very-targeted ads were random.


Targeted ads are almost universally pretty terrible. My reaction to seeing a targeted ad usually falls into one of these categories:

- I already bought that recently, why would I need another one?

- I already decided which variety of thing you're showing me I will buy, this ad isn't helpful to me.

- Why are you spying on me? That's creepy, I'm not going to click that ad even if it was exactly the thing I wanted to buy right now.

I expect that this experience isn't extraordinarily unique. Having less targeted ads is actually somewhat helpful because it maintains the pretense that ad companies aren't spying on you constantly.

Why does Steve like the BMW 3 series?

I contend it is due to untargeted ads.

Or because Steve regularly goes on business trips and has rented the 3 series in the past. That's basically why I want a Mazda 3, because I drove it for 2 weeks and it was fun.

You and I might, a prior, believe that they would be much more valuable, but after 10 years and with some of the sharpest minds in the world working on the problem, facebook has never once shown me an ad for something I wanted. I have as far as a recall only twice clicked on ad on google with intent to acquire the item (as opposed to just clicking on the ad rather than go to the homepage of a company).

If you could actually create ads for things that I would want, personalized ads are probably worth more. I just don't think it is possible with current technology.

Sometimes it would pick up on keywords in my messages, but I rarely talk about things I wanted to buy. I'm a very satirical individual so if I mention a product it's most likely because I'm making fun of it. Making the advertisements extra ineffective.

I am in my mid-30s and I have never ever bought anything that was advertised to me. Ever. I couldn't even say why. Sure, I click on the first links in the google SERP, but just because I am lazy ass.

>I would be absolutely astounded if targeted ads did not provide significant long-term advantage to a big player like Google. How can it be that having intimate knowledge of someone would not allow you to sell them more stuff?


People know when they're being manipulated and we usually don't like it. All the data in the world won't help you if using it creates antipathy.

You're listing interests... Using that just reminds people of their desires.

John has a 10y anniversary - will make John focused in buying stuff, he'll need exposure to new products... Not ads for stuff that he already looked at. John is probably not an amnesiac.

Mary is single and likes gym - doesn't make her more interested in gym clothes, she's already well aware of the gym clothes...

Steve getting a promo bundled with him liking BMW will be a total waste of a ad target. He already knows the cars and will buy a BMW with or without an ad from any car manufacturer.

Going on a diet implies that you know what diet you're going on.

In short retargeting is great at reminding people of what they like. It's really not great at acquiring new customers.

I worked on this stuff and it's flashy, but ultimately just a great way to siphon off marketing budgets out of corporations.

Don't sell your stock just yet. The product is useless, but it's a scam that is as prevalent as organized religion

It's not necessarily Google but the advertisers using the platform and how the ad is configured. Google itself addresses these kinds of ambiguities all the time but if the user puts coarse configuration on it, then it will do what the user told it to do like show French ads to non-French speakers who just visited the country.

All your examples presume that the people targeted lack the initiative to seek what they want, no? One could argue that getting all the junkmail targeted ads out of one's face _delays_ actual transactions with retailers that would-be customers can adequately find themselves, with even greater information.

My biggest issue is, i can't remember when i clicked on a ad and was not closing that tab asap.

It would be really nice to see those people buying to understand it more.

Amazon for example still shows me washing machines after i purchased one through amazon. But that feature is still here on a high traffic user landing page!

Perhaps it matters for products which are clearly build for a very specific target (like a startup selling a shopping service to 30 year olds which work all the time) but there might also be a lot of standard products which everyone would buy?

Dog food, dog toys etc.

Also i often enough see companies paying (at least those are ads) for there own name. When you enter 'Miele' (well known german company for washing machines and other stuff). You get an ad for miele.de but miele.de is already the first hit. The same when googling 'Samsung S9'.

Google is in general not selling stuff to people. It is selling ad space to companies. Targeted ads only provide advantage if these companies will pay more, or if companies will spend more of their advertising budget for targeted ads.

I would be astounded if revenue dropped at all if Google stopped providing targeted ads, at least in the short term, because there is no serious competition. Longer term it may well be an advantage as you think, but I don't think it is a sure thing. Unless targeting gets an awful lot smarter, I think companies may choose to spend their dollars on cheaper, dumber services that provide the same conversion rates. Especially if privacy concerns and law changes enable a lot more people to drop out of the targeting.

