4 ms·

What's the fraud scenario? Page owners presenting fraudulent visitor/click-through numbers to advertisers? In that scenario, it seems like advertisers would pi

What's the fraud scenario? Page owners presenting fraudulent visitor/click-through numbers to advertisers? In that scenario, it seems like advertisers would pick up on that pretty quickly when they realize the conversion rate on that supposed traffic is terrible and doesn't warrant the inflated price. In the case they're using an ad network, the network could ban the page owner from their network if they see this pattern from them. Since page owners are materially benefiting from the network, proof of identity should be (probably is? I don't work in the space) applied between the page owner and network to prevent repeat fraud via identity laundering. > you will not be able to pick as good of an ad compared to if you had more data. In theory I lean towards agreeing. I was arguing for tech-powered hyper-personalized ads back when I was studying advertising 2008-2012 (and did a bit of "stealth marketing" in that period where I built relationships with bloggers to share our product before the term "influencer" had hit the mainstream vocabulary). In practice, advertisers do not personalize ads. Facebook has become pretty good about selecting ads that map to my interests thanks to the reach of their spy network, but the ads themselves still aren't personalized at all (they take my interests into account, but not my spending history to realize that I don't have the budget for what they're trying to sell me) and my conversion to a sale because of them is still very, very low.

stevesearer 5y ago
We sell and host our own advertising which is content-based (office furniture ads on office design content) and think it is a good solution. Instead of selling space by impressions or clicks, we use length of time (monthly) and find it to be a good way to prevent ourselves from trying to game impressions with clickbait or clicks with fake users.
michaelt 5y ago
> What's the fraud scenario? 1. Page owner / Ad network / Ad space auction market middleman fakes clicks to get click revenue 2. Page owner's rival fakes clicks to devalue ad spots 3. Advertiser's agency fakes clicks to make numbers go up 4. Advertiser's rivals fake clicks, to waste advertiser's budget 5. Ad networks 'accidentally' classifying legitimate clicks as fraud, to reduce payouts to page owners.
- b3morales 5y ago
  How are these mitigated by pervasive end user tracking and surveillance?
- squiggleblaz 5y ago
  Sibling commenter identified an alternative as to charge for (presumably a proportion of) ads in a timeframe. That seems to satisfactorily mitigate almost everything: 1. Clicks are not charged for, so increasing clicks does not change the cost. 2. The only question available is how much you pay and how much you make. Page owner's rival could attempt to confuse you but there's not actually a difference between paying $1000 for 0.1% of ads shown in July, getting 100 clicks out of 10000 impressions and making $2000 vs paying $1000 for 0.1% of ads shown in July, getting 10000 clicks out of 19900 impressions and making $2000. 3. Shorter 2: Numbers are irrelevant unless they have a dollars/euro/yen/pound sign before them. 4. Clicks to not come out of the advertisers budget. 5. Surveillance advertising enables this fraud by attempting to distinguish between real and fake traffic. If you are paying for something that is robust to that, then the fraud is irrelevant. In particular, it doesn't matter if 0.1% of ads displayed to bots are yours and 0.1% of ads displayed to humans are yours if you've paid for 0.1% of ads to be displayed to be yours using a random rotation. The only thing like any of these hacks that seems relevant is that a very crafty adversary might be able to determine your traffic rate and identify the timeframe they need to request to particularly target a certain ad, making more of the 0.1% of displays get shown to a bot. But this is easily handled by using a secure random rotation rather than say a round robin rotation. In any case, I can independently verify the efficiency of my spend by getting a sim card that the ad network doesn't know about and repeatedly reloading the page for a month. If I receive only 0.01% of displays because some third party or the network is trying to defraud me I now have pure evidence that they aren't upholding their side of the deal. If they try to verify humanity before deciding what to do, then they can just say "no, we identified these requests as coming from a bot and graciously didn't display your ad to them". What can you do? The existence of surveillance is not in the interest of the advertiser who is concerned about fraud. (If you don't trust your network, any fraud on their part isn't going to get solved of course. But then having less data is an advantage. There will be fewer variables so it will be easier to catch, and the capital requirements will be simpler --- code and business partners vs code, business partners and user surveillance data --- so it will be easier for a trustworthy competitor to emerge.)

stevesearer 5y ago

We sell and host our own advertising which is content-based (office furniture ads on office design content) and think it is a good solution. Instead of selling space by impressions or clicks, we use length of time (monthly) and find it to be a good way to prevent ourselves from trying to game impressions with clickbait or clicks with fake users.

michaelt 5y ago

> What's the fraud scenario? 1. Page owner / Ad network / Ad space auction market middleman fakes clicks to get click revenue 2. Page owner's rival fakes clicks to devalue ad spots 3. Advertiser's agency fakes clicks to make numbers go up 4. Advertiser's rivals fake clicks, to waste advertiser's budget 5. Ad networks 'accidentally' classifying legitimate clicks as fraud, to reduce payouts to page owners.

b3morales 5y ago
How are these mitigated by pervasive end user tracking and surveillance?
squiggleblaz 5y ago
Sibling commenter identified an alternative as to charge for (presumably a proportion of) ads in a timeframe. That seems to satisfactorily mitigate almost everything: 1. Clicks are not charged for, so increasing clicks does not change the cost. 2. The only question available is how much you pay and how much you make. Page owner's rival could attempt to confuse you but there's not actually a difference between paying $1000 for 0.1% of ads shown in July, getting 100 clicks out of 10000 impressions and making $2000 vs paying $1000 for 0.1% of ads shown in July, getting 10000 clicks out of 19900 impressions and making $2000. 3. Shorter 2: Numbers are irrelevant unless they have a dollars/euro/yen/pound sign before them. 4. Clicks to not come out of the advertisers budget. 5. Surveillance advertising enables this fraud by attempting to distinguish between real and fake traffic. If you are paying for something that is robust to that, then the fraud is irrelevant. In particular, it doesn't matter if 0.1% of ads displayed to bots are yours and 0.1% of ads displayed to humans are yours if you've paid for 0.1% of ads to be displayed to be yours using a random rotation. The only thing like any of these hacks that seems relevant is that a very crafty adversary might be able to determine your traffic rate and identify the timeframe they need to request to particularly target a certain ad, making more of the 0.1% of displays get shown to a bot. But this is easily handled by using a secure random rotation rather than say a round robin rotation. In any case, I can independently verify the efficiency of my spend by getting a sim card that the ad network doesn't know about and repeatedly reloading the page for a month. If I receive only 0.01% of displays because some third party or the network is trying to defraud me I now have pure evidence that they aren't upholding their side of the deal. If they try to verify humanity before deciding what to do, then they can just say "no, we identified these requests as coming from a bot and graciously didn't display your ad to them". What can you do? The existence of surveillance is not in the interest of the advertiser who is concerned about fraud. (If you don't trust your network, any fraud on their part isn't going to get solved of course. But then having less data is an advantage. There will be fewer variables so it will be easier to catch, and the capital requirements will be simpler --- code and business partners vs code, business partners and user surveillance data --- so it will be easier for a trustworthy competitor to emerge.)