Supply side optimisation in online display advertising (Chapter 3)

Xuezhi Cao; Haokun Chen; Xiaofan Xuejian Xiaoling Wang; Weinan Zhang; Yong Yu; Lantao Yu; Kan Ren; Guanyu Tao; Weinan Zhang; Yong Yu; Jun Wang; Xiao Ma; Liqin Zhao; Guan Huang; Zhi Wang; Zelin Hu; Xiaoqiang Zhu; Kun Gai; Sepp Hochreiter; Jürgen Schmidhuber; Minjoon Seo; Aniruddha Kembhavi; Ali Farhadi; Hannaneh Hajishirzi; Eustache Diemert; Julien Meynet; Pierre Galland; Damien Lefortier; PwC; Interactive Advertising Bureau; Shuangfei Zhai; Keng-hao Chang; Ruofei Zhang; Zhongfei Mark Zhang; Kan Ren; Weinan Zhang; Keng-hao Chang; Yifei Rong; Yong Yu; Jun Wang; Shuai Yuan; Sahin Cem Geyik; Abhishek Saxena; Ali Dasdan; Junchi Yan; Chao Zhang; Hongyuan Zha; Min Gong; Changhua Sun; Jin Huang; Stephen Chu; Xiaokang Yang; Philip Kotler; Gary Armstrong; Vibhanshu Abhishek; Peter Fader; Kartik Hosanagar; Ritwik Sinha; Shiv Saini; N Anadhavelu; Ron Berman; Gabriel Hughes; Damien Allison; Jun Song; Jun Xiao; Fei Wu; Haishan Wu; Tong Zhang; Zhongfei Mark Zhang; Wenwu Zhu; Wendi Ji; Xiaofan Xuejian Xiaoling Wang; David A Wooff; Jillian M Anderson; Brian Dalessandro; Claudia Perlich; Ori Stitelman; Foster Provost; Eva Anderl; Ingo Becker; Florian V Wangenheim; Jan Hendrik Schumann; Wendi Ji; Xiaofan Xuejian Xiaoling Wang; Dell Zhang; Lizhen Xu; Jason A Duan; Andrew Whinston; Ya Zhang; Yi Wei; Jianbiao Ren; Xuhui Shao; Lexin Li; John Chandler-Pepelnjak; Bhaskar Bhattacharya; Desale Habtzghi; Anatol Rapoport; Albert M Chammah; Steven Kuhn; Flavio M Menezes; Paulo Klinger Monteiro; Weinan Zhang; Yuchen Wang; Kan Ren; Weinan Zhang; Jun Wang; Yong Yu; Yanru Qu; Han Cai; Kan Ren; Weinan Zhang; Yong Yu; Ying Wen; Jun Wang; Haifeng Zhang; Weinan Zhang; Jun Wang; Yifei Rong; Han Cai; Kan Ren; Weinan Zhang; Kleanthis Malialis; Jun Wang; Yong Yu; Defeng Guo; Chi-Chun Lin; Kun-Ta Chuang; Wush Chi-Hsuan Wu; Ming-Syan Chen; Kan Ren; Weinan Zhang; Yifei Rong; Haifeng Zhang; Yong Yu; Jun Wang; Claudia Perlich; Brian Dalessandro; Rod Hook; Ori Stitelman; Troy Raeder; Foster Provost; John Shawe-Taylor; Nello Cristianini; S Muthukrishnan; Weinan Zhang; Shuai Yuan; Jun Wang; Kan Ren; Jun Wang; Tianxiong Zhou; Jun Wang; Jian Xu; Thore Graepel; Joaquin Q Candela; Thomas Borchert; Ralf Herbrich; Xinran He; Junfeng Pan; Ou Jin; Tianbing Tao Xu; Bo Liu; Tianbing Tao Xu; Yanxin Shi; Antoine Atallah; Ralf Herbrich; Stuart Bowers; Richard J Oentaryo; Ee-Peng Lim; Jia-Wei Low; David Lo; Michael Finegold; H Brendan McMahan; Gary Holt; David Sculley; Michael Young; Dietmar Ebner; Julian Grady; Lan Nie; Todd Phillips; Eugene Davydov; Daniel Golovin; Matthew Richardson; Ewa Dominowska; Robert Ragno; Ying Cui; Ruofei Zhang; Wei Wentong Li; Jianchang Mao; Olivier Chapelle; Weinan Zhang; Shuai Yuan; Jun Wang; Suleyman Cetintas; Datong Chen; Luo Si; Weinan Zhang; Jun Wang; Yifei Rong; Jun Wang; Tianchi Zhu; Xiaofan Xuejian Xiaoling Wang; Steffen Rendle; Christoph Freudenthaler; Lars Schmidt-Thieme; Hairen Liao; Lingxiao Peng; Zhenchuan Liu; Xuehua Shen; Kareem Amin; Michael Kearns; Peter Key; Anton Schwaighofer; Google; Kuang-Chih Lee; Burkay Orten; Ali Dasdan; Wei Wentong Li; Ali Jalali; Ali Dasdan; Anh-Phuong Ta; Aditya Krishna Menon; Krishna-Prasad Chitrapura; Sachin Garg; Deepak Agarwal; Nagaraj Kota; Bee-Chung Chen; Pradheep Elango; Rahul Agrawal; Rajiv Khanna; Nagaraj Kota; Weinan Zhang; Tianming Du; Jun Wang; Benjamin Edelman; Michael Ostrovsky; Michael Schwarz; Ye Chen; Pavel Berkhin; Bo Anderson; Nikhil R Devanur; Xiang Li; Devin Guan; Jian Xu; Xuhui Shao; Jianjie Ma; Kuang-Chih Lee; Hang Qi; Quan Lu

