I was very impressed by this article. It is a new approach, contextual ads but dynamically sold and highly micro-targeted, just like current user-targeted ads, except based on web page content....
I was very impressed by this article. It is a new approach, contextual ads but dynamically sold and highly micro-targeted, just like current user-targeted ads, except based on web page content.
And the idea, specifically, that this approach could reverse the death spiral of Internet journalism, that's very appealing.
But then I went and visited the actual websites mentioned in the article ... you know, that ones that "completely abandoned using cookies to track users"?
And guess what? Cookies.
And my uMatrix and NoScript plugins both identified several google-based libraries — maybe trackers, maybe just pretty menus — but suspicious, either way.
Cutting out the middlemen and using contextual rather than microtargeted advertising helped increase revenues. Hopefully this approach will be used more widely in the future.
Cutting out the middlemen and using contextual rather than microtargeted advertising helped increase revenues. Hopefully this approach will be used more widely in the future.
While cutting out the middlemen is nice, I think there's something to say about sort of relevant but not hyper-relevant ads. The example of hungry people looking at food more willing to buy...
While cutting out the middlemen is nice, I think there's something to say about sort of relevant but not hyper-relevant ads.
The example of hungry people looking at food more willing to buy snickers I think holds true across a lot of what we do on the internet. If I see an ad about a tech service on a news article about tech, it doesn't stand out to me in the same way that an add for say, dishwashing detergent because they happen to know I buy this and it's been x amount of time since I've bought some.
How many times have we all told or heard a story about talking about a product and then an ad showing up for said product minutes or hours later when we were doing something else - joking that alexa or google must have overheard the conversation. More likely it's these hyper targeted ads that are tracking our behaviors, but I think the key here is that these hyper targeted ads have the ability to focus attention on the ad and make it very clear that it is a targeted ad instead of a more subconscious oh hey that looks like a good idea, let me click that link and buy some.
I was also very taken by the Snickers analogy. I work in recommendation/personalization and so I was very intrigued by the implications of it to recommendation systems—it helped me to better put...
I was also very taken by the Snickers analogy. I work in recommendation/personalization and so I was very intrigued by the implications of it to recommendation systems—it helped me to better put words to some thoughts I'd been having for a while. I'll re-jigger some tweets I wrote into a comment here:
Focusing only on user-based personalization can lead to a form of over-fitting where we always assume that past behavior can tell us what's most relevant to a user. But if the past is a good signal for user's latent taste, it can still completely miss their current needs.
There's ongoing work on learning a user's context real-time, but I think it's often ok to just let the user tell the application what they're up to. For Netflix, that might be as simple as selecting the genre, for adtech it could mean letting the user pick topics of interest. My Youtube viewing habits are a mix of semi-educational tech videos, live music performances, movie trailers, and occasional gaming stuff—and they're all mixed together in my homepage recommendations. It would be lovely to open Youtube and say "this is a music browsing session" and have the experience tailored to that.
If we think of personalization as a "push" model, where the user's wants are inferred and given to them without asking, I'm interested in more integration of search/"pull" model where the user can guide the personalization.
That's an interesting idea, sort of like an ad targeting uncanny valley where they're simultaneously really good and not good enough? I definitely wonder if that might be the case, since this...
That's an interesting idea, sort of like an ad targeting uncanny valley where they're simultaneously really good and not good enough? I definitely wonder if that might be the case, since this isn't the first time some internet site has found that contextual ads are just as good as targetted ones. As another example, DuckDuckGo only uses ads based on the context of your search term and they've said it works well for them.
Yeah I think there's a good middle point where they're unobtrusive and not too targeted in which people are likely to notice them rather than treat them as noise, but also not so targeted as to be...
sort of like an ad targeting uncanny valley where they're simultaneously really good and not good enough?
Yeah I think there's a good middle point where they're unobtrusive and not too targeted in which people are likely to notice them rather than treat them as noise, but also not so targeted as to be creepy.
But that's just my highly nonacademic theory. Frankly I haven't dived into the science of ads all that much.
So, I posted a comment here, saying "cool article, but the website in question doesn't seem to match the story". Several people asked for more detail. First issue is, the article itself never...
So, I posted a comment here, saying "cool article, but the website in question doesn't seem to match the story". Several people asked for more detail.
First issue is, the article itself never actually identifies which website(s) to which it refers. I originally assumed NPO website was https://www.npostart.nl/ , that's the first hit I got on my search. Others have pointed out https://over.npo.nl/ , plus the company is a large media group. They have many subsidiary sites, and exactly which sites are included, I'm just not sure.
Near the end of the article, they link to a Brave study, which confirms the article's basic theme, and lists many NPO subsidiary websites (but also no parent site?). The article also links directly to one subsidiary site, https://www.omroepmax.nl/ .
Now, all that said, you can just go visit any of these sites, and the first thing you get (ahem, in Dutch) is the familiar "we use cookies" pop-up.
Furthermore, upon closer reading, the article says they got rid of all "third party tracking/cookies" (that is to say, advertising cookies). I guess that means they still use cookies for their own website. It's still a huge step, but it definitely loses some of its impact.
Keep in mind, I do not speak Dutch. IDK what those "we use cookies" messages actually say. Might be worth getting those details.
Finally, between uMatrix and NoScript, on various subsidiary websites, I see things like
ajax.googleapis.com
gstatic.com
googletagmanager.com
fonts.googleapis.com
fonts.gstatic.com
Notably, right now, on what I think is the parent site, https://www.npostart.nl/ , I see absolutely no google-like, nor any other tracking-like, libraries. I could have sworn I saw a couple there the other day, but perhaps I mixed up my sites at the time.
