|
Activity Number:
|
396
|
|
Type:
|
Invited
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics and Marketing
|
| Abstract - #303048 |
|
Title:
|
Bayesian Statistics in Internet Advertising
|
|
Author(s):
|
Paul Dagum*+
|
|
Companies:
|
MarkMonitor
|
|
Address:
|
303 Second Street Suite 800N, San Francisco, CA, 94107,
|
|
Keywords:
|
Internet Advertising ; Bayesian Statistics ; Mixture Models
|
|
Abstract:
|
The internet allows advertisers to target the long-tail of consumer interests in a cost-effective model. The ability for search engines, ad networks, and ad exchanges to monetize the interaction between advertisers and an audience hinges on their ability to measure the audience response. For example, we have previously shown that performance based auctions used for monetization are incentive compatible only if inferences of audience response are unbiased. The disaggregated nature of the long-tail audience and the short half-life of consumer interests make it very difficult to reliably measure that response. At Business.Com we have successfully implemented Bayesian estimation of audience response using conditionally conjugate exponential family priors. We review the methodology and results applied to the Business.Com search site and the RHDi leads-network platform.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2009 program |