Supply side optimisation in online display advertising (Chapter 3)

Xuezhi Cao, Haokun Chen, Xiaofan Xuejian Xiaoling Wang, Weinan Zhang, Yong Yu, Lantao Yu, Kan Ren, Guanyu Tao, Weinan Zhang, Yong Yu, Jun Wang, Xiao Ma, Liqin Zhao, Guan Huang, Zhi Wang, Zelin Hu, Xiaoqiang Zhu, Kun Gai, Sepp Hochreiter, Jürgen SchmidhuberMinjoon Seo, Aniruddha Kembhavi, Ali Farhadi, Hannaneh Hajishirzi, Eustache Diemert, Julien Meynet, Pierre Galland, Damien Lefortier, PwC, Interactive Advertising Bureau, Shuangfei Zhai, Keng-hao Chang, Ruofei Zhang, Zhongfei Mark Zhang, Kan Ren, Weinan Zhang, Keng-hao Chang, Yifei Rong, Yong Yu, Jun Wang, Shuai Yuan, Sahin Cem Geyik, Abhishek Saxena, Ali Dasdan, Junchi Yan, Chao Zhang, Hongyuan Zha, Min Gong, Changhua Sun, Jin Huang, Stephen Chu, Xiaokang Yang, Philip Kotler, Gary Armstrong, Vibhanshu Abhishek, Peter Fader, Kartik Hosanagar, Ritwik Sinha, Shiv Saini, N Anadhavelu, Ron Berman, Gabriel Hughes, Damien Allison, Jun Song, Jun Xiao, Fei Wu, Haishan Wu, Tong Zhang, Zhongfei Mark Zhang, Wenwu Zhu, Wendi Ji, Xiaofan Xuejian Xiaoling Wang, David A Wooff, Jillian M Anderson, Brian Dalessandro, Claudia Perlich, Ori Stitelman, Foster Provost, Eva Anderl, Ingo Becker, Florian V Wangenheim, Jan Hendrik Schumann, Wendi Ji, Xiaofan Xuejian Xiaoling Wang, Dell Zhang, Lizhen Xu, Jason A Duan, Andrew Whinston, Ya Zhang, Yi Wei, Jianbiao Ren, Xuhui Shao, Lexin Li, John Chandler-Pepelnjak, Bhaskar Bhattacharya, Desale Habtzghi, Anatol Rapoport, Albert M Chammah, Steven Kuhn, Flavio M Menezes, Paulo Klinger Monteiro, Weinan Zhang, Yuchen Wang, Kan Ren, Weinan Zhang, Jun Wang, Yong Yu, Yanru Qu, Han Cai, Kan Ren, Weinan Zhang, Yong Yu, Ying Wen, Jun Wang, Haifeng Zhang, Weinan Zhang, Jun Wang, Yifei Rong, Han Cai, Kan Ren, Weinan Zhang, Kleanthis Malialis, Jun Wang, Yong Yu, Defeng Guo, Chi-Chun Lin, Kun-Ta Chuang, Wush Chi-Hsuan Wu, Ming-Syan Chen, Kan Ren, Weinan Zhang, Yifei Rong, Haifeng Zhang, Yong Yu, Jun Wang, Claudia Perlich, Brian Dalessandro, Rod Hook, Ori Stitelman, Troy Raeder, Foster Provost, John Shawe-Taylor, Nello Cristianini, S Muthukrishnan, Weinan Zhang, Shuai Yuan, Jun Wang, Kan Ren, Jun Wang, Tianxiong Zhou, Jun Wang, Jian Xu, Thore Graepel, Joaquin Q Candela, Thomas Borchert, Ralf Herbrich, Xinran He, Junfeng Pan, Ou Jin, Tianbing Tao Xu, Bo Liu, Tianbing Tao Xu, Yanxin Shi, Antoine Atallah, Ralf Herbrich, Stuart Bowers, Richard J Oentaryo, Ee-Peng Lim, Jia-Wei Low, David Lo, Michael Finegold, H Brendan McMahan, Gary Holt, David Sculley, Michael Young, Dietmar Ebner, Julian Grady, Lan Nie, Todd Phillips, Eugene Davydov, Daniel Golovin, Matthew Richardson, Ewa Dominowska, Robert Ragno, Ying Cui, Ruofei Zhang, Wei Wentong Li, Jianchang Mao, Olivier Chapelle, Weinan Zhang, Shuai Yuan, Jun Wang, Suleyman Cetintas, Datong Chen, Luo Si, Weinan Zhang, Jun Wang, Yifei Rong, Jun Wang, Tianchi Zhu, Xiaofan Xuejian Xiaoling Wang, Steffen Rendle, Christoph Freudenthaler, Lars Schmidt-Thieme, Hairen Liao, Lingxiao Peng, Zhenchuan Liu, Xuehua Shen, Kareem Amin, Michael Kearns, Peter Key, Anton Schwaighofer, Google, Kuang-Chih Lee, Burkay Orten, Ali Dasdan, Wei Wentong Li, Ali Jalali, Ali Dasdan, Anh-Phuong Ta, Aditya Krishna Menon, Krishna-Prasad Chitrapura, Sachin Garg, Deepak Agarwal, Nagaraj Kota, Bee-Chung Chen, Pradheep Elango, Rahul Agrawal, Rajiv Khanna, Nagaraj Kota, Weinan Zhang, Tianming Du, Jun Wang, Benjamin Edelman, Michael Ostrovsky, Michael Schwarz, Ye Chen, Pavel Berkhin, Bo Anderson, Nikhil R Devanur, Xiang Li, Devin Guan, Jian Xu, Xuhui Shao, Jianjie Ma, Kuang-Chih Lee, Hang Qi, Quan Lu