Google doesn't target ads themselves, they sell advertising targeting to the highest bidder. If someone's bidding high numbers for a French-language ad and not bothering to exclude non-French-speakers, Google is more than happy to take their money.

> How can it be that having intimate knowledge of someone would not allow you to sell them more stuff?

If the unknowns have a greater influence. So, a little knowledge makes google overconfident.

Half the money I spend on advertising is wasted; the trouble is I don't know which half.

Now removing the names:

>>> User1 has a 10th wedding anniversary coming up. User2 is single and goes to the gym every day. User3 just got a promotion and likes the BMW 3 series. User4 is overweight but has just gone on a diet.

There, no identities and you can still advertise. Is it hard?

>Are such details of no significance determining what ads to show these people?

Perhaps the question is whether that turns to sales and engagement, or if people who don't click ads don't click ads regardless of content, and those that do, do.

Doesn't that assume that the targeting information gets used appropriately? My understanding is that advertisers bid on placement and the largest big gets the prize, regardless of whether or not they are using it "correctly". So it wouldn't matter if BMW wants to target people with new job promotions if, e.g., Rolex is outbidding them for "luxury".

The vast majority of ads I see are of no relevance to me. It makes me assume that the companies that would have relevant products just aren't spending as much as broader interest stuff, bringing us right back to square one.

The content of ads rarely matches with something I'm actually interested but uninformed in. The most wasteful ads are probably the ones about products I already bought, or the ones about products referring to a conversation partner or shared links to websites which hold interest to the sending party, not the receiving one. The ads are very targeted but not targeted at someone who would ever buy the advertised thing.

If only one could 'scan' the product they just bought to no longer see any ads about it...

A few (potential) ways targeted ads may not be advantageous:

Often I see targeted ads for products I looked up and did some research on. I either bought the product or decided not to buy the product. Unless the ads convince me to buy another one or change my mind about whether or not I need a first one, they're not going to do any good.

Targeted ads can also be creepy and offputting. It's weird to look something up and then see ads for that specific version of that specific product everywhere.

I've worked on recommenders for a big brick & mortar with significant e-commerce.

Our team's recommender algorithms were marginally effective. Despite having significant data on our customers.

The new new thing is "personalization". We'll see if that helps.

If I had to guess, broadcast and individual targeting both (can) work, but the middle ground is fool's gold.

IMHO, I'd like to see renewed emphasis on improving search, browse, foraging. I generally know what I want, but still struggle to find it.

Then you should try buying some Google Display ads sometime. See how quickly Google wastes your money, because the tools it provides are nowhere close to the targeting you may have in mind. The audiences you'll get to target even after a lot of manual fine-tuning will still feel too generic.

But they'll also still be expensive because everyone else also thinks the data is super-targeted so they throw a lot of money at it, increasing bids.

The efficiency of targeted ads depends greatly on there being a sufficient number of ads and sophisticated buyers on the platform. If you only have one unsophisticated advertiser on Adsense, it's a high likelihood that Adsense will just end up showing that ad to its entire network and the targeting advantage is completely negated. This is the problem left to be solved in advertising and it's really hard to solve

> Steve just got a promotion and likes the BMW 3 series. Michael is overweight but has just gone on a diet. Are such details of no significance determining what ads to show these people?

Really valuable - but is it best to be done when they're reading something completely irrelevant? If they're reading a piece on the latest political issue of the day, are they really in the mindset to explore that purchase?

Twitter is even worse - the adverts when travelling are almost entirely in the local language.

I have used Twitter for a decade. They know where I live, and that I’ve never once indicated that I know Spanish or Japanese or whatever, yet somewhow their targeting setup seems to be “you are in Japan therefore here are Japanese adverts” which is fucking baffling for a company with that amount of engineering resource.

A nitpick, but Google stopped scanning emails for advertising back in like 2015. I see this still thrown around way too much...

Except that more typically, if I buy a pair of shoes I am then followed around by footwear ads for the next month.

Or you talk about shoes... :)

Here’s a thought: targeted advertisements sell me crappier versions of things I’m likely to research, while non targeted ads actually aid in discoveribility of things I would not be already thinking of purchasing

Maybe it's the willingness to click a random ad from a curated set that I haven't seen before instead of clicking on things that have been forecasted as interesting to me, but I already know about them?

