Hacker News Reader

Comment by spiderfarmer | Hacker News Reader

What I would like in a new kind of Ad network:

- all ads are responsive HTML5 ads

- all resources are loaded over HTTPS

- ads are based on your content, not on a profile of the user

- ads can be requested server side by sending the URL where the ad will be displayed through an API. The response contains the ad code, as well as the expiration date

- if you want, you can even download a package with all resources so ad blockers can't block you. (unless they target you specifically)

- if you don't get an ad in return, you can fill that space with your own fallback ads

- the ad network also does some kind of sentiment analysis so it doesn't show ads for Donald Trump on a page that's critical about him

- the ad network immediately severs ties with anyone who abuses the system

View on HN

We already built this - https://instinctive.io

It's the fastest ad network around. Simple, fast, static, all user-initiated, and we only focus on content (articles, videos, etc that you read on the same site you're already on). We also don't work with any 3rd parties so there's no security or malware risk. Happy to share any technical details.

There are good ad networks out there but in this strange industry, it's all about politics and connections, not tech or UX (which is why we're in this mess in the first place). This is the biggest battle we're fighting with everyone from advertisers and agencies to publishers.

You had me quite intrigued, at least until I saw native advertising being mentioned so prominently on the front page. Unbridled resource consumption and tracking are far from the only ethical concern around the modern ad industry. The breakdown of the barrier between journalistic content and paid content is another big one.

Our site is years old and needs a refresh but there's a lot of nuance in this.

Native is a complicated term that gets overused but the way we see it is a combination of look/feel + context + behavior. Most "native" ads today are just the look/feel but crappy irrelevant content that makes no sense on the site or isn't content at all but just takes you to some sales landing page like any other banner ad.

However, there can be good quality content (educational, informative, entertaining, and not just primarily selling) that happens to be sponsored by a brand. Native isn't about tricking the user but about being unobtrusive. The same way you would skip a news headline you're not interested in, you would skip a native ad placement as well (at least this is how it should work).

People read what they want to read and sometimes it's advertising. There's nothing wrong with that. What we do is make sure everything is properly labeled (to FTC standards) and user-initiated so that users know what it is and make their own choice - and I believe transparency and choice are the best things we can offer.

This is literally what HaloAds (HaloAds.com) aims to fix. We have been irritated with the state of online advertising for long enough now.

We are looking for people to try it out once it launches. Show your interest at https://HaloAds.com

My English websites are not very big but I'll check it out if possible.

Since you seem to specialise in this, FYI:

> Mixed Content: The page at 'https://haloads.com/' was loaded over HTTPS, but requested an insecure stylesheet 'http://fonts.googleapis.com/css?family=Lato:400,300,700,900'. This request has been blocked; the content must be served over HTTPS. https://haloads.com/assets/img/favicon.png Failed to load resource: the server responded with a status of 404 ()

For Google Fonts, it's best practice to leave out the protocol and link to '//fonts.googleapis.com/...'. It's unhelpful that their link generator still suggests 'http://fonts...'.

Or just download them and serve them from your own domain. If you enable HTTP/2 there is little to no advantage of using a third party for hosting fonts.

Wont they be cached from other peoples websites saving the download?

No. Well, maybe, if you're using one of the most popular fonts, but each combination of weights is a seperate CSS file, that's probably a unique combination and only cached for 24 hours. A reason not to use Google fonts is that it's just another tracking tool in their arsenal. For each request, cookies are sent to Google's servers, so they have enough reason to avoid cache hits.

nowadays I just use 'https://...' instead of '//...' because

* security * I will immediately know if a server doesn't support https, because the image won't load

Thanks for the spot. The issue has been resolved.

Signed up to be notified, but new companies in my experience have a lot of issues with fill and CPM. How are you planning to compete?

Thanks for signing up.

Our primary aim is to rid the internet of spam advertising and to turn online ads into an acceptable thing to have on your site as opposed to focusing on the highest bidder regardless of the quality of what your viewers end up seeing.

Those are all great ideas for a user-friendly ad network. I think that you could add "so the user makes fewer network requests and loads the page faster" as a reason for the option to download a resource pack.

But the key requirement for an ad network is that it makes money. I am optimistic that your system would produce user-friendly sites that would, given good content, eventually develop a positive reputation, earn more traffic, and make more money than their competitors. But that requires long-term vision and user-first business ethics, which most companies are not very good at. At any time, once that reputation and traffic pattern was established, switching to a more aggressive ad network would make the site or ad seller more money. Then it would dip back down as users reacted, but I am not convinced that a sufficient number of businesses would have the wherewithal to see it through.