Research output: Contribution to journal › Article › peer-review

Abstract

On the Internet there are publishers (the supply side) who provide free contents (e.g., news) and services (e.g., email) to attract users. Publishers get paid by selling ad displaying opportunities (i.e., impressions) to advertisers. Advertisers then sell products to users who are converted by ads. Better supply side revenue allows more free content and services to be created, thus, benefiting the entire online advertising ecosystem. This thesis addresses several optimisation problems for the supply side. When a publisher creates an ad-supported website, he needs to decide the percentage of ads first. The thesis reports a large-scale empirical study of Internet ad density over past seven years, then presents a model that includes many factors, especially the competition among similar publishers, and gives an optimal dynamic ad density that generates the maximum revenue over time. This study also unveils the tragedy of the commons in online advertising where users’ attention has been overgrazed which results in a global sub-optimum. After deciding the ad density, the publisher retrieves ads from various sources, including contracts, ad networks, and ad exchanges. This forms an explorationexploitation problem when ad sources are typically unknown before trail. This problem is modelled using Partially Observable Markov Decision Process (POMDP), and the exploration efficiency is increased by utilising the correlation of ads. The proposed method reports 23.4% better than the best performing baseline in the real-world data based experiments. Since some ad networks allow (or expect) an input of keywords, the thesis also presents an adaptive keyword extraction system using BM25F algorithm and the multi-armed bandits model. This system has been tested by a domain service provider in crowdsourcing based experiments. If the publisher selects a Real-Time Bidding (RTB) ad source, he can use reserve price to manipulate auctions for better payoff. This thesis proposes a simplified game model that considers the competition between seller and buyer to be one-shot instead of repeated and gives heuristics that can be easily implemented. The model has been evaluated in a production environment and reported 12.3% average increase of revenue. The documentation of a prototype system for reserve price optimisation is also presented in the appendix of the thesis.