It's not Google that's bad at this, but the people buying the ads and selecting the targets. And if targeted ads wasn't a thing then a large portion of your ads would be in a different language.

Well, you remember that French advertisement, don't you? Just because you don't think an ad is effective doesn't mean it wasn't effective.

Over-targetting creeps people out, so I sometimes wonder if Google (and others) intentionally slip in "bad" targeting to lower peoples guard.

I get amazon ads all the time, "based on your interest in X you might like Y".

They've been wrong 100% of the time.

Maybe I'm just weird.

Advertising seems to be pretty bad at suggesting things that would be targeted at me.

I've heard directly from Google ads eng that it does not.

What if I told you the NYT has been doing "untargeted" advertising for 150 years?

Besides if you advertise in the NYT you ARE technically targeting people: namely NYT readers. Thats a demographic right there. And a lucrative one.

National newspapers have several ways to target print ads:

- Calendar: You can buy an ad for one day, a whole week, the weekend, just Sunday, etc.

- Section: You can run your ad in the Travel section instead of Sports.

- Region: For national papers, you can have your ad printed in only one region--for example buying a full-page NY Times ad that is only printed in the papers printed in DC.

- Edition: Papers use to run a morning edition and an updated afternoon and evening edition. Some papers now produce a free thin version that is handed out free to commuters.

Big papers have also long offered ad campaign management services, where you work with their staff to optimize a campaign (including creative sometimes) to target the audience you want.

> National newspapers have several ways to target print ads:

Broad content-based targeting kind of targeting we or the GDPR is talking about. The issue here are individually targeted ads that are only possible with computerized ad delivery.

The NYT is one of the premier ad properties just because of exactly that: the demographics of their readers. Breaking google ads works out great for the very tippy-top of premier properties like the nyt. It doesn't work out well for non-premier properties.

>Besides if you advertise in the NYT you ARE technically targeting people: namely NYT readers

At that point wouldn't all ads be targeted, to some extend? The only sites that would be in "trouble" is massive news aggregations sites.

Examples of effective targeting:

- Intent targeting via search remains by far the most effective form of targeting. Type life insurance into Google, very high value ads. Mortgage, a car model, etc. This is why Google is so valuable.

- For display, the biggest lift from targeting is via retargeting (showing you the shirt you considered on a commerce site)

New York Times was definitely using ad exchanges only for their remnant inventory, so there is nothing surprising or interesting about this article.

> Intent targeting

I wonder how effective it is actually.

The narrative we are lead to believe is that the user will be thankful to get the information it needs surfaced right in front of them. Or the difference between the paid ad and the organic results will be blurry enough to not have an impact.

Yet I think it works only if:

- the ad effectively matches what the user is looking for (that's not a given, even amazon throws a lot of random things supposed to be in relation to search results), and if the product pushed by the ad is seen as legit.

- the ad is from a brand the user somewhat trust in the first place. For instance most people won't choose an ad from an unknown phone maker if they were searching for an iPhone (or they're getting scammed, and that's another issue). I see very few things average people would just search on Google and buy from a random supplier, especially for mortgage or anything high value.

What I am getting at is, there's a significant amount of effort needed on the vendor part to make an ad work, and I wouldn't be surprised if the brands going these length are not already appearing pretty high in the search results in the first place. Having the top stop could still make a difference, but that's just a nudge, and not something critical or that valuable compared to the rest.

> The narrative we are lead to believe is that the user will be thankful to get the information it needs surfaced right in front of them.

This is not the point of advertising at all. You’re operating on incorrect assumptions.

There are many different objectives of advertising. Imagine you’re in market for a vacuum. You might only be aware of a few brands, so other brands need to advertise so you become aware of their product. You may have forgotten about a brand that you previously liked, so they advertise. You might be considering one vacuum, but are potentially open to spending more on a premium model. You might be doubting the trustworthiness of a brand, until you seen them advertising on NYT. You might also need a new broom if you’re buying a new vacuum. What about this new product I invented, scented bags. How am I supposed to inform you about those?

Targeted online advertising is at its most effective if you are a small player targeting a very niche market. When you're a well-known brand, you won't see the same ROI. Also, you'll be able to get away with a lot more than you would with a TV or print ad, that would require human verification.

Take something like Alex Jones' male virility supplements. There are numerous 'brands' selling this type of snake oil to vulnerable, ideologically zealous people. Brands like these are fly by night, they wouldn't invest in an ad on say, The Economist.

