Live data from Hacker News

Ask HN: Does targeted advertising work at all?

news.ycombinator.com

101–110 of 158 posts

Re: Ask HN: Does targeted advertising work at all?

#101
post #59
post #8

Earlier quoted context omitted.

> It’s hard to tell when someone has bought a product. How about, "I've literally got it in my last-day purchase history on eBay and they keep showing me the same thing and a bunch of related stuff to the item that hasn't even arrived yet, but that they've taken my money for"? ... and then eBay follows me around showing the same stuff on other sites. The thing I've purchased. From eBay. At least, they did, until I go…

My standard response to this: Prior purchases do have predictive power. They can very well be positively correlated to further purchases. I've often bought stuff that didn't work for some reason. It was broken, or it had something that was bothering me. Maybe the power adapter was annoyingly loud (looking at you, Kyocera). So I sent it back. But before I managed to go to the postal office a few days later I was alrea…

But before I managed to go to the postal office a few days later I was already searching for a replacement online.

While plausible, that sort of situation doesn't sound common enough that it should underpin the entire online advertising industry.

Even right after I have bought a graphics tablet I'm infinitely more likely to buy a graphics tablet tomorrow, compared to my parents who have never bought one in their lives, and never will.

But you're more likely still to buy another stylus, or art software, or an accessory, or physical art materials, etc. Why don't advertisers understand that?

Re: Ask HN: Does targeted advertising work at all?

#102

What is the strongest signal that someone will be buying blinds the near future? a) Someone recently bought blinds b) Someone bought blinds 5 years ago c) Someone hasn't bought blinds ever (maybe they're renting, or just tape newspaper to their windows) Turns out, on average, the answer is a. Either because you're returning blinds you don't like and are in the market for replacements, or because you're doing a home r…

Where does this data come from? What do you think if you included the option d) Someone recently made a search that included the word "blinds" ?

A person in category a probably also searched. The thing is just that a recent purchase isn’t a strong enough signal to not retarget. It’s a too big risk to take because the probable recent buyers are fewer so it’s not worth filtering them out. At least that’s my impression of how it works.

Re: Ask HN: Does targeted advertising work at all?

#103
post #34

Earlier quoted context omitted.

It is entirely possible that large corporations measure the effectiveness of their advertising by comparing it to sales, and are confusing correlation with causation on a massive scale . They spend more on advertising, and point to sales in the same time frame, as a cause, even if reality disagrees. There are zero incentives for anyone to figure this out and fix it within corporate structures. I'd be thankful it make…

You clearly have no idea what you are talking about. Entire business are built on testing the effectiveness of online advertising, and providing that as a consulting device to brands. Do you think Nike would be happy if their vendor just said, “yeah everything drives positive lift”. Nobody is confusing correlation and causation. These are rigorous studies completed by literal data scientists and statisticians. They a…

And then there's the example of P&G, or Unilever or some other FCMG company that stopped advertising and saw no change in their business

Bob Hoffman's take on advertising e.g. Advertising for Skeptics, is well worth reading - he's an ex-adman who thinks the ad industry and particularly the personalised ad industry is full of shit

When you look at the online ad industry it's reasonably clear that the people who are making the money are the ad tech companies and not the advertisers

Re: Ask HN: Does targeted advertising work at all?

#104

What is the strongest signal that someone will be buying blinds the near future? a) Someone recently bought blinds b) Someone bought blinds 5 years ago c) Someone hasn't bought blinds ever (maybe they're renting, or just tape newspaper to their windows) Turns out, on average, the answer is a. Either because you're returning blinds you don't like and are in the market for replacements, or because you're doing a home r…

Back of a napkin calculation: showing OP 100 ads of blinds probably costs 0.1$ in the US. The cost of a blind is 100$ (I have no idea, don't live in the US). So if the probability of buying a second blind is at least 0.1% it was worth running the campaign.

The fact is, online ads are super cheap.

Re: Ask HN: Does targeted advertising work at all?

#105

Mostly no, it does not. Actually there is nothing wrong with targeted advertising, products and ads are meant to target specific people in its nature. But online advertising are a scheme for tech companies to make money. Adtech was created to make the buying and selling of online advertising so much more efficient. Today, about $350 billion dollars is spent on online advertising. 70%+ of it is bought programmatically…

I can't follow your comment. You say it doesn't work, then you say there's nothing wrong with it, then you list a bunch of things wrong with it. Also, a few things: - Products and services are make to make money for their creators. They generally try to achieve that by creating something of value others will buy. So this isn't really an argument for or against it. A Prius works and it is a scheme for Toyota to make m…

Advertising and online advertising are different things. Advertising was a thing, long before technology. Great ad people (like David Ogilvy) was created effective, targeted advertising and achieved success.

How to Create Advertising That Sells Ad by David Ogilvy [1972] https://boraoztunc.medium.com/how-to-create-advertising-that...

What I meant was, there is nothing wrong with advertising, or targeting right people to deliver value, products, services to them with creative approaches. Online advertising though, especially programmatic, is totally a different scheme.

I suggest reading more about; "Surveillance Capitalism"

Professor Shoshana Zuboff, author of “[The Age of Surveillance Capitalism](https://en.wikipedia.org/wiki/The_Age_of_Surveillance_Capita..., who coined the term surveillance capitalism in 2014, defines it as;

" …the unilateral claiming of private human experience as free raw material for translation into behavioral data. These data are then computed and packaged as prediction products and sold into behavioral futures markets — business customers with a commercial interest in knowing what we will do now, soon, and later."