I don't know how lucrative and / or effective it is, but a lot of high quality websites united in using The DECK, an ad network without a lot of bells and whistles. Since it's been around for quite some time, I guess they're doing well.

The ad network I'm proposing will never be as large as Google is right now. If it's a small network that only connects entrepeneurs with solid long-term visions, I am fine with that. I don't believe in unicorns anyway.

*http://decknetwork.net/

The Deck's revenue is about 1.5 million annually. They publish their advertisers list and rates. However, it might be less if they make deals.

My intuition is that The Deck is in slow decline due to fewer advertisers and declining interest in its member sites (Metafilter, Instapaper, a lot of web 2.0 design resources.) I still think it is an excellent example to follow, though.

As a user, this is what I want: The ad network provides the javascript, not the advertisers. No ad-hoc javascript provided by each advertiser. They deal with a standard API where they provide the content using pre-configured blocks of code. Like "we provide this standard tracking tool, this standard slideshow/animation system, etc".

This would be sweet and all except for the fact that ad networks typically don't care about a) providing proper stats and b) detecting display/click/affiliate fraud.

It's not really the advertisers providing the JS, it's that there are tons of adtech platforms and everyone just mixes and matches.

This is why you have 1 tag that loads dozens more because each agency is using a different system to serve an image or track an impression. There are some standards around formats but the actual tech delivery is terrible and this industry (for all the great tech that is actually built) just seems to have major problems at making things that work well. Probably because the consumer has never really been the major focus.

Either way, the amount of malicious code served by even the most respectable ad networks is ludicrous and unacceptable.

Or perhaps, all ads are just plain html... no scripts or videos.

Ideally, this would be a choice for the publisher. I spent a lot of time making my pages fast, so I would not request video ads.

> ads are based on your content, not on a profile of the user

Have we given up on relevance? I was always hopeful that one day in the future the sites I love could be supported by showing _only_ ads that interested their users. With our post-modern privacy fetish, though, I guess I'm becoming resigned to hygiene ads just so long as they're served over HTTPS.

Serving personalised ads in the EU means you will have to show a cookie warning. That for me is reason enough to focus on contextual ads. Not that I am showing cookie warnings right now, I'm counting the days till they revoke that stupid law.

I think it's worth at least considering that we should give up on relevance. As far as I can tell, nobody has figured out a good way to decide what ads are actually relevant to an individual. So personalized ads tend to come come in one of two flavors:

  1. A blast of ads for something you just bought.
  2. A blast of ads for something you just decided not to buy.

At best, ads like this deliver zero value to both advertisers and consumers. At worst, they creep people the heck out. Which means they hurt consumers by making them feel stalked, hurt advertisers because nobody wants their brand associated with creepy stalker behavior, and hurt publishers because it only accelerates the proliferation of ad blockers.

This doesn't mean relevance needs to go out the window. People were targeting their ads well before it became vogue to collect people's toenail clippings and feed them into collaborative filtering models. For example, if you're a high end diamond store then you shouldn't need a bunch of whiz kids from San Francisco to tell you that your ad would be better-placed on The New Yorker's site than on The Village Voice.

I dispute that serving ads relevant to the content of the site is "giving up on relevance." If anything it's far more likely to be relevant. And it doesn't require tracking the user.

If I'm reading a site about auto maintenance, for example, what's more relevant? Ads for auto parts suppliers, or ads for cookware because that's what I last bought on Amazon?

How much have you spent in the past year on advertising?

If the answer is none or not much, you may see the problem with this analysis, as it has as much bearing on what the market will provide as my opinion on what I would like to see in woman's grooming products.

I ran a small Internet Marketing agency for the last 15 years (15 people) and have been a publisher in my spare time, I sold it last year to focus on being a publisher.

This small agency easily spent 500k a year on Adwords alone. Not much maybe, but there are tens of thousands of agencies just like that one.

As a self-taught developer, I rebuilt my main website from the ground up and relaunched it last month. I have an A A A A A rating at webpagetest.org, but have the feeling that Adsense is ruining my hard work.

Right now I am not focussed on maximizing revenue, but when the time comes I hope I can do it in the most user-friendly way possible, without having to setup a sales department and doing everything myself.