But you would definitely want that ad showing on someone's feed when they've already indicated they follow pages and personalities that make it obvious you're the target customer, i.e. an easy mark.

I keep seeing the same ad over and over again even though I'm clearly not interested. The variety that I see is really narrow. Wouldn't it be better to give people a much higher variety of ads rather than bombarding them with the same ones over and over?

It sounds like you're being hit with a retargeting ads (you look at a fridge, and then suddenly the entire internet showing you fridges from that, or related, retailers).

Retargeting is considered to be a pretty efficient ad spend. Though sometimes I do wonder if the numbers are inflated by people like me: I'll tend to look at an item, throw it in the back of my mind, and purchase weeks/months later after some consideration. It's a longstanding habit, and I block the very vast majority of ads -- so retargeting doesn't factor in as far as I'm aware.

NYT advertising is still targeted to their demographic though. NYT would know their demographic pretty intimately at this point.

I would be surprised if there is no difference, but advertising worked for decades without hyper-targeted ads, I'm sure it will keep working.

Targeting ads matter, this is not really a debate.

It's a big paradox that I personally loathe Facebook but would be dead in the water without them as an ad platform: he are the only place that gives us material ability to target people. We don't go that specific or that deep, but it's far more than most others provide.

This is again an issue of the EU killing business with 'well intentioned' but possibly maligned legislation. The NYT as mentioned in the article is a major brand, and they are effectively selling that. Buyers know roughly what they are getting, and transactions back and forth are not small.

Small companies do not have even the budget to access NYT, even then it would be a bad idea in most circumstances.

Smaller companies depend generally on very actionable ads, whereas only at a certain scale does the marginal value of ads matter for brand campaigns etc..

So kudos to the NYT, but this is not good for many as we now have one (major) less place to advertise.

Does this include "remarketing"? From what I read it's much more effective. For those that don't know, remarketing involves showing ads to only people who have visited your website, or interacted with your brand before. Is this considered targeted?

I always wonder if this really works? Amazon uses it a lot and all I get ads for are products that I already bought. Which is pointless.

Anecdotal, but I've ran ads cold, targeted and re-targeted ads on Facebook for clients and eventually dropped re-targeting from the service I offer because it's extra time to build the campaign, extra cost and little to no benefit in terms of lead acquisition.

It is possible for advertisers to configure it so that this does not happen (in theory), but a lot simply do not bother.

I've seen some compelling case studies that it does work, and even at low conversion % for big-ticket items (like hotels etc) it's insanely cost effective.

If I shop around for a lamp on 5 different websites and finally buy from a 6th site, how do those first 5 websites learn that I bought a lamp?

The pervasive tracking of which retargeting is the most visible form is mostly what drove me to use an ad blocker. I wonder if the cost effectiveness calculation factors in lost future purchases?

> how do those first 5 websites learn that I bought a lamp?

Even Amazon's ad servering systems don't seem to be told that I bought the lamp from them never mind some other store.

Maybe they are hoping I liked it so much I'd decide that I need another lamp, so they want to remind me where I got the first one from?

> is the most visible form is mostly what drove me to use an ad blocker

My reason for blocking ads was because of the increasing incidence of malware riddled ad hosts often being used by high(ish) profile sites.

imgur.com was the final bail of hay that made me install network-wide protection at home. I also rehost any image I decide to share from there, to protect people outside my LAN. I don't want the tech support workload of cleaning other people's machines after what I'd too often seen their (obviously not sufficiently verified) ad partners try to pull.

My privacy is fairly shot already, but at least I can try to protect myself and others from other malware.

Assuming that five different websites would all run a re-targeting campaign for the same lamp - it's tricky, and you would probably be relying on the retargeting cookie expiring, or you hitting the frequency cap for each.

It's possible that at the level of the ad network the products could be disambiguated, and then cancelled so that other companies did not waste ad spend. This would be difficult as the products would have different SKUs.

At the level of a single retailer, the retailer places a burn pixel on the checkout page which registers to no longer place ads for that product.

Certain hotels and services are already targeting a captive audience.

If you're looking for hotels in London, you're already intent on finding a hotel.

A single 3 night booking in New York City will easily produce $100 of revenue for an online travel agency. So... Throwing $10 for a click is hardly a bad investment.

Other products and services are not like that.