Original language	American English
Pages (from-to)	1-6
Number of pages	6
Journal	Kdd
Volume	56
Issue number	3
State	Published - 2016

Cite this

Cao, X, Chen, H, Wang, XXX, Zhang, W, Yu, Y, Yu, L, Ren, K, Tao, G, Zhang, W, Yu, Y, Wang, J, Ma, X, Zhao, L, Huang, G, Wang, Z, Hu, Z, Zhu, X, Gai, K, Hochreiter, S, Schmidhuber, J, Seo, M, Kembhavi, A, Farhadi, A, Hajishirzi, H, Diemert, E, Meynet, J, Galland, P, Lefortier, D, PwC, Bureau, IA, Zhai, S, Chang, K, Zhang, R, Zhang, ZM, Ren, K, Zhang, W, Chang, K, Rong, Y, Yu, Y, Wang, J, Yuan, S, Geyik, SC, Saxena, A, Dasdan, A, Yan, J, Zhang, C, Zha, H, Gong, M, Sun, C, Huang, J, Chu, S, Yang, X, Kotler, P, Armstrong, G, Abhishek, V, Fader, P, Hosanagar, K, Sinha, R, Saini, S, Anadhavelu, N, Berman, R, Hughes, G, Allison, D, Song, J, Xiao, J, Wu, F, Wu, H, Zhang, T, Zhang, ZM, Zhu, W, Ji, W, Wang, XXX, Wooff, DA, Anderson, JM, Dalessandro, B, Perlich, C, Stitelman, O, Provost, F, Anderl, E, Becker, I, Wangenheim, FV, Schumann, JH, Ji, W, Wang, XXX, Zhang, D, Xu, L, Duan, JA, Whinston, A, Zhang, Y, Wei, Y, Ren, J, Shao, X, Li, L, Chandler-Pepelnjak, J, Bhattacharya, B, Habtzghi, D, Rapoport, A, Chammah, AM, Kuhn, S, Menezes, FM, Monteiro, PK, Zhang, W, Wang, Y, Ren, K, Zhang, W, Wang, J, Yu, Y, Qu, Y, Cai, H, Ren, K, Zhang, W, Yu, Y, Wen, Y, Wang, J, Zhang, H, Zhang, W, Wang, J, Rong, Y, Cai, H, Ren, K, Zhang, W, Malialis, K, Wang, J, Yu, Y, Guo, D, Lin, C-C, Chuang, K-T, Wu, WC-H, Chen, M-S, Ren, K, Zhang, W, Rong, Y, Zhang, H, Yu, Y, Wang, J, Perlich, C, Dalessandro, B, Hook, R, Stitelman, O, Raeder, T, Provost, F, Shawe-Taylor, J, Cristianini, N, Muthukrishnan, S, Zhang, W, Yuan, S, Wang, J, Ren, K, Wang, J, Zhou, T, Wang, J, Xu, J, Graepel, T, Candela, JQ, Borchert, T, Herbrich, R, He, X, Pan, J, Jin, O, Xu, TT, Liu, B, Xu, TT, Shi, Y, Atallah, A, Herbrich, R, Bowers, S, Oentaryo, RJ, Lim, E-P, Low, J-W, Lo, D, Finegold, M, McMahan, HB, Holt, G, Sculley, D, Young, M, Ebner, D, Grady, J, Nie, L, Phillips, T, Davydov, E, Golovin, D, Richardson, M, Dominowska, E, Ragno, R, Cui, Y, Zhang, R, Li, WW, Mao, J, Chapelle, O, Zhang, W, Yuan, S, Wang, J, Cetintas, S, Chen, D, Si, L, Zhang, W, Wang, J, Rong, Y, Wang, J, Zhu, T, Wang, XXX, Rendle, S, Freudenthaler, C, Schmidt-Thieme, L, Liao, H, Peng, L, Liu, Z, Shen, X, Amin, K, Kearns, M, Key, P, Schwaighofer, A, Google, Lee, K-C, Orten, B, Dasdan, A, Li, WW, Jalali, A, Dasdan, A, Ta, A-P, Menon, AK, Chitrapura, K-P, Garg, S, Agarwal, D, Kota, N, Chen, B-C, Elango, P, Agrawal, R, Khanna, R, Kota, N, Zhang, W, Du, T, Wang, J, Edelman, B, Ostrovsky, M, Schwarz, M, Chen, Y, Berkhin, P, Anderson, B, Devanur, NR, Li, X, Guan, D, Xu, J, Shao, X, Ma, J, Lee, K-C, Qi, H & Lu, Q 2016, 'Supply side optimisation in online display advertising (Chapter 3)', Kdd, vol. 56, no. 3, pp. 1-6.