When you were running the internet marketing agency and spending six figures on ads would your clients have been OK with the model you outlined above? Would you have steered their dollars there rather than to platforms that allow more rich media ads, more tracking/demographics, retargeting, and similar?

You've poked and prodded the parent commenter, but he has provided more discussion than you have. What is your issue with his suggestion? What are your qualifications to have issue with his suggestion?

Thanks, but I didn't mind. Seemed like a valid question.

They are genuine questions.

It seems to me that it's pointless to discuss product development without considering the customers of that product, so I'm trying to elicit discussion of that specific topic.

Oh yes, absolutely.

As a marketer my problem never was "I don't have enough data", but it always was "I don't have enough traffic". Just give me quality traffic and I'll hand you my money.

I don't need all bells and whistles as it makes it far too hard (and expensive) to set up a campaign for a small SMB.

As for the clients, they only want to know the results and maybe a list of sites where their ads were shown.

> ads can be requested server side by sending the URL where the ad will be displayed through an API. The response contains the ad code, as well as the expiration date

So, while adserver waiting for an ad from the RTB waterfall, then your server would wait too. It's not best practice I guess.

Besides that, many of your points already adopted by all ad networks. Many of the ad sector's problems directly related to advertisers. They want to measure everything, everything! There is a section in VAST definition called TrackingEvents. start, firstQuartile, midpoint, thirdQuartile, complete, pause, mute, click, skip ... goes on.

Agency make plans and send Vast codes to networks (for example sizmek vast code). We have many publishers well we have to measure that actions too. Stats has to be match. Then we are putting incoming vast to our adserver and it generates new one with wrapped vast. Then we are sending to publisher and of course they want to measure these actions too. They putting our vast code to their DFP and it generates new codes and ad goes to public.

Dfp Code > Our Adserver Code > Agency code

Then user sees the ad, for example video ad then tracking starts.

Impression: [Dfp, Our adserver, Sizmek]

Ad start: [Dfp, Our adserver, Sizmek]

First Quartile: [Dfp, Our adserver, Sizmek]

Midpoint: [Dfp, Our adserver, Sizmek]

Complete: [Dfp, Our adserver, Sizmek]

Click: [Dfp, Our adserver, Sizmek]

See, list goes on. One ad and many requests and counting. I am not talking about RTB waterfalls it is something else already.

We all have to measure this stats because many of the advertisers pay based on this measurements.

(you name it the currency)

%25 == 0

%50 == 0.003

%75 == 0.0055

%100 == 0.007

Requesting ads doesn't have to be on pageload.

Advertisers may want to track everything, but that doesn't mean you have to. If everyone is doing it, you can differentiate by saying NO.

As a marketer my problem never was "I don't have enough data", but it always was "I don't have enough traffic". Give me quality traffic and I'll hand you my money.

As a publisher THAT's what I want to offer. Quality traffic, without all the bullshit.

Maybe that gets me 10% of the revenue that all scummy 'best practices' get me, but I think it will work out great in the long run.

I only need an ad network to facilitate this, without having to set it up myself.

How can I provide you a good quality traffic without having those stats?

I am classifying publishers based on those stats and also requesting money from you again based on those stats.

You can not differentiate by saying not, because there is no such an option like NO!

I define good quality traffic by what they are doing on my website, not by who they are or what they are doing before they end up here.

Sounds like you are a direct response advertiser (as am I). For brand advertisers, ad engagement is a huge part of their measurement. Likewise, for direct response advertisers growing awareness through display and video, these ad engagement signals are critical for optimizing. On-site performance is just one measure, but when you start going down the whole view-through attribution rabbit hole, you'll want to look at engagement metrics like these to provide additional signals to steer you in the right direction.

Well, you would certainly not load ads in sync with page loads.

About measurements, you can measure almost anything (except for user identity). As long as you don't monetize it. You can monetize clicks or conversions.

> They want to measure everything, everything!

They can learn to live without spying on everything (this means no javascript), or they can learn to deal with more and more people installing an adblocker. Wanting something doesn't mean they get it.

If each site creates a CNAME (ex: ads.example.com or not-ads.example.com or more likely [a-z][a-z0-9]{15}.example.com) pointing at the ad broker, they could use something like letsencrypt to dynamically set up HTTPS.

Just like CDN77 does with their CDN.. Great example of using Letsencrypt to move the web forward.

Mind you, HTML5 allows for advertisers to do a lot of resource-hogging crap too, and not directly sandboxed like flash was / is.

As the ultimative target of your ads, I just want ads for products that doesn't suck and that actually cater to my needs and interests - I honestly don't care if they are HTML5 or responsive, though I am not sure I even have a flash player anymore so I suggest HTML5 as the most practical solution.

But the problem with ads isn't the technical issue, it is that they suck, horribly, because the products on offer are useless, offensive, or just plain not interesting. Facebook kept showing me offers for a shitty dating site that I am almost sure was a scam, for magic healing crystals, for utility services from a company 100s of miles away. At one point they even offered me to take the degree I already had, from the university I had already attended, despite the fact that facebook knew this.

Google kept giving me ads for the same shitty apartment search site after I moved into my new apartment, rather than something useful, such as curtains, furniture, etc.

Get an ad network that can actually deliver interesting ads I won't care how many requests it makes.

That's because Google and Facebook don't give a rats ass about who uses their network. The only thing they care about is to mine this advertising bubble for as long as possible.

Another thing that would be nice is getting rid of the layers of reselling that many of these ad networks do. Ever follow the links in a VAST Wrapper? Each one is the previous ad network not finding an ad, then requesting from another parter (while taking a cut of the CPI), then another partner, etc. Request times go through the roof and the final CPI is lousy.

About as realistic as any of the other things on the wishlist but what can you do.

I am convinced you don't need stuff like that. Only if you want to run a business that makes so much money that you'll start thinking about flying cars, balloons, satellites, robots, contact lenses, glasses and hundreds of other things, not related to your core mission.

A competitor can make do with much lower margins.

If you throw in "No Flash, Java, or JavaScript" that sounds like a killer idea for a YC submission.

Could be, but my passion lies elsewhere so if anyone would like to give it a go, be my guest.

> ads can be requested server side by sending the URL where the ad will be displayed through an API

THIS THIS THIS THIS THIS is my condition for viewing ads. Request and render the ads server-side, and I'll happily view them.

what about fraud?

As in? Both views and clicks can still be tracked so fraud will have to be dealt with in the same way as it is right now.

If you are saying that fraud is harder to detect without building user profiles, that might be so. But it's not impossible.

If you are saying that ads can be changed into clickbait I guess you'll have find a way to deal with that. I think an ad network should have an active relationship with both advertisers and publishers and not the anonymous, fully automated, crappy relationship Google has right now.

Furthermore I don't believe that fraud is being stopped now. I have paid google for a lot of meaningless clicks on their search network (people come to a page for a second or two). Not just a few- but the bulk of the clicks are like that.

And don't get me started on mobile.

That's why I'm staying away from ppc and am advertising with CPM. My spidey-sense tells me it hasn't been as thoroughly gamed yet.

If you want to be truly disruptive, you could build an ad network that doesn't require you to pay for bounces. An improved version of Google's quality score (read: one that works) would also improve the quality of the network as a whole.

CPM does that (cost per 1000 views as opposed to pay-per-click).

The quality score is quality from Google's point of view. If I put in things that help me qualify my leads (and get fewer people clicking and discovering that they don't want it), my quality score goes down. (The best way to do that, by the way, is to put in a price. Then the people who are just doing research tend to not click it. But you get penalized for that.)

They are throwing no intelligence at it. I have an entire site set up around study skills. My quality score for "study skills?" 1/10.

Fraud is hard to detect even with user profiling. Among other tricks, sophisticated fraudsters replay genuine user sessions collected on hacked computers to build fake profiles.

Whatever the solution, it will always be a game of cat and mouse.

Also commented elsewhere: If I'm correct (and this might be different in other countries), in the Netherlands TV ratings aren't measured by the TV channels themselves, but they use a trusted third party (Stichting KijkOnderzoek) that has nothing to do with selling ads.

You could setup a non-profit for tracking views and maybe even clicks, that only uses this information for statistical analysis. Revenue can be shared based on these statistics. The fraud issue would also have to be tackled by this entity.

There're 3rd party trackers for ad network such as double verify, comscore, moat. But the technology on this works horribly. For example, if an ad network want to track thru a vpaid, it wraps vpaid on top of ad content and that's it. ( Say the ad is hosted by the ad network itself.) But if you want third party tracking, you first wrap your ad with one layer, then you wrap third party tracking as second layer on top of that, which makes things worse.

Reply on news.ycombinator.com