Look for food delivery on Google and check how many ads are unnecessary for you

Actual retargeting (= targeting a previous visitor/buyer of your product) works, and in some cases works very well. Reactivating a customer vs acquiring a new customer can be up to 10x cheaper, depending on how well you do it. And if you know how to do that, then you can have a much higher initial CAC, which is often the hardest thing to optimize.

I think what you're referring to is different or more advanced. Simple remarketing involves putting a tracking pixel on your website, and then your ads show only to those users on another platform such as Facebook. It usually (on the base level) does not include purchase history.

I think that's the exact problem to which GP poster is referring: tracking only knows about URLs visited but not purchase history. A visit to the product page "vacuum cleaner WheezyMatic 3K" will get you flagged as interested in that product and you'll get shown ads for it — even though you already bought it on the original site.

Maybe the ads could have content along the lines of "The WheezyMatic 3K: The Best On The Planet" and so when someone who already has bought the product sees those ads, they're instead reassured about their purchasing decision.

I find the more effective version is showing you different products that either complement something you bought or are substitutions or variations on something you considered but did not buy. (Either browsed or added to cart but did not complete purchase)

This has nothing to do with effectiveness of targeting. It's about advertisers having no data to target so they do bigger blind ad deals and deal with fewer but bigger publishers.

It's why TV ads are so expensive and wasteful, but lucrative for networks.

I would bet targeted advertising works best for certain classes of things, but I rarely actually see targeted ads for them.

Like, fashion and clothing can be pretty individualistic so it might actually help to micro-target. Not even just for styles, but if you're a swim suit maker who specializes in plus-sized clothes you can target ads at them. You could even not bother sending ads for a sale on jeans to people with, say, size 30 inseams if you're sold out of jeans in that size. I never see them used this way though.

But almost every use case I can think of that is mutually beneficial (like the sizing), seems to be demographic information rather than behavioral analytics. I'm having trouble coming up with a way to leverage behavioral analytics in a way that doesn't seem invasive or impertinent. I guess it might work for content curation/recommendation engines (movies, books, music, etc.) But even then, you just need to know people's backlogs and basic demographics, you don't need to start snooping on their web browsing behavior.

> I honestly would not be surprised if the difference in effectiveness of targeted vs untargeted advertisement turns out to be none, negligible or even unfavorable to targeted.

The devil in the detail is how well the targeting is done.

Is there a big difference between poor targeting & untargeted? Not really.

Is there a big difference between good targeting & untargeted? Massive amounts of data to say yes.

I'd like to see some data which was collected in a quantifiable, carefully controlled and rigorous way. I'm not trying to be snarky or saying it doesn't exist, I'd just like to see some credible data.

I was able to find this study after banging my head against a bunch of paywalled sites trying to charge money for public domain research:


Working on reading through. Seems like it would be difficult to control an experiment like this, since the subjects know they will be seeing advertisements as part of a study and won't really be purchasing these items. I don't know, I'm looking forward to learning more about the experimental procedure. Also they seem to be measuring somewhat vague self-reported impressions about the ad in context of being targeted / untargeted, not the actual sales they produce. Drawing a line between these impressions and real ad performance seems to be mostly a matter of opinion. As I said it seems like a tough thing to quantify, and I'm hoping that people smarter than myself have found credible methods to truly measure it.

Do you have anything you'd like to recommend?

Ads that are not in context, do not convert, unless they appeal to a majority (e.g. the big brands).

Note that ads can still be targeted, except that the article's context (or whatever the user is looking at) can be used instead of the user's profile and you're not violating anybody's privacy. And on the web this works great due to the long tail, available articles, movies, searches, etc. telling you a lot about what the user wants, without having to profile that user. It's not like on TV where the content has to appeal to a wide demographic, quite the contrary.

The problem with ad exchanges is that they are designed to violate the user's privacy by leaking user data on purpose, like the user's IP address and now with GDPR they rely on good actors, because the ad exchanges can't control who gets that data based on consent, with the threshold for vendors signing up being really low.

I think there is evidence that in some cases targeted advertising works fairly well. In the book, "The Power of Habit," for instance, it goes into the Target story about their marketing to expecting mothers. This for instance is an opportunity for marketers and businesses to acquire new customers at a point in their life when they're likely to shop somewhere else and become regular customers.

They were able to do this by analyzing purchasing patterns and according to the book were fairly successful.