In an interview with The Gurdian, Zuboff explains how surveillance capitalism is a human creation;

>It lives in history, not in technological inevitability. It was pioneered and elaborated through trial and error at Google in much the same way that the Ford Motor Company discovered the new economics of mass production or General Motors discovered the logic of managerial capitalism.

>Surveillance capitalism was invented around 2001 as the solution to financial emergency in the teeth of the dotcom bust when the fledgling company faced the loss of investor confidence. As investor pressure mounted, Google’s leaders abandoned their declared antipathy toward advertising. Instead they decided to boost ad revenue by using their exclusive access to user data logs (once known as “data exhaust”) in combination with their already substantial analytical capabilities and computational power, to generate predictions of user click-through rates, taken as a signal of an ad’s relevance.

>Operationally this meant that Google would both repurpose its growing cache of behavioural data, now put to work as a behavioural data surplus, and develop methods to aggressively seek new sources of this surplus.

>The company developed new methods of secret surplus capture that could uncover data that users intentionally opted to keep private, as well as to infer extensive personal information that users did not or would not provide. And this surplus would then be analysed for hidden meanings that could predict click-through behaviour. The surplus data became the basis for new predictions markets called targeted advertising.

Here was the origin of surveillance capitalism in an unprecedented and lucrative brew: behavioural surplus, data science, material infrastructure, computational power, algorithmic systems, and automated platforms. As click-through rates skyrocketed, advertising quickly became as important as search. Eventually it became the cornerstone of a new kind of commerce that depended upon online surveillance at scale.

The success of these new mechanisms only became visible when Google went public in 2004. That’s when it finally revealed that between 2001 and its 2004 IPO, revenues increased by 3,590%.

Surveillance capitalism is no more limited to advertising than mass production was limited to the fabrication of the Ford Model T. It quickly became the default model for capital accumulation in Silicon Valley, embraced by nearly every startup and app.

---

Further sources if you are interested:

VPRO Documentary Shoshana Zuboff on surveillance capitalism https://www.youtube.com/watch?v=hIXhnWUmMvw

The Intercept The Rise of Surveillance Capitalism https://www.youtube.com/watch?v=2s4Y-uZG5zk

High tech is watching you https://news.harvard.edu/gazette/story/2019/03/harvard-profe...

Re: Ask HN: Does targeted advertising work at all?

#106
post #34

Earlier quoted context omitted.

You clearly have no idea what you are talking about. Entire business are built on testing the effectiveness of online advertising, and providing that as a consulting device to brands. Do you think Nike would be happy if their vendor just said, “yeah everything drives positive lift”. Nobody is confusing correlation and causation. These are rigorous studies completed by literal data scientists and statisticians. They a…

And then there's the example of P&G, or Unilever or some other FCMG company that stopped advertising and saw no change in their business Bob Hoffman's take on advertising e.g. Advertising for Skeptics, is well worth reading - he's an ex-adman who thinks the ad industry and particularly the personalised ad industry is full of shit When you look at the online ad industry it's reasonably clear that the people who are ma…

Coca-Cola's advertising is so good, the sugar water literally tastes better if you know it's Coke, compared with a blind taste test.

Re: Ask HN: Does targeted advertising work at all?

#108
post #59
post #8

Earlier quoted context omitted.

> It’s hard to tell when someone has bought a product. How about, "I've literally got it in my last-day purchase history on eBay and they keep showing me the same thing and a bunch of related stuff to the item that hasn't even arrived yet, but that they've taken my money for"? ... and then eBay follows me around showing the same stuff on other sites. The thing I've purchased. From eBay. At least, they did, until I go…

My standard response to this: Prior purchases do have predictive power. They can very well be positively correlated to further purchases. I've often bought stuff that didn't work for some reason. It was broken, or it had something that was bothering me. Maybe the power adapter was annoyingly loud (looking at you, Kyocera). So I sent it back. But before I managed to go to the postal office a few days later I was alrea…

None of this sounds super frequent?

Re: Ask HN: Does targeted advertising work at all?

#109
post #34

Earlier quoted context omitted.

It is entirely possible that large corporations measure the effectiveness of their advertising by comparing it to sales, and are confusing correlation with causation on a massive scale . They spend more on advertising, and point to sales in the same time frame, as a cause, even if reality disagrees. There are zero incentives for anyone to figure this out and fix it within corporate structures. I'd be thankful it make…

You clearly have no idea what you are talking about. Entire business are built on testing the effectiveness of online advertising, and providing that as a consulting device to brands. Do you think Nike would be happy if their vendor just said, “yeah everything drives positive lift”. Nobody is confusing correlation and causation. These are rigorous studies completed by literal data scientists and statisticians. They a…

Honestly, after lifetime of working in companies on capitalist market, the theory that they/we make decisions based on rigorous studies strikes me as unplausible. So does the idea of effective neutral third party. I have seen how consultants work too.

Re: Ask HN: Does targeted advertising work at all?

#110

I recently started dabbling in Google ads for my side project. And I have learnt a lot about advertising through actually running campaigns: - Different countries have drastically different CPA (cost per action). You can acquire users easily in low-income countries but not US or Canada. So you have to set different CPA target for different countries. Currently Google ads doesn't seem to do adjustments automatically f…

> Running ad campaign for my app feels surprisingly similar to the experience of hyper-parameter tuning of ML models.

Maybe becouse most adclickers are bots?

Post reply on HN