Yes, your intuition derived from a single anomalous data point (yourself) is definitely more accurate than the companies which have seen demonstrable success from targeted advertising.

If you wanted to argue that brand advertising doesn't work, I might agree. Brand advertisers tend not to have high standards of data analysis.

Your publisher experience seems to actually be a case of wanting the former (targeted advertising) but getting the latter (brand advertisers), which can be a real problem because brand advertisers often overspend and undertarget.

What WB discusses is a common misconception. Just because it doesn't appear to work for yourself, doesn't mean that targeting is not worthwhile. It's not really designed for an improved user experience, but to provide an advantage to those running the ad campaigns.

The crucial thing that people forget, is that that targeting only has to be a little bit better than random to be worth doing (from the ad operations perspective). Given a large number of ads shown every day, with nominal cost, we've only got to push a fraction of those toward users with a slightly better propensity to interact to increase the interaction rate, and get paid more from the advertiser (presuming a model where the advertiser pays us based upon interactions that we generate).

One of the reasons that it is profitable is that the amount we might get paid for an interaction could 50p, £1 or even £10 depending upon the client and media. The cost of showing that ad is a few pence per thousand! So, we can get it wrong many many times, and still show a statistically significant improvement in interaction rates and make a huge dent in profitability for ourselves.

I discuss all of this in Chapter 5 of Algorithms of the Intelligent Web. 2nd Ed. https://www.manning.com/books/algorithms-of-the-intelligent-...

And here we descend into spammer-logic. It took a decade or more, but spam filters are good enough now that spam is almost a non-issue; the flip-side is that email, which was designed to be reliable above all else, is now flaky because some legitimate mail can't find its way through the maze of filters, many operating according to machine learning that defies introspection. Ad blockers will eventually be that good, will break web pages in equally unpredictable ways, and the "Open Web" will be a worse place for it.

But I guess you make some money from your book.

"Single anomalous data point"? His description is exactly how targeted advertising works for pretty much everybody. He's not some weirdo outlier -- we're all show the same things we've bought and tangential things to our searches with the stupidest ad AI.

He might be wrong that "it doesn't work" (e.g. doesn't pay since it's crude), and probably he is, since so many businesses follow it, but he's totally right in that it's crude and far less effective than its potential effectiveness.

Now we can debate if it ought to work, if it's ethical etc., but let's get the facts straight, it does work.

The parent described how it's broken and works crudely and bad.

That it still works and gets some sales, he probably doesn't doubt.

FWIW, since we're doing anecdata, I just checked my Amazon front page. Many things are irrelevant, a few comical (I recently looked at a 80-book box-set of Penguin classics[1], now it's systematically recommending each individual member of that box-set to me :-/ ) -- and a few are quite relevant and I clicked on them to check it out.

1: http://www.amazon.co.uk/Little-Black-Classics-Box-Penguin/dp...

And that's in a nutshell both the solution and the problem because users have only so much to spend on product. So you end up with an arms race and that's exactly what we are seeing.

It's like the joke about two guys in a jungle, one with boots, the other with running shoes. The one with boots asks why are you wearing running shoes. The second one answers: In case we get chased by wild animals. "You won't be able to outrun a wild animal so don't bother, wear boots". To which the guy with the running shoes says: "I only have to outrun you".

I don't see the problem? It's probably much more expensive per impression than traditional advertising, but probably much less expensive per sale (which is the metric that matters). This arms race means that if I'm a plumbing supply company, I can now spend my money on displaying ads for specific plumbing goods to people searching for specific plumbing goods, even when that means I'm going to hit some people no longer in the market for these goods -- but even hitting people who are no longer in the market is vastly better then hitting people who were never in the market. This lowers the barrier to entry (less cash required to advertise your products) which means more and better competition which means prices go down.

They don't appear to be targeting people who are in the market for plumbing supplies and hitting a few who already have them, from what we can tell. They appear to not be targeting people who are in the market at all until after they're out of the market.

The only places in which people are really competing directly with each other are in the labor market (which drives the price of labor down) and in professional sports or other one-on-one contests.

Almost everywhere else we co-operate before we compete.

So far this model is a bit like democracy, it may not be optimal but it seems to be the best we have. If you can improve on that model you'll go places but you're going to be ironically in competition with everybody else!

Also interesting note: from my experience, most tech companies aren't really competitive. The market is big enough that most successful companies either compete superficially or sometimes outright cooperate, even when working in the same market.