I'd imagine that online tracking when combined with other sources of data provide a lot of insight into what customers are looking for and in some cases can provide a big payoff.

> Sure, NYT is not enough data to be significant

Even just a few hours of NYT ad traffic ought to be statistically significant enough to say that the difference is within a few percent.

The problem is that the data might not be _representative_. NYT traffic might behave differently enough from yours that their results don't apply to you.

You mean that following the users with the same repetitive ads across the net (for things that he might have even bought already) is not so effective?

Yeah, tracking probably throws the ads into a local minima. But ad algos don't care about conversion, they care about charging the ad buyer more.

> But ad algos don't care about conversion, they care about charging the ad buyer more.

Bingo. The users clicking the ads aren't the ones funding the advertising networks - it's the people buying the ads. If you can convince them that targeted ads are more effective (and it would certainly be a safe assumption based on most peoples' understanding) then you can charge them more.

I'm not sure why you're being so heavily downvoted for this, but the presence of so many people "in the industry" in this thread might be some indication.

Yeah, I've been struck by this possibility while reading a lot of the rebuttals here. "Targeted ads work because otherwise why would they be so prevalent?" or "targeted ads work because x, y and z." There could be a kind of meta-advertising at work here, where adtech is often actually marketing towards the ad market itself. Would that mean it works? In a way--clearly lots of people are buying into it. And people wouldn't necessarily notice the scheme, because after all even if targeting isn't as effective as they claim it's still a form of advertisement and would generate some sales as dumb ads regardless of whether the targeting is really effective. We'd have to measure how much above that baseline the ads are driving actual sales somehow.

But I've yet to see some actual controlled evidence showing clear data in terms of sales that we can be reasonably sure were generated by targeted advertising. I think such evidence probably does exist, I'm not saying I don't believe it just yet. But all the arguments here seem to confuse science with scientism--what we have is a lot of reasoning and academic explanations of how people think it works, which is just not the same as evidence. It doesn't matter if something sounds like the way it might work in real life, we have to make observations to call something data. I haven't seen much data here.

From professional and personal experince, remarketing ads, what you are describing, are much more effective ($ spent / products sold) for the advertiser than traditional ads.

So why do ads follow you even after you bought the product? Well the ads are still less than 1% effective (products sold / by ads shown), so for the average company (not Amazon) it is just not worth the engineering effort for a < 1% savings on your remarketing budget. Secondly, this is more speculation, but for many products the chance of buying a product given you have already bought one (e.x. a spare or a gift) is still higher than chance for a random person. As such the ads are still effective.

So what is the deal with the NYTimes? They a huge popular publisher (especially amount rich people) already command high ad prices through traditional adverstising. The value of a remarking ad (to the advertiser) is largely independent of the site that it is shown on. So The New York Times would naturally have fewer of these ads because the adverstiser gets more ads/$ on other sites. A diffrent site like a local paper might have a diffrent story. However remarketing ads have never been a large percentage of ads shown so from a revenue stand point publishers care a lot less about them.

That's not a good definition of effective. The advertiser actually cares about ($ spent / additional products sold that would not have been otherwise), which is much harder to actually measure as it involves a real causal question as to whether the ads influenced a purchase. Much of remarketing spend is likely wasted on customers who were going to buy the product anyways, especially so for people making a repeat purchase.

Its a not very well kept secret of internet advertising that it's so much more "measurable" than other types of advertising, but it's not actually measuring anything useful most of the time. There is the traditional quote that "I am wasting half my advertising budget, but I can't tell which half" With internet remarketing ads the answer might very well be both ;)

You mean advertising a Widget to someone who has just bought one? Yeah, I can see how that would be counter-productive.

I think even Amazon suggests other cellphone models "because I bought 'Y phone' recently"

YouTube keeps presenting me French advertisements as a Flemisch person. So I guess "targeted ads" will remain not as targeted as you think.

Also, when I just bought some appliance online, maybe it's time to stop showing me ads for it on every page that I visit.

What’s your browser language set to? Are you actually in Belgium?

If your browser is sending “Accept-Language: nl-BE, <other stuff>” I wouldn’t be surprised if they showed French ads. It’s one of the official languages of Belgium after all.

They may not have any relevant Dutch inventory to show, or they’re getting more money for the French-language ads which also target Belgium.