@article{adebda8a25cc41ab8b0e7ff1531e75ef,

title = "Supply side optimisation in online display advertising (Chapter 3)",

abstract = "On the Internet there are publishers (the supply side) who provide free contents (e.g., news) and services (e.g., email) to attract users. Publishers get paid by selling ad displaying opportunities (i.e., impressions) to advertisers. Advertisers then sell products to users who are converted by ads. Better supply side revenue allows more free content and services to be created, thus, benefiting the entire online advertising ecosystem. This thesis addresses several optimisation problems for the supply side. When a publisher creates an ad-supported website, he needs to decide the percentage of ads first. The thesis reports a large-scale empirical study of Internet ad density over past seven years, then presents a model that includes many factors, especially the competition among similar publishers, and gives an optimal dynamic ad density that generates the maximum revenue over time. This study also unveils the tragedy of the commons in online advertising where users{\textquoteright} attention has been overgrazed which results in a global sub-optimum. After deciding the ad density, the publisher retrieves ads from various sources, including contracts, ad networks, and ad exchanges. This forms an explorationexploitation problem when ad sources are typically unknown before trail. This problem is modelled using Partially Observable Markov Decision Process (POMDP), and the exploration efficiency is increased by utilising the correlation of ads. The proposed method reports 23.4\% better than the best performing baseline in the real-world data based experiments. Since some ad networks allow (or expect) an input of keywords, the thesis also presents an adaptive keyword extraction system using BM25F algorithm and the multi-armed bandits model. This system has been tested by a domain service provider in crowdsourcing based experiments. If the publisher selects a Real-Time Bidding (RTB) ad source, he can use reserve price to manipulate auctions for better payoff. This thesis proposes a simplified game model that considers the competition between seller and buyer to be one-shot instead of repeated and gives heuristics that can be easily implemented. The model has been evaluated in a production environment and reported 12.3\% average increase of revenue. The documentation of a prototype system for reserve price optimisation is also presented in the appendix of the thesis.",