It doesn't matter if it's proven or not. We're (hopefully) not automata just doing what's proven to make us excel without consideration for the others.

But a very low one. Conventional advertising is what, 0.1% effective?

I find my Amazon recommendations are often pretty good. And I'm willing to spend time curating them. If advertising on the rest of the web was as relevant as my Amazon recommendations then I'd probably stop blocking it.

I'll raise you mails that I get from Alibaba - after few months of receiving offers of aerospace-grade struts, I'm now getting mails advertising... stones.

(I do know how this happened though; the first is the result of me searching for aerospace-grade struts after CSR-7 launch failure, seeing if I can find the unnamed vendor by looking if someone is dumping cheap struts on Alibaba; the second is the result for me accidentally typing in "Pebble", meaning Pebble Watch).

So there's probably a significant number of people where targeted advertising is able to get the click that leads to the purchase. In reality, it's probably just taking organic traffic that would have purchased anyway and counting it as a targeted advertising win versus an organic purchase.

Or did the ad nudge you toward the conclusion that it is right for you, before you got over it and forgot?

Are you sure?

If this is an unassailable fact because data, where can I see that data?

The anomaly is not how targeting works, the anomaly is that it "doesn't work" on him. The fact is that most people do occasionally buy things because of advertising, particularly targeted advertising.

"Doesn't work" is in quotes because I suspect that he has in fact bought something based on a targeted ad in the past.

I'm not so sure. It's quite a big claim to say online advertising is effective on "most" people. It's also quite possible that it's not, and that online advertisers are all chasing an anomalous minority population that is not yet ad-blind.

But for 99.88% of people seeing those plumbing ads -- including many folks who have already bought their sinks -- such ads are stupid distractions. The net result: there's a big surge in the use of ad blockers, and the swamp is starting to drain.

Another case in point: I now follow Twitter trending reports for Yokohama, even though I live 5,000 miles away and can barely read a few characters of hiragana. It's my primitive hack to disable the trending "service." I'd rather have incomprehensible squiggles on a corner of my screen than a bunch of garbage content that's clamoring for attention, even though it's irrelevant to me.

And yet, if I buy a coffee maker, the front page is covered in coffee makers for weeks. I've never bought anything based on what's on the front page and I'm not sure I've ever even clicked those suggestions. I'm sure targeted advertising can work in the long run, but at the moment the success rate seems so low that it's more often an annoyance than a convenience.

* buy an item from Amazon (coffee maker or otherwise)

* realize the build quality/size is rubbish (which from some of the less well known sellers, it really can be)

* return the item using Amazon's generous returns policies (never had one rejected, I've used it many times)

* buy a different item in the same category.

* repeat?

If this type of purchase happens a non-negligible number of times, could that be the reason we see so many 'buy more of X!' style ads?

Haven't you ever wish for a 'Punch Advertiser' button during commercials? This would be like hearing they built it themselves.

(To be clear, by having someone making $0.03/click sign their customer up for buyer's remorse thus destroying the profit from serving them because those return policies, the customer service time, etc, aren't free.)

The following is a very interesting read: http://www.ft.com/cms/s/2/cd1722ba-8333-11e5-8e80-1574112844...

As far as the success is concerned, the advertising company is hopping from one aggressive technology to another mostly because their techniques work but only for a while and then users tune out. So it may be that WB has a point that is rather stronger than you make it out to be.

If we're comparing anecdotes, I'll offer the 3 companies I've helped build online advertising campaigns for which had a positive ROI.

> hopping from one aggressive technology to another

I'd agree with that if the argument was that "this specific ad technology" is going to fail. I totally agree that ex. clickbait headlines are a temporary boom which will disappear soon.

But to argue that advertising as an industry is in a bubble, we'd need evidence that the industry is bigger than historically or what it can support. Which there just isn't evidence of (take a look at my other post).

Making an ad campaign that has a positive ROI is doable, but if you are that well embedded in the industry then you will quite likely agree with me that it has become quite a bit harder to achieve that today than it was in the past. The solution (performance based advertising) shifts the burden of the effectiveness of a campaign onto the wrong party (the owner of the inventory) which has resulted in a series of start-ups that optimize on behalf of inventory owner.