My browser language is English, and I live in Flanders. I never search, read or watch anything in French. So there is no reason to show me French ads, except that I live in a country where there is also a French speaking part.

I honestly don't care why they show French ads, but you cannot call such things "targeted", if you cannot even get the language correct, out of a choice of either Dutch or English.

But also does targeted advertising increases the size of the advertising pie? Companies have an advertising budget, they will spend it where they think it will yield the most sales. But is that budget increasing because they can do more targeted ads?

This makes sense to me. If you are targeting me, I have told you I like it, and therefore don't need you to tell me about it.

The worst kind of targeting is when I go to a website, browse some similar items comparing their features and price, buy one... and now you decide to show me ads about those same products for the next week even if it should be clear that I've already made up my mind (I'm looking at you, Amazon).

Customers who bought this television also bought:

    1) different television
    2) another television
    3) alternate listing of the same television and 4x the
       price because it's controlled by some idiot algorithm
    4) more television
    5) yet another television
Suuuuure they did

One possible explanation I've heard (of which I have no idea if it is true) for both this and ads showing after you've made a big purchase is that people actually are somewhat more likely to buy an appliance close to another purchase of one, since they only buy them a) after several years, if the previous one has broken or is outdated, or b) shortly after they bought another one, which they returned for some reason.

Pretty sad state of affairs if that's the case, you'd think Amazon would do something to exclude repeat purchases after a return. But it's been this way for years.

As it stands, 99% of their product recommendations are saying "I assume what we just sold you was a total piece of shit and you'll probably need another one after you return it. Here are some options!"

Can you not use fixed-width for block quotes please? It’s very difficult to read on smaller screens!

I'd avoid it for anything important, but if you can read the first ten characters of each line I think you get the point of my list there. Didn't want to waste the space with paragraph breaks between items.

This is a matter of custom, fixed-width is for code [0]

It is more typical to italicize a blockquote, perhaps with leading `>` to mark it out:

> like this

- and this is a list-item

- and so is this

it's just a matter of courtesy to those using smaller screens and assistive technologies


[0] https://news.ycombinator.com/formatdoc

I agree in principle, but

- I don't

- think that

- individual paragraphs

- for a list

- is a very good

- use of space

- on small screens

- or on large ones

Code blocks are crappy for small screens, but so are lists written as a series of paragraphs.

Unfortunately, HN seems opposed to supporting non-paragraph linebreaks, lists, blockquotes, or any other sort of formatting. That's fine, it's up to dang and the other admins who run the place and they don't want to support other formatting, but it means that the code format gets abused as a poor approximation for these things.

EDIT: I should also point out that your source says "fixed-width is intended for code." You left that word out and it's an important one. If it said "is only for code" the meaning would be different.

Side note, reddit lets you do a line break by ending a line with two spaces before hitting enter. Just tested to double check, it doesn't work here. I wish it did, and replacing space-space-newline with <br> seems like a pretty trivial and unobjectionable feature, but here we are.

In terms of usability, vertical space almost doesn't matter anymore. The vast majority of people have an interface that makes vertical scrolling effortless, whether it's a trackpad gesture, mouse scroll wheel, or touchscreen drag / flick.

Horizontal scrolling is harder to do and less common/conventional (chicken and egg as to why, but it's the reality today).

You might not like the way it looks, but it's clearly better for usability to spend more vertical space in order to prevent horizontal scrolling.

I don't know, I guess this just doesn't seem unreadable to me:


It's a list making fun of Amazon's stupid duplicate product recommendations and you don't need to scroll horizontally to get that out of it.

Screenshot from an iPhone SE which is a damn small viewport by 2019 standards.

sorry it wasn't an intentional omission. I was being brief.

   The problem is that anything beyond a certain length gets clipped and has to be scrolled from side-to-side which is a very hard way to read.
However I can quite


scroll down all day and it

is more harmonious with how

my brain likes to read!

As a HN reader I understand and accept the limitations of the formatting.

There are different sections, "also bought" and "related". They both show up in the same formatting.

We notice you're interested in car insurance, how about some car insurance?

Yeah, I was, you're right, I spent a couple of hours looking at car insurance stuff and now it's done for the year, so you're spending your advertising budget chasing people who are the absolute least likely to be interested...

That sounds vastly oversimplified. So you’ve discovered that I’m into hiking/camping gear. That doesn’t mean I will never buy another piece of outdoor gear or that I have heard of every brand and product in that category.