None of these domains are related to ads or tracking though, are they? At most GTM can be used to bootstrap analytics, but it's not like you're going through DoubleClick or something.
None of these domains are related to ads or tracking though, are they? At most GTM can be used to bootstrap analytics, but it's not like you're going through DoubleClick or something.
I was very impressed by this article. It is a new approach, contextual ads but dynamically sold and highly micro-targeted, just like current user-targeted ads, except based on web page content.
And the idea, specifically, that this approach could reverse the death spiral of Internet journalism, that's very appealing.
But then I went and visited the actual websites mentioned in the article ... you know, that ones that "completely abandoned using cookies to track users"?
And guess what? Cookies.
And my uMatrix and NoScript plugins both identified several google-based libraries — maybe trackers, maybe just pretty menus — but suspicious, either way.
Cookies are definitely only part of the problem. We've still got analytics, fonts, fingerprinting and so on.
Do you mind taking a screenshot or any confirmation? Because this is basically saying this article is fake news and should be treated as such.
Chrome's 'cookies.txt' extension reports these cookies for https://over.npo.nl/:
uBlock Origin lists these domains on https://www.omroepbrabant.nl/
None of these are advertising related. Well, GTM can be, but it looks to be only used for analytics.
I'm not claiming they are (I'm not OP), I was just trying to provide more information.
Same here :)
I'm posting a new top-level comment, expanding on this.
Cutting out the middlemen and using contextual rather than microtargeted advertising helped increase revenues. Hopefully this approach will be used more widely in the future.
While cutting out the middlemen is nice, I think there's something to say about sort of relevant but not hyper-relevant ads.
The example of hungry people looking at food more willing to buy snickers I think holds true across a lot of what we do on the internet. If I see an ad about a tech service on a news article about tech, it doesn't stand out to me in the same way that an add for say, dishwashing detergent because they happen to know I buy this and it's been x amount of time since I've bought some.
How many times have we all told or heard a story about talking about a product and then an ad showing up for said product minutes or hours later when we were doing something else - joking that alexa or google must have overheard the conversation. More likely it's these hyper targeted ads that are tracking our behaviors, but I think the key here is that these hyper targeted ads have the ability to focus attention on the ad and make it very clear that it is a targeted ad instead of a more subconscious oh hey that looks like a good idea, let me click that link and buy some.
I was also very taken by the Snickers analogy. I work in recommendation/personalization and so I was very intrigued by the implications of it to recommendation systems—it helped me to better put words to some thoughts I'd been having for a while. I'll re-jigger some tweets I wrote into a comment here:
Focusing only on user-based personalization can lead to a form of over-fitting where we always assume that past behavior can tell us what's most relevant to a user. But if the past is a good signal for user's latent taste, it can still completely miss their current needs.
There's ongoing work on learning a user's context real-time, but I think it's often ok to just let the user tell the application what they're up to. For Netflix, that might be as simple as selecting the genre, for adtech it could mean letting the user pick topics of interest. My Youtube viewing habits are a mix of semi-educational tech videos, live music performances, movie trailers, and occasional gaming stuff—and they're all mixed together in my homepage recommendations. It would be lovely to open Youtube and say "this is a music browsing session" and have the experience tailored to that.
If we think of personalization as a "push" model, where the user's wants are inferred and given to them without asking, I'm interested in more integration of search/"pull" model where the user can guide the personalization.
That's an interesting idea, sort of like an ad targeting uncanny valley where they're simultaneously really good and not good enough? I definitely wonder if that might be the case, since this isn't the first time some internet site has found that contextual ads are just as good as targetted ones. As another example, DuckDuckGo only uses ads based on the context of your search term and they've said it works well for them.
Yeah I think there's a good middle point where they're unobtrusive and not too targeted in which people are likely to notice them rather than treat them as noise, but also not so targeted as to be creepy.
But that's just my highly nonacademic theory. Frankly I haven't dived into the science of ads all that much.
So, I posted a comment here, saying "cool article, but the website in question doesn't seem to match the story". Several people asked for more detail.
First issue is, the article itself never actually identifies which website(s) to which it refers. I originally assumed NPO website was https://www.npostart.nl/ , that's the first hit I got on my search. Others have pointed out https://over.npo.nl/ , plus the company is a large media group. They have many subsidiary sites, and exactly which sites are included, I'm just not sure.
Near the end of the article, they link to a Brave study, which confirms the article's basic theme, and lists many NPO subsidiary websites (but also no parent site?). The article also links directly to one subsidiary site, https://www.omroepmax.nl/ .
Now, all that said, you can just go visit any of these sites, and the first thing you get (ahem, in Dutch) is the familiar "we use cookies" pop-up.
Furthermore, upon closer reading, the article says they got rid of all "third party tracking/cookies" (that is to say, advertising cookies). I guess that means they still use cookies for their own website. It's still a huge step, but it definitely loses some of its impact.
Keep in mind, I do not speak Dutch. IDK what those "we use cookies" messages actually say. Might be worth getting those details.
Finally, between uMatrix and NoScript, on various subsidiary websites, I see things like
Notably, right now, on what I think is the parent site, https://www.npostart.nl/ , I see absolutely no google-like, nor any other tracking-like, libraries. I could have sworn I saw a couple there the other day, but perhaps I mixed up my sites at the time.
None of these domains are related to ads or tracking though, are they? At most GTM can be used to bootstrap analytics, but it's not like you're going through DoubleClick or something.