This is all like watching eb-and-flood erode a beach and just like real life erosion there are turning points at which the situation can go from steady state to dynamic quite rapidly and it is anybody's guess what the new steady configuration will be once all the potential for movement has been exhausted.

The ad industry is imo not exactly in a bubble as much as that they are more than a little bit worried about why the usual bag of tricks no longer works and how they plan on going back to when conversion rates were still something to be proud of. I don't doubt that this will initially be tackled with a wave of even more invasive advertising, or advertising masquerading as content.

The problem is that what was initially an absolute boon for the advertising industry, that you can 'measure' engagement ended up being the exact problem, you can measure engagement, including a steady drop in engagement over time. And so, by making advertising more and more aggressive the industry has made it impossible - or nearly so - to make a non-aggressive campaign because it has to compete with all the aggressive ones for attention (from the users perspective it is a zero sum game, they only have 24 hours in a day and usually something else to do besides watching and engaging with ads, and if they don't then they are not exactly prime material for some demographic to begin with).

So now there is a problem, which is that the ad industry can't suddenly stop reporting all these engagement metrics, nor can they pull more technological rabbits out of the hat that spawned all these new tricks because these too will work for a while and then run out.

It's not a simple problem.

There has yet to be a point at any time in the past decade where I said, this doesn't work, and shut all of the ad campaigns off. There definitely are issues that exist. There is a problem with mobile app ad quality. There definitely are bot traffic problems. The ad exchanges will be pretty brutal on inexperienced advertisers, but this was the case pre-exchange too. One "bubble" point is I suspect many of the adtech companies involved in exchange inventory may fine themselves on the line for the bot traffic. It will be enough to wipe out the investor equity. This follows based on class actions Google, Yahoo, and others have had to settle in the past and Google especially has always been aggressive about this stuff (although less so in the past two years.)

I read the original blog post and the writer doesn't seem to make a very good case outside of things that happen in normal capitalism. Lots of middle men, tools, and supply chains attempting to add value along with investors dumping in a ton of money in an attempt to gain some return -- possibly more than they will ever get back out. Sounds like a healthy marketplace to me. Some ideas are terrible, some are good, some are amazing but will never return their investment. That is ok.

The bigger problem is for small publishers and likely medium sized ones too. Ad inventory and the users online has ballooned many times over over the past decade. It also is a problem for traditional media companies like radio stations, newspapers, and television channels. I don't need to extrapolate on that one.

Advertisers need to be sophisticated. They had to in the past too but few had access to tools that told them they sucked. I'm not sure that is any more extraordinary than asking a commodity traders purchasing cocoa beans to know what they are looking at and know how to be sure that is really what shows up at port.

You're right that this is part of the normal cycle of capitalism, with all the creative destruction that entails. It's quite normal for a new, growing industry to support a whole forest of new businesses, and then as the industry matures some of those will die off. But I think there's one extraordinary thing here: the advertising industry is growing like a new market, the kind of market you see when the customer base is growing by 100% every year, like the early days of the Internet. But that growth simply isn't there; advertising is by most measures a mature market. So what's fueling all this expansion?

Advertising was a mature market when Google started. Media was a mature market when Facebook started. (And, to make a slightly different point, people believed computers were a mature market when Apple started.)

Is the question the expansion in adtech companies? If that is so, I think a huge part has been the Facebook traffic dumps to publishers like news sites. Basically the available display inventory something like doubled. That isn't sustainable, Facebook fully intends to recapture that external traffic back in to their platform. The drop is already visible to some publishers. Many companies certainly are turning a blind eye to fraud so as to hide the bad news from their clients and investors.

At some point the shift from offline to digital ends and the growth moves more in line with GDP. That will bring down valuations. Google and Facebook probably want to integrate the ad model with every human transaction. One could imagine getting in a Google car and saying "pizza" and be driven to the highest bidding pizza shop.

Certainly ad costs get baked in to some products in a big way: Coca-Cola, Geico, hotel rooms, for profit universities, and so on. Sometimes the costs are so much the profit center moves to another area, such as investing the float from insurance or maybe just straight up stealing money out of a person's bank account (pay day loan lead aggregators selling bank info to criminals, has happened many times, see FTC.)

As a whole, I don't know how this plays out. If the ad competition is brutal enough the costs will be driven up to the point where the parties not spending money on advertising are just far better deals. If it isn't the prices running up, the cost could be offloaded somewhere else like cutting maintenance at a hotel. Eventually the advertiser's reputation tanks.