author = "Xuezhi Cao and Haokun Chen and Wang, \{Xiaofan Xuejian Xiaoling\} and Weinan Zhang and Yong Yu and Lantao Yu and Kan Ren and Guanyu Tao and Weinan Zhang and Yong Yu and Jun Wang and Xiao Ma and Liqin Zhao and Guan Huang and Zhi Wang and Zelin Hu and Xiaoqiang Zhu and Kun Gai and Sepp Hochreiter and J{\"u}rgen Schmidhuber and Minjoon Seo and Aniruddha Kembhavi and Ali Farhadi and Hannaneh Hajishirzi and Eustache Diemert and Julien Meynet and Pierre Galland and Damien Lefortier and PwC and Bureau, \{Interactive Advertising\} and Shuangfei Zhai and Keng-hao Chang and Ruofei Zhang and Zhang, \{Zhongfei Mark\} and Kan Ren and Weinan Zhang and Keng-hao Chang and Yifei Rong and Yong Yu and Jun Wang and Shuai Yuan and Geyik, \{Sahin Cem\} and Abhishek Saxena and Ali Dasdan and Junchi Yan and Chao Zhang and Hongyuan Zha and Min Gong and Changhua Sun and Jin Huang and Stephen Chu and Xiaokang Yang and Philip Kotler and Gary Armstrong and Vibhanshu Abhishek and Peter Fader and Kartik Hosanagar and Ritwik Sinha and Shiv Saini and N Anadhavelu and Ron Berman and Gabriel Hughes and Damien Allison and Jun Song and Jun Xiao and Fei Wu and Haishan Wu and Tong Zhang and Zhang, \{Zhongfei Mark\} and Wenwu Zhu and Wendi Ji and Wang, \{Xiaofan Xuejian Xiaoling\} and Wooff, \{David A\} and Anderson, \{Jillian M\} and Brian Dalessandro and Claudia Perlich and Ori Stitelman and Foster Provost and Eva Anderl and Ingo Becker and Wangenheim, \{Florian V\} and Schumann, \{Jan Hendrik\} and Wendi Ji and Wang, \{Xiaofan Xuejian Xiaoling\} and Dell Zhang and Lizhen Xu and Duan, \{Jason A\} and Andrew Whinston and Ya Zhang and Yi Wei and Jianbiao Ren and Xuhui Shao and Lexin Li and John Chandler-Pepelnjak and Bhaskar Bhattacharya and Desale Habtzghi and Anatol Rapoport and Chammah, \{Albert M\} and Steven Kuhn and Menezes, \{Flavio M\} and Monteiro, \{Paulo Klinger\} and Weinan Zhang and Yuchen Wang and Kan Ren and Weinan Zhang and Jun Wang and Yong Yu and Yanru Qu and Han Cai and Kan Ren and Weinan Zhang and Yong Yu and Ying Wen and Jun Wang and Haifeng Zhang and Weinan Zhang and Jun Wang and Yifei Rong and Han Cai and Kan Ren and Weinan Zhang and Kleanthis Malialis and Jun Wang and Yong Yu and Defeng Guo and Chi-Chun Lin and Kun-Ta Chuang and Wu, \{Wush Chi-Hsuan\} and Ming-Syan Chen and Kan Ren and Weinan Zhang and Yifei Rong and Haifeng Zhang and Yong Yu and Jun Wang and Claudia Perlich and Brian Dalessandro and Rod Hook and Ori Stitelman and Troy Raeder and Foster Provost and John Shawe-Taylor and Nello Cristianini and S Muthukrishnan and Weinan Zhang and Shuai Yuan and Jun Wang and Kan Ren and Jun Wang and Tianxiong Zhou and Jun Wang and Jian Xu and Thore Graepel and Candela, \{Joaquin Q\} and Thomas Borchert and Ralf Herbrich and Xinran He and Junfeng Pan and Ou Jin and Xu, \{Tianbing Tao\} and Bo Liu and Xu, \{Tianbing Tao\} and Yanxin Shi and Antoine Atallah and Ralf Herbrich and Stuart Bowers and Oentaryo, \{Richard J\} and Ee-Peng Lim and Jia-Wei Low and David Lo and Michael Finegold and McMahan, \{H Brendan\} and Gary Holt and David Sculley and Michael Young and Dietmar Ebner and Julian Grady and Lan Nie and Todd Phillips and Eugene Davydov and Daniel Golovin and Matthew Richardson and Ewa Dominowska and Robert Ragno and Ying Cui and Ruofei Zhang and Li, \{Wei Wentong\} and Jianchang Mao and Olivier Chapelle and Weinan Zhang and Shuai Yuan and Jun Wang and Suleyman Cetintas and Datong Chen and Luo Si and Weinan Zhang and Jun Wang and Yifei Rong and Jun Wang and Tianchi Zhu and Wang, \{Xiaofan Xuejian Xiaoling\} and Steffen Rendle and Christoph Freudenthaler and Lars Schmidt-Thieme and Hairen Liao and Lingxiao Peng and Zhenchuan Liu and Xuehua Shen and Kareem Amin and Michael Kearns and Peter Key and Anton Schwaighofer and Google and Kuang-Chih Lee and Burkay Orten and Ali Dasdan and Li, \{Wei Wentong\} and Ali Jalali and Ali Dasdan and Anh-Phuong Ta and Menon, \{Aditya Krishna\} and Krishna-Prasad Chitrapura and Sachin Garg and Deepak Agarwal and Nagaraj Kota and Bee-Chung Chen and Pradheep Elango and Rahul Agrawal and Rajiv Khanna and Nagaraj Kota and Weinan Zhang and Tianming Du and Jun Wang and Benjamin Edelman and Michael Ostrovsky and Michael Schwarz and Ye Chen and Pavel Berkhin and Bo Anderson and Devanur, \{Nikhil R\} and Xiang Li and Devin Guan and Jian Xu and Xuhui Shao and Jianjie Ma and Kuang-Chih Lee and Hang Qi and Quan Lu",

year = "2016",

language = "American English",

volume = "56",

pages = "1--6",

journal = "Kdd",

number = "3",

}

TY - JOUR

T1 - Supply side optimisation in online display advertising (Chapter 3)

