4 ms·

This may not be a great solution to combat click fraud, but could google just limit the payout to 1 click per IP per website per day?

mistercow 15y ago
You would think, but I think they have always felt that a "tough on crime" mentality was necessary to inspire confidence in advertisers. This is a direct result of the imbalance of clout implicit to the long-tail/short-tail dichotomy of AdSense; Google packages a (relative) many dime-a-dozen content creators into a product (commodified ad space) which they sell to a (relative) few customers (advertisers). In addition, they can bet on content creators seeing the risk as low stakes (even though getting banned from a monopolistic ad network is anything but low stakes), while they can bank on advertisers seeing the risk as high stakes (which it is). A better solution is not to use cost-per-click in the first place, instead relying on metrics that are harder to game, like unique visitors and page views. These are arguably a better reflection of an advertising space's value even without taking click fraud into account. That's what Project Wonderful did, and it has worked out well for some of its users, but its model is also more long-tail/long-tail, which I think hurts its profitability.
- pbhjpbhj 15y ago
  >instead relying on metrics that are harder to game, like unique visitors and page views. // Surely page views are as easy to imitate as clicks (possibly easier). Unique visitors would be reasonably easy to game as well I'd imagine.
  - mistercow 15y ago
    The reason that page views and unique views are harder to game is that they worth far less individually, while not being correspondingly easier to produce. Say I'm a fraudster with a thousand IPs and a well-anonymized browser. If I'm paid per click, I can easily turn that into a thousand apparently legitimate clicks, which, depending on how I play my cards, could net me a few hundred dollars. By contrast, if I'm paid per unique visitor, I can turn that into a thousand apparently legitimate unique visitors, which will very optimistically pull in maybe $5 (judging by Project Wonderful). Now, neither of those schemes are optimal, but just about any improvement you suggest for the unique visitor model will reap ten times the rewards applied to the per-click model.
SpikeGronim 15y ago
Proxies ruin this idea. There are entire ISPs whose web traffic originates from a handful of IPs.

mistercow 15y ago

You would think, but I think they have always felt that a "tough on crime" mentality was necessary to inspire confidence in advertisers. This is a direct result of the imbalance of clout implicit to the long-tail/short-tail dichotomy of AdSense; Google packages a (relative) many dime-a-dozen content creators into a product (commodified ad space) which they sell to a (relative) few customers (advertisers). In addition, they can bet on content creators seeing the risk as low stakes (even though getting banned from a monopolistic ad network is anything but low stakes), while they can bank on advertisers seeing the risk as high stakes (which it is). A better solution is not to use cost-per-click in the first place, instead relying on metrics that are harder to game, like unique visitors and page views. These are arguably a better reflection of an advertising space's value even without taking click fraud into account. That's what Project Wonderful did, and it has worked out well for some of its users, but its model is also more long-tail/long-tail, which I think hurts its profitability.

pbhjpbhj 15y ago
>instead relying on metrics that are harder to game, like unique visitors and page views. // Surely page views are as easy to imitate as clicks (possibly easier). Unique visitors would be reasonably easy to game as well I'd imagine.
- mistercow 15y ago
  The reason that page views and unique views are harder to game is that they worth far less individually, while not being correspondingly easier to produce. Say I'm a fraudster with a thousand IPs and a well-anonymized browser. If I'm paid per click, I can easily turn that into a thousand apparently legitimate clicks, which, depending on how I play my cards, could net me a few hundred dollars. By contrast, if I'm paid per unique visitor, I can turn that into a thousand apparently legitimate unique visitors, which will very optimistically pull in maybe $5 (judging by Project Wonderful). Now, neither of those schemes are optimal, but just about any improvement you suggest for the unique visitor model will reap ten times the rewards applied to the per-click model.

SpikeGronim 15y ago

Proxies ruin this idea. There are entire ISPs whose web traffic originates from a handful of IPs.