Hacker News Reader

Comment by mike_d | Hacker News Reader

I used to work detecting ad fraud. Publishers would do bad things, call in, and try to get their account rep to get details.

Obviously I can't say "of the last 2500 ad clicks zero of them had any mouse movement over the ad before the click event" because then the publisher obviously just fixes their fraud software.

This isn't specific to Google or even advertising. Every company has figured out when dealing with abuse and fraud sharing the minimum amount of information is beneficial to the health of the ecosystem as a whole.

View on HN

In a case like that, sure. But they don't provide any information even when they want the publisher to make a change. Our Adsense account once got suspended because ads were appearing on pages that contained user-entered search keywords. Occasionally users would enter keywords that google considered 'naughty', and didn't want their ads appearing alongside. If they'd just told us that, we could have added a filter to not show ads with the list of keywords they had a problem with. Instead it was an infuriating, weeks-long process of pulling teeth to get clues as to what the problem might even be, and then making a list of every conceivably bad word we could find or imagine (admittedly that part was a bit fun) before we were finally able to get re-approved. And presumably we only got that much leeway because we were a reasonably large account.

Now your app has a bunch of offensive language compiled into it. Surely that's grounds for nuking your account.

Obviously the words were hashed before being added to the codebase :P

The subtility is they don’t actively _want_ these small publisher to do something.

It could be better if the issue is fixed, but Google’s skin in the game is small enough it doesn’t matter, and they already chose to get rid of the publisher as an efficient solution.

A valid explanation, but not one I believe applies here.

This ban is not only not explaining how it detected unwanted activity, it is not explaining what activity it detected.

"We detected you faking ad impressions, though we won't tell you how we (believe we) know" is very different to "We detected you (or your app) doing something wrong, stop doing it and you will be fine. We won't tell you what you did wrong".

"Our ML system registered a hit on your account, which is almost always associated with policy violations, but we don't know what the trigger was, or why that set of data about your account is almost always associated with policy violations... it just is."

We're on the cusp of independent artificial intelligence and it's incredibly lame.

"and we know the ML system always works, because those that get banned from it never get their account reinstated"

"and we know that it doesn't always work and don't care because despite our ML model being opaque and hard to introspect when we tested it against our database of known bad actors it performed better than both human review and our old heuristic based system."

There is an opposite side to this. If you have a human appeal process your accuracy drops to min(ml_model, humans) and I have bad news about which one of these is smaller.

I would believe this if Google wasn't notorious for providing the worst customer service that will avoid someone going to prison. Cable monopolies have a better reputation.

op didn't say he worked for Google

If the justice system started acting this way too where they don't tell you what crime you committed, what proof they have and how others can not commit such crimes, then I don't think that would mean "minimum amount of information is beneficial to the health of the society as a whole". This is basically letting get away with arbitrary rules without any accountability.

That is how the US "No Fly" list is managed, and is exactly what makes it unconstitutional extrajudicial punishment.

Is there anything in place to prevent bad actors from deploying fraudulent ad-click tools to takedown accounts?

(What you said makes sense to me, I just don’t know if/how ad networks could differentiate between the account-owner committing fraud and a malicious 3rd party.)

This is the issue I have with macs being the “go to dev machine.” People forget there are entire lines of devices with touch screens…

Not saying this is what was happening here, or that y’all didn’t think of it; just a general observation I’ve noticed over the years.

Perhaps it is just a symptom of most ad systems being fraud in general.

The expression "lie down with dogs, wake up with fleas" comes to mind...

What about false positives? How did you account for that?

You make your peace with the fact that you'll have a certain rate of false positives, where you'll intentionally lose also some legitimate business in order to keep most of the "ecosystem" cleaner. Perhaps an unsatifying answer, but that's it.

It's not a situation like putting someone in prison where "beyond all reasonable doubt" is the appropriate mark; you can refuse to do business based on mere suspicion that may be mistaken. There's a limit where extra investigation or appeals is too costly compared to just accepting the lost revenue, and for small-scale customers, that limit is quite low. With fraud detection, you have to balance the tradeoff between false positives and false negatives, but you'll certainly have both.

In Google’s case, this is not enough. They exert too much control over the online advertising industry that it’s simply unfair to ban anyone with no explanation and recourse. It should be illegal. It’s almost impossible to effectively monetize an app or website using ads without including various Google technologies and services, and that’s Google’s own doing; they’re the ones who purchased all of those companies and integrated their own products in a way that makes them inseparable.

People have this misconception that ad networks some how get joy from turning people off for no reason. Every ad shown is a penny in their pocket, even if its 100% fraud. When advertisers start asking for money back is when investigations are launched and accounts are terminated.

There are literally no false positives. It may be fraud, it may be the ad is too close to a back button and gets accidently clicked, it could be the ads don't display right. But at the end of the day, it is a revenue decision.

I also worked on adtech.

We viewed them as a cost of doing business. Some small accounts got nuked. :shrug: If we had to have humans investigate everything, and produce reports / interpretations that nuked customers found satisfactory, we wouldn't have been willing to service accounts under probably $40k/year.

And keep in mind the ecosystem is filthy with fraud, particularly on the low end. There very much are groups of organized thieves actively exploiting adtech.

And as @PeterisP says... look, we're not a court. We're a private business that is refusing to do further business with someone. Our right to do this was very clearly explained before the beginning of any relationship, and agreed to by that someone. If that someone doesn't like it, their recourse is to not do business with us.

> And as @PeterisP says... look, we're not a court. We're a private business that is refusing to do further business with someone. Our right to do this was very clearly explained before the beginning of any relationship, and agreed to by that someone. If that someone doesn't like it, their recourse is to not do business with us.

So you are basically justifying Google behavior. You are not a court, that's right. But every ban process should be easily and quickly prosecutable to settle the issue right in a court. Obviously real fraudsters will never appeal like that, because they know they would incur in even bigger problems.

> Obviously real fraudsters will never appeal like that

I now know you have no experience at all fighting online fraud. People, um, lie.

On a serious note, if you require a prosecutable ban process -- whatever that means, because prosecuting is something the government does -- where you'll end is my original point. Ad companies will refuse to do business with publishers that aren't above some minimum threshold. My guess is $40k a year. Because remember eg google or whoever keeps about 1/3 of that money, so a $1k/mo minimum to staff humans and deal with arguing feels ballpark reasonable.

Separately, I'm not justifying anything. I'm explaining the economics driving behavior. If you want to be mad at me for behavior I don't control or influence... :shrug:

Their recourse is to not do business, full stop, because Google has monopolized the ad space.

Reply on news.ycombinator.com