AU - Cao, Xuezhi

AU - Chen, Haokun

AU - Wang, Xiaofan Xuejian Xiaoling

AU - Zhang, Weinan

AU - Yu, Yong

AU - Yu, Lantao

AU - Ren, Kan

AU - Tao, Guanyu

AU - Zhang, Weinan

AU - Yu, Yong

AU - Wang, Jun

AU - Ma, Xiao

AU - Zhao, Liqin

AU - Huang, Guan

AU - Wang, Zhi

AU - Hu, Zelin

AU - Zhu, Xiaoqiang

AU - Gai, Kun

AU - Hochreiter, Sepp

AU - Schmidhuber, Jürgen

AU - Seo, Minjoon

AU - Kembhavi, Aniruddha

AU - Farhadi, Ali

AU - Hajishirzi, Hannaneh

AU - Diemert, Eustache

AU - Meynet, Julien

AU - Galland, Pierre

AU - Lefortier, Damien

AU - PwC,

AU - Bureau, Interactive Advertising

AU - Zhai, Shuangfei

AU - Chang, Keng-hao

AU - Zhang, Ruofei

AU - Zhang, Zhongfei Mark

AU - Ren, Kan

AU - Zhang, Weinan

AU - Chang, Keng-hao

AU - Rong, Yifei

AU - Yu, Yong

AU - Wang, Jun

AU - Yuan, Shuai

AU - Geyik, Sahin Cem

AU - Saxena, Abhishek

AU - Dasdan, Ali

AU - Yan, Junchi

AU - Zhang, Chao

AU - Zha, Hongyuan

AU - Gong, Min

AU - Sun, Changhua

AU - Huang, Jin

AU - Chu, Stephen

AU - Yang, Xiaokang

AU - Kotler, Philip

AU - Armstrong, Gary

AU - Abhishek, Vibhanshu

AU - Fader, Peter

AU - Hosanagar, Kartik

AU - Sinha, Ritwik

AU - Saini, Shiv

AU - Anadhavelu, N

AU - Berman, Ron

AU - Hughes, Gabriel

AU - Allison, Damien

AU - Song, Jun

AU - Xiao, Jun

AU - Wu, Fei

AU - Wu, Haishan

AU - Zhang, Tong

AU - Zhang, Zhongfei Mark

AU - Zhu, Wenwu

AU - Ji, Wendi

AU - Wang, Xiaofan Xuejian Xiaoling

AU - Wooff, David A

AU - Anderson, Jillian M

AU - Dalessandro, Brian

AU - Perlich, Claudia

AU - Stitelman, Ori

AU - Provost, Foster

AU - Anderl, Eva

AU - Becker, Ingo

AU - Wangenheim, Florian V

AU - Schumann, Jan Hendrik

AU - Ji, Wendi

AU - Wang, Xiaofan Xuejian Xiaoling

AU - Zhang, Dell

AU - Xu, Lizhen

AU - Duan, Jason A

AU - Whinston, Andrew

AU - Zhang, Ya

AU - Wei, Yi

AU - Ren, Jianbiao

AU - Shao, Xuhui

AU - Li, Lexin

AU - Chandler-Pepelnjak, John

AU - Bhattacharya, Bhaskar

AU - Habtzghi, Desale

AU - Rapoport, Anatol

AU - Chammah, Albert M

AU - Kuhn, Steven

AU - Menezes, Flavio M

AU - Monteiro, Paulo Klinger

AU - Zhang, Weinan

AU - Wang, Yuchen

AU - Ren, Kan

AU - Zhang, Weinan

AU - Wang, Jun

AU - Yu, Yong

AU - Qu, Yanru

AU - Cai, Han

AU - Ren, Kan

AU - Zhang, Weinan

AU - Yu, Yong

AU - Wen, Ying

AU - Wang, Jun

AU - Zhang, Haifeng

AU - Zhang, Weinan

AU - Wang, Jun

AU - Rong, Yifei

AU - Cai, Han

AU - Ren, Kan

AU - Zhang, Weinan

AU - Malialis, Kleanthis

AU - Wang, Jun

AU - Yu, Yong