Yeah but crowding out my browsing experience with things I’ve already bought you’re losing the opportunity to sell me other stuff I mightn’t have even thought about!

if you really are into hiking/camping gear, you probably know where to find it and at what kind of price, or perhaps you know you will actively search and compare when buying, so the ad is of little relevance.

I think advertisers would be overjoyed to hear your low estimate of your susceptibility to ads.

I often get targeted ads for upcoming shows in my area. Just because you have identified that you like a subject does not necessarily mean you have spent all the money you are going to spend on that subject.

Depends on how the targeting is done. You can be aware of a product but not looked into it or purchased it. In that case the targeting is trying to move you along to become a customer.

How would a small business advertise a new short-term promotion to people in the area when all local papers are gone?

I wonder how many people are turned off by targeted ads because they find them creepy.

targeted advertisement is not by itself inferior or superior to untargeted advertisement. Method can only be effective when used correctly.

If you believe that, have I got a Koi Carp to sell you.

most of the big websites can afford to stop tracking. The problem is that smaller / weird niche websites can't , and gdpr is disproportionally hurting them.

Why do weird niche websites need more tracking? I'd assume their user base is way more specific, and thus can be targeted better without needing tracking. That I visit the NYT tells you way less about my interests than that I'm reading your blog about aquarium filters.

I can only give you speculation on what I’ve seen in online video.

If you are a smaller publisher, unless you get particularly active in courting advertisers (i.e. spend time going into sales), the niche advertisers will never even look at your content. The 20-something digital ad manager has no motivation to risk his job doing speculative ad spend. As a result your ads are less relevant and your inventory’s CPM drops as a result.

If you are a large property, then tracking affects you less (in fact tracking might be worse for you). Large publishers already have sales and marketing teams and ad managers will just choose you because you are “safe”. If brands are moving to a world where tracking data is less and less reliable, then by default they will move to spending more money on a smaller amount of “safe” inventory

If you are a small publisher, why would your ad prices endanger any but the smallest of advertising budgets?

As an advertiser, you go down the tracking route and you decide to spend 1M targeting women 21-35.

With tracking, you spend all on an exchange that has a ton of supply. ELLE competes with JennyBlogger. You potentially hit a ton of women across different properties. You may be advertising to women on male dominated outlets (the “promise” of digital tracking)

With the direct route, an ad manager may spend that 1M on specific properties, maybe between ELLE, Vogue and InStyle. This is a bit of a simplification as most firms have separate direct and programmatic spend, but the amount of money allocated to each depends on the perceived return of each inventory type.

this logic (barely) works for blogs. if your game is an aquarium game, it doesn't mean that your users are only interested in aquarium cleaning products.

> if your game is an aquarium game, it doesn't mean that your users are only interested in aquarium cleaning products.


On the other hand, the topic of the game is just one of the contexts. A trivial example:

At least we can infer that the player likes to play games, so advertising for mobile games, discount gaming consoles etc might be a better match.

yeah but the category "plays games" is too broad, rivaled by "reads blogs". And after the users have seen enough of gaming ads (and not responded) they are offeered the bottom of the barrel: russian brides (true story).

If we are going to go down the road of doing increasingly detailed ad customization as publishers, we will have come full circle to reinventing targeted ads by another name.

> yeah but the category "plays games" is too broad, rivaled by "reads blogs".

Plays x game on y website at z time of the days however should tell you something though.

the question is how it overlaps with what the advertiser wants to promote. e.g. "is looking for new jeans"

It doesn't overlap with everything a random advertiser might want to promote.

But based on the site, game and other indicators it might be a good place to place an ad for:

- stupid mobile games (assuming the audience is just bored)

- discount consoles/games (assuming audience is playing it because they don't have access to a console)

- toys (assuming audience is kids)

- etc

Targeted ads have always been around. Everyone targets their ads. Decide whether to put your ad in the Times or the Sun.

Its the unsolicited tracking/stalking of individuals across communication networks GDPR forbids.

Source for this claim?

Some of this shows it's hurting smaller adtech companies (read: advertisers who aren't google, facebook, amazon) but nothing here really shows it's hurting smaller site's revenue in general. There's some speculative stuff like how investors are reacting.

Which is a non-argument if it turns out that tracking doesnt contribute to higher ad revenue

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact