Comments on: Winning In An Ad Exchange World http://www.adexchanger.com/data-driven-thinking/winning-in-an-ad-exchange-world/ Digital media optimization news and discussion website. Fri, 10 Feb 2012 23:33:42 +0000 hourly 1 http://wordpress.org/?v=3.3.1 By: Pascal Bensoussan http://www.adexchanger.com/data-driven-thinking/winning-in-an-ad-exchange-world/#comment-4412 Pascal Bensoussan Thu, 14 Jan 2010 03:00:18 +0000 http://www.adexchanger.com/?p=13838#comment-4412 Huayin, Thanks for the comment, which I agree with. Let me clarify. As you point you, the quality of both "look-alike" and "like-minded" comparisons is relatively low when done at the segment/attribute level (e.g. "car intender" or "high income"). However, in our experience, those comparison techniques become extremely powerful if they can run in real-time down to comparing individual users based on their event streams, without a-priori segmentation (e.g. ads they clicked on, products they added to their cart or purchased, articles they read). To your point, real-time bidders and dynamic ad servers have to be able to perform real-time, user-level similarity computations to be most effective. This is a requirement that we, at AK, are very familiar with. I am happy to take this discussion off-line if you would like more details. Best, Huayin,

Thanks for the comment, which I agree with.

Let me clarify.

As you point you, the quality of both "look-alike" and "like-minded" comparisons is relatively low when done at the segment/attribute level (e.g. "car intender" or "high income").

However, in our experience, those comparison techniques become extremely powerful if they can run in real-time down to comparing individual users based on their event streams, without a-priori segmentation (e.g. ads they clicked on, products they added to their cart or purchased, articles they read).

To your point, real-time bidders and dynamic ad servers have to be able to perform real-time, user-level similarity computations to be most effective. This is a requirement that we, at AK, are very familiar with.

I am happy to take this discussion off-line if you would like more details.

Best,

]]> By: Huayin Wang http://www.adexchanger.com/data-driven-thinking/winning-in-an-ad-exchange-world/#comment-4410 Huayin Wang Thu, 14 Jan 2010 02:15:26 +0000 http://www.adexchanger.com/?p=13838#comment-4410 Nicely done! Agree with most of what you said except a couple of things related to data analytics: I do not think finding look-a-like audience is the right framework; also, segmentation could be a bit outdated/insufficient analytics technique to handle the level of complexity unique to this. Nicely done!
Agree with most of what you said except a couple of things related to data analytics: I do not think finding look-a-like audience is the right framework; also, segmentation could be a bit outdated/insufficient analytics technique to handle the level of complexity unique to this.

]]> By: zach coelius http://www.adexchanger.com/data-driven-thinking/winning-in-an-ad-exchange-world/#comment-4405 zach coelius Wed, 13 Jan 2010 22:04:52 +0000 http://www.adexchanger.com/?p=13838#comment-4405 Well said Pascal, you are describing the system that some of us have spent the last year building. Pretty exciting stuff. Well said Pascal, you are describing the system that some of us have spent the last year building. Pretty exciting stuff.

]]>