5 ms·

The unintended consequence of paying per click is that it heavily incentives bot fraud, despite what Google might say about 'amazing performance' (Taboola is al

The unintended consequence of paying per click is that it heavily incentives bot fraud, despite what Google might say about 'amazing performance' (Taboola is also a CPC model for context). This is also another step towards unifying/supplanting the 'Google' ad network and DoubleClick.

Scoundreller 3y ago
If clicks can be botted, why can’t impressions? What this does kill is a small time publisher opening up their own website while waiting at the Apple Store and clicking some ads they know will pay a lot. Not that I’d ever do such a thing.
- crazygringo 3y ago
  Is it just a question of scale? You can easily have 100x or 1000x more impressions than clicks. So detecting fraudulent clicks is much harder than fraudulent impressions, in terms of a proportion of fraud, just because there's so much less of a baseline genuine signal. E.g. 20 fraudulent clicks out of 40 total clicks per day, who can tell? But 20,000 fraudulent impressions out of 40,000 total, now you can do a lot more pattern recognition to filter them out.
  - Scoundreller 3y ago
    If those 20 fraud clicks didn’t come from a source similar to 2000-20000 pageviews, they should be able to pick out something happening. Like. You’re right: it’s easier for a small-time fraudster to do cpc fraud, but they should still be able detect suspicious page views, whether there’s a click involved or not. At cpc large scale fraud, there’s a lot of data either way.
    - crazygringo 3y ago
      Not necessarily at all. The profile of people who click on an ad may be very different from those who just see it. The point is that small numbers just don't let you make confident conclusions period.
  - crowcroft 3y ago
    A good example is looking for 'unique devices' that are doing the action. It is fairly trivial to have 20 unique devices generating 20 clicks, quite a bit more challenging to have 20,000 unique devices generate 20,000 impressions.
    - Scoundreller 3y ago
      It’s a website: you can spoof a lot of details to show uniqueness. Proving you’re not headless is where it gets difficult.
pests 3y ago
I am really curious about the long term result of the Google and DoubleClick merger. I saw a commentor here a few weeks ago calling it a "reverse-takeover" of Google by DoubleClick. I was around and remember that era fondly so just wondering what was going on behind the scenes and what long-term effects it had.
- crowcroft 3y ago
  On one hand Google really wrapped up the whole market quite effectively (especially on the publisher side). On the other hand, I don't know anyone that considers the DoubleClick stack 'nice to use' (best of a bad bunch maybe). I'm probably naive, but I really believe digital media could be cleaned up significantly simplified if we started over. Programmatic media is massively overcomplicated mess compared to solutions from platforms like Meta.

Scoundreller 3y ago

If clicks can be botted, why can’t impressions? What this does kill is a small time publisher opening up their own website while waiting at the Apple Store and clicking some ads they know will pay a lot. Not that I’d ever do such a thing.

crazygringo 3y ago
Is it just a question of scale? You can easily have 100x or 1000x more impressions than clicks. So detecting fraudulent clicks is much harder than fraudulent impressions, in terms of a proportion of fraud, just because there's so much less of a baseline genuine signal. E.g. 20 fraudulent clicks out of 40 total clicks per day, who can tell? But 20,000 fraudulent impressions out of 40,000 total, now you can do a lot more pattern recognition to filter them out.
- Scoundreller 3y ago
  If those 20 fraud clicks didn’t come from a source similar to 2000-20000 pageviews, they should be able to pick out something happening. Like. You’re right: it’s easier for a small-time fraudster to do cpc fraud, but they should still be able detect suspicious page views, whether there’s a click involved or not. At cpc large scale fraud, there’s a lot of data either way.
  - crazygringo 3y ago
    Not necessarily at all. The profile of people who click on an ad may be very different from those who just see it. The point is that small numbers just don't let you make confident conclusions period.
- crowcroft 3y ago
  A good example is looking for 'unique devices' that are doing the action. It is fairly trivial to have 20 unique devices generating 20 clicks, quite a bit more challenging to have 20,000 unique devices generate 20,000 impressions.
  - Scoundreller 3y ago
    It’s a website: you can spoof a lot of details to show uniqueness. Proving you’re not headless is where it gets difficult.

pests 3y ago

I am really curious about the long term result of the Google and DoubleClick merger. I saw a commentor here a few weeks ago calling it a "reverse-takeover" of Google by DoubleClick. I was around and remember that era fondly so just wondering what was going on behind the scenes and what long-term effects it had.

crowcroft 3y ago
On one hand Google really wrapped up the whole market quite effectively (especially on the publisher side). On the other hand, I don't know anyone that considers the DoubleClick stack 'nice to use' (best of a bad bunch maybe). I'm probably naive, but I really believe digital media could be cleaned up significantly simplified if we started over. Programmatic media is massively overcomplicated mess compared to solutions from platforms like Meta.