Hacker News new | past | comments | ask | show | jobs | submit login

> It seems pretty simple to audit your 10,000 largest ad-spend domains and check to see if they're actually just ad fraud. You'd need maybe 300 person-hours to do so, at an overhead-in cost of, let's high-ball it, 100$/hr, which puts you at a cost of $30,000.

You're figuring 10k domains in 300 hours, or 1.8min per domain. It's way more expensive than this, and it's an adversarial system. If you have rules about how spammy a site can be and still show your ads many sites are going to try and get as close to the boundary as possible without going over. This means it takes substantial effort to figure out which side of it they're on.

Then, the spammier sites run many different domains with multiple accounts, which means they're not that likely to show up as top domains. Spammy sites also can "cloak" so they show up one way to a reviewer at the ad network and another way to a real site visitor.

Manual inspection is definitely a valuable tool, and is something that ad networks use, but that a process like you described would catch 99% of fraud is massively optimistic.

(Disclosure: I work at Google on ad stuff, but not this sort of ad stuff. Speaking for myself and not the company.)




>It's way more expensive than this, and it's an adversarial system

Sure, but I've overstated the hourly costs by an order of magnitude. You can get body shops to do this work very inexpensively.

Making a call as to whether or not they're over the line isn't difficult; you flag the clear bullshit calls and leave the borderline ones after making the contractual terms more conservative than your in practice review would accept.

>Then, the spammier sites run many different domains with multiple accounts, which means they're not that likely to show up as top domains

Sure, but we both know your distribution of revenue isn't difficult to parse, nor is auditing methodology difficult.

We have active fraudsters at the top of the iOS and Play stores. The issue isn't that you don't know how best to cast the net to grab everyone. It's that you don't cast it at all to grab anyone.

>but that a process like you described would catch 99% of fraud is massively optimistic.

Not really. Most fraudsters are terrible. Sure a few will slip through the cracks, but that's the case in every industry. That's not an excuse to do nothing. It's especially not an excuse when the dichotomy being pushed here is that the difficulty underlies a need for illegal anti-competitive behavior.


> We have active fraudsters at the top of the iOS and Play stores. The issue isn't that you don't know how best to cast the net to grab everyone. It's that you don't cast it at all to grab anyone.

What apps are you thinking of? My guess is that they're apps that are being really careful to stay just inside the bounds of the store policies.

> That's not an excuse to do nothing.

I'm confused why you think ad networks are doing nothing?

> the dichotomy being pushed here is that the difficulty underlies a need for illegal anti-competitive behavior.

I wasn't trying to defend that. I just saw your "it's pretty simple to..." and as someone close to the issue wanted to explain ways it's not that simple.


>I wasn't trying to defend that. I just saw your "it's pretty simple to..." and as someone close to the issue wanted to explain ways it's not that simple.

Oh, I thought you were the original poster I was replying to. Sorry about that. Regardless, it's not actually all that complicated at all. Catching all fraud is almost impossible. Catching 90-99% is dead simple. If you don't like my quick and dirty breakdown, make another - the problem's size needs to be 4 orders of magnitude larger before the cost implications become a concern.

That said, there's a very consistent pattern at play here: Organizations and groups that refuse to take action on the basis that they can't catch everything are consistently those that intentionally look the other way in respect of fraud, and digital advertising has a bad rep in this area for a reason: most networks ignore as much fraud as they can get away with.

In any event, locking out other networks doesn't change the fact that the excuse provided for exclusivity is about how many containers are around, not how many networks can bid on that inventory. The rationale for the illegal activity doesn't address the problem at all.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: