3 ms·

Sorry for the throwaway, I work in ad tech. Other posters are saying they see hardly any fraud on Facebook. My employer is a more traditional ad tech company (t

by throwaway198374 11y ago

Sorry for the throwaway, I work in ad tech. Other posters are saying they see hardly any fraud on Facebook. My employer is a more traditional ad tech company (think exchange and direct buys) and we estimate 60-70% of our exchange inventory is fraudulent, and at least 15% of our impressions are delivered against fraudulent sites (of course we don't tell the customers). In reality the numbers are probably higher. The problem is that literally everyone except the end customer is making money off this. As soon as anyone talks about the need to do something about it the whole "let's not rock the boat", "what about our revenue targets", "well you can't tell Sales that" crowd comes out. I personally think there's a reckoning coming soon to traditional ad tech. There is so much fraud, such ludicrous and impossible revenue and impression targets, so much pushback against shitty and annoying advertising, and so little incentive to be honest with the customer. It can't last like this.

hayksaakian 11y ago
unless you're talking about really sophisticated fraud, i think advertisers and publishers both care about it. I'm on both sides of the coin. As an advertiser, click fraud is obvious when I see a huge disparity between google analytics and the ad network. You'd have to put a lot of effort into sending fake traffic that actually engages with the advertiser's landing page. As a publisher, I know that if I send a bunch of BS traffic to my advertisers they will blacklist me and starve my website. Regarding your estimates, I agree for really broad traffic. If your targeting options are very vague/loose you can buy ad traffic for very cheap, but you end up with a large number of click fraud. Most advertisers should want to be as narrow as possible with their targeting. The tricky thing is that when everyone is bidding on the golden goose egg demographics the price jumps quickly. In some cases, you're better off buying the 'general audience' traffic for pennies just because of sheer volume multiplied by chance.
ghshephard 11y ago
I would think that customers are reasonably sophisticated when it comes to stuff like this, particularly when they are spending real $$$. How difficult is it, for something like brand advertising, to do before/after poll's to discern how effective the dollar spend is in promoting a brand (with conversion to sales, it's even more trivial.) Done properly, the customer doesn't actually care if 99% of the exchange inventory and impressions are fraudulent, as long as they accomplish whatever objectives they set out to. Put more explicitly, if the objective for an advertiser (or, more likely, their agency) is to build awareness of a brand, they can measure a measure awareness among their target group, run a few million dollars of advertising, and then measure awareness again to see the difference. Done with enough customers, and enough programs, you start to get a very clear sense of where the dollar is best spent.
- ethanbond 11y ago
  Your hand wavy usage of "measure awareness" is the problem. How do you do such a thing? When you're advertising on multiple channels, even if you did have a solid measurement of awareness (you don't), you still wouldn't know what was responsible. This is the appeal behind things like Facebook ads. It seems you have a very clear sense of what return you're getting because there's such precise targeting and tracking. The issue is that neither of those have anything to do with the first portion of your comment: measuring awareness. You still need to measure awareness objectively and no one knows how to do it. The people best equipped and most ready to act like they do, however, is — you guessed it — Facebook and similar ad networks! Besides, the ad industry has been kept afloat with exactly the type of behavior that grandparent describes for decades if not centuries. What makes you think they're more sophisticated now?
  - ghshephard 11y ago
    Two ways to measure brand awareness - polling, and are people buying your product. And you run many, many programs on lots of channels in different demographics, and geographies - and see which program is most effective.
    - throwaway198374 11y ago
      The polling is often done by randomly inserting Vizu polls into the page somewhere and hoping someone fills them out. Of course hardly anyone does - who can be bothered? Still, they've managed to establish themselves as a credible third-party verifier (not saying they actually are credible - I know sometimes people running the campaign will fill these out to show that the campaign is working) Also, I'm not saying it's all fraudulent and doesn't work at all - of course if that were the case everyone would know. What I'm saying is that the brand is massively overpaying, because a large percentage of their delivery goes to fraudulent bot traffic. And it's pretty much in everyone's financial interest except the brand's to turn a blind eye. Also, it's important to know how campaigns are actually delivered: firstly, Sales sells something wildly optimistic like delivering 1m impressions a day to 45-50 y/o women in Florida. As time goes on and delivery can't be met we loosen the targeting parameters in order to "deliver in full" even though this delivery is no longer to the customers specifications. We might tell them we're doing that, or we might not and hope we get away with it. Then after the campaign we take whatever numbers we got and create a story out of it that makes us look good. If we're working with an agency we give it to them and they presumably spin it some more before presenting it to the customer.

hayksaakian 11y ago

unless you're talking about really sophisticated fraud, i think advertisers and publishers both care about it. I'm on both sides of the coin. As an advertiser, click fraud is obvious when I see a huge disparity between google analytics and the ad network. You'd have to put a lot of effort into sending fake traffic that actually engages with the advertiser's landing page. As a publisher, I know that if I send a bunch of BS traffic to my advertisers they will blacklist me and starve my website. Regarding your estimates, I agree for really broad traffic. If your targeting options are very vague/loose you can buy ad traffic for very cheap, but you end up with a large number of click fraud. Most advertisers should want to be as narrow as possible with their targeting. The tricky thing is that when everyone is bidding on the golden goose egg demographics the price jumps quickly. In some cases, you're better off buying the 'general audience' traffic for pennies just because of sheer volume multiplied by chance.

ghshephard 11y ago

I would think that customers are reasonably sophisticated when it comes to stuff like this, particularly when they are spending real $$$. How difficult is it, for something like brand advertising, to do before/after poll's to discern how effective the dollar spend is in promoting a brand (with conversion to sales, it's even more trivial.) Done properly, the customer doesn't actually care if 99% of the exchange inventory and impressions are fraudulent, as long as they accomplish whatever objectives they set out to. Put more explicitly, if the objective for an advertiser (or, more likely, their agency) is to build awareness of a brand, they can measure a measure awareness among their target group, run a few million dollars of advertising, and then measure awareness again to see the difference. Done with enough customers, and enough programs, you start to get a very clear sense of where the dollar is best spent.

ethanbond 11y ago
Your hand wavy usage of "measure awareness" is the problem. How do you do such a thing? When you're advertising on multiple channels, even if you did have a solid measurement of awareness (you don't), you still wouldn't know what was responsible. This is the appeal behind things like Facebook ads. It seems you have a very clear sense of what return you're getting because there's such precise targeting and tracking. The issue is that neither of those have anything to do with the first portion of your comment: measuring awareness. You still need to measure awareness objectively and no one knows how to do it. The people best equipped and most ready to act like they do, however, is — you guessed it — Facebook and similar ad networks! Besides, the ad industry has been kept afloat with exactly the type of behavior that grandparent describes for decades if not centuries. What makes you think they're more sophisticated now?
- ghshephard 11y ago
  Two ways to measure brand awareness - polling, and are people buying your product. And you run many, many programs on lots of channels in different demographics, and geographies - and see which program is most effective.
  - throwaway198374 11y ago
    The polling is often done by randomly inserting Vizu polls into the page somewhere and hoping someone fills them out. Of course hardly anyone does - who can be bothered? Still, they've managed to establish themselves as a credible third-party verifier (not saying they actually are credible - I know sometimes people running the campaign will fill these out to show that the campaign is working) Also, I'm not saying it's all fraudulent and doesn't work at all - of course if that were the case everyone would know. What I'm saying is that the brand is massively overpaying, because a large percentage of their delivery goes to fraudulent bot traffic. And it's pretty much in everyone's financial interest except the brand's to turn a blind eye. Also, it's important to know how campaigns are actually delivered: firstly, Sales sells something wildly optimistic like delivering 1m impressions a day to 45-50 y/o women in Florida. As time goes on and delivery can't be met we loosen the targeting parameters in order to "deliver in full" even though this delivery is no longer to the customers specifications. We might tell them we're doing that, or we might not and hope we get away with it. Then after the campaign we take whatever numbers we got and create a story out of it that makes us look good. If we're working with an agency we give it to them and they presumably spin it some more before presenting it to the customer.