These are marketplaces. On a micro level there are many perverse distortions because of some condition at a given time. Over time they should re-balance.

Some things like Coca-Cola, which has probably physically harmed and killed people on the scale of cigarettes, just last a really long time. Coca-Cola has a truly perverse combination of trademark law enforcement, addiction (caffeine+sugar), general scientific ignorance by consumers, and a long list of government subsidized cost advantages. All of these strengths save trademark law are being chipped away at right now. Which, at some point it flips and they have the McDonalds problem (toxic brand.)

Most of the time, I've been at the receiving end of advertising as a publisher (web/mobile). And I'm looking at the money stream from a different angle. What I hope (and fear the opposite of) is that for all the money I make with advertising (plus Google's share) there's actually more money earned (profit) by the advertisers. Because only then the system as a whole is sustainable as it is now.

The thing to remember about advertising is that at the end of the day it's really about finding ways to be different than it is about anything else. And so when everyone is doing TV ads, start doing online ads, when everyone is doing online ads, start making centent marketing, when everyone is doing content marketing use email and so on. It's an ever moving target. Thats why it's effective when it works.

My purchasing decisions tend to be pretty long and drawn out (it will take me months to decide that I want, for example, a camera and then to decide which one I want), but targeted ads have several times caused me to pick up the slack and continue my line of investigation into what to buy.

Their aggressiveness is an asset – and they also don’t really annoy me. Comical, sure, but they also remind me: Ah, I wanted to check out those cameras.

Obviously it’s a blunt instrument and not a scalpel (we know for to little about how human brains work to make that work – we need the blunt approach), but it’s a different blunt instrument compared to other advertising.

The problem is that although these stories exist, they're so rare. Drops in the ocean. All they say is that highly targeted advertising definitely works for a small part of one per cent of the advertisers.

(Which isn't to say that it doesn't work for more. Maybe it does, I wouldn't know.)

(a) On HN, it doesn't match the ideological overtones so it won't be voted up.

(b) In general, it's not interesting. Most business owners already know that advertising works and aren't terribly interested in hearing the obvious.

I'll trust his numbers. But is 20-100% enough to say "targeting works"?

And yes, targeting works. But non-animated banner ads also used to work, and they worked even better than quite a few targeted campaigns do today (hence the shift to performance based advertising rather than CPM and CPC campaigns, they are also much more resilient against fraud but much harder to properly set up).

There's an exception: Targeting geographical areas obviously works. If your market is the city of Toulouse, targeting by geoip gives you most of the intended audience and eliminates almost everyone else, at low cost.

The rest of the targeting techniques? Trying to pick male users, trying to pick users over 50 or between 20 and 30, trying to pick people who've visited faucet FAQs in the past? There's a long-term strategic cost in user revulsion and there's an immediate technical cost in implementing the tracking.

I won't try to put a number to the cost of creeping out users and making them install adblock, ghostery etc, but it does not seem safe to assume that the cost is smaller than the value of those 20-100% additional clicks.

But after all, those ads leave an impression on me, that coca-cola is still there, in a red glassy bottle/can, and people seem happy drinking it. That is the reason I might choose it as my go-to soft drinks other than some unknown small brands. So that is advertising works, laziness.

"...advertising was positively linked to brand prices in manufactured food products. Prices of weakly advertised manufacturer brands, private labels and generic products were substantially lower than the prices of leading brands, especially highly advertised leading brands. Since the products selected for analysis were relatively homogenous, real quality and cost differences do not appear to account for the price differences across brands."

This may or may not be a result of the paradox of choice. That would be an interesting paper: "brand advertising increases prices when there are fewer than five choices of brands" would be a testable hypothesis, but I haven't been able to find that paper.

Targeted advertising does in fact work. I know this because I work for an ad company, and it is very clear in both click- and conversion-rates.

The product specific level of targeting you are talking about, might be feasible for amazon... might be. But it isn't for our company since it is relatively difficult to maintain the segmentation and targeting rules required for it to work. The ROI is simply too low.

There are lots of other easier targeting options with a proven effect such as, frequency, geo, audience segmentation, site/product segmentation (such as the retargeting campaigns you have probably experienced), predictive behavioral targeting, and so on.

We typically see 1.2 to 2 times as many clicks on campaigns using a mix of the mentioned targeting options above, compared to campaigns with no targeting.

Which are, over time steadily dropping. Even simple non-aninmated banner ads used to have a high click-trough and conversion rate, higher than the best and most aggressive campaigns today.

But novelty can exist within a specific frame. The print ads of the 1950s stopped working, and a new approach to print ads was taken in the 1960s. The TV ads of the 1960s stopped working and a new approach was taken in the 1970s. Flashing banner ads that say "Click Here!" don't work the way they did in the 1990s, so a different approach is being taken. That approach will stop working and something new will be ginned up. But that doesn't necessarily entirely invalidate the technology being used or the basic form. Chances are the tech and form evolve, and the evolution will be driven by the dominant companies in the category. Revolutionary change is rare.

In addition to battle between adblocking and adblocker evasion, I think there are going to be a lot more sites with fusions between advertising and editorial (sponsored posts, product placement, and just good pr) making it more difficult to distinguish between the content and the advertising.

Targeting definitely works. A big entrepreneurial mistake is looking at things through your own eyes instead of those of users. Many brilliant coders can't get a handle on the business stuff because they just assume everyone else thinks and behaves like they do.

Ironically, the successful products I've created in my career I built to please myself. The ones I built to please others have all been failures. Apparently I'm not the unique snowflake others allege I am :-)

Given the amount of tracking they do, I'm surprised this is still a problem.

> soon, on a website near you

Yeah, screw you advertisers. I honestly tried to browse with ads, but I couldn't any longer and finally reinstalled AdBlock a few months ago.

Amazing you held out that long. The difference between 'with ads' and 'without ads' is so large I can't understand how people in the tech scene (for who it is trivial to set this up, luckily for the advertising industry not all that trivial for others but getting easier all the time) still browse without some kind of blocking. Besides the security implications of running third party code on just about every website.

Ultimately it was YouTube ads that play before videos that were the last straw; I sometimes run YT as music source and I got too pissed off about having to listen to 30 seconds of unskippable advertising in order to get to 120 - 240-seconds long video. Mostly it's mood-breaking.

Why do complicated things when something simple already works pretty well? Your thinking that just because something is possible it will be done is not very close to the practical reality in the trenches. Because something is possible it might be done if resources are there. If not, best effort, done.

All of this analysis is expensive, so don’t expect it all to be done.

Theory A: He wasn't shopping for faucets at all, he's just a bot following links and who had recently visited a faucet portal page.

Theory B: He searched, found a faucet he liked and bought it.

Theory C: He searched, then installed Ghostery, and is still in the market for a faucet.

All are possible, but one of them sounds a lot more likely than the other two to my ears.

Advertising to someone who you have reason to believe is going to buy your product is much more effective than advertising to an unknown person.

Totally depends on the product and the person you are marketing to. Marketing hammers to carpenters makes sense: they use many different types, and presumably go through them quickly. Marketing faucets to regular folks is a dumb thing to do. A faucet is something you buy once in a blue moon.

People buy stuff for others. People buy stuff for professional reasons. And yes people do need more than one faucet when they are building or redecorating a home for instance.

Also, showing people stuff they have recently bought may provide an incentive to recommend it to others or write a positive review online.

(Unless you meant just clicking "I bought it" to get rid of retargeting ads in general, not that you actually bought the product)

And even if such a button were present one could press it even though the product had not been bought.

Actually, you can: http://www.google.com/settings/ads

I think Facebook enjoys the current Adblock situation, the native advertising is the future. Good luck blocking those ads in their app which everyone is using. Also paid tweets, paid Facebook shares - I don't think many people on HN realize they are being tried to sell something all the time, just the fact your Adblock doesn't block it, it doesn't mean it's not happening.

I can't see myself browsing the web without Adblock, but I think the future is really dark because more and more people feel that way.

I wonder why they don't do this? For example, offer $1 Play credit for filling out a short form of your preferences.

Before I upgraded to no advertising on Hulu, they would sometimes ask after showing an ad if you were interested in the target market for the ad.

Machine learning/recommendation systems?

http://www.google.com/settings/ads

NaNaNaNaNaNaN Batman

More seriously: "20 to 40 year old male"

Programming -> Nerd -> Comics -> Batman

You can see plenty of this in, say, music recommendations where Youtube will follow a Foo Fighters song with Taylor Swift because they're both in the same "top 100 popular" category. It's easier than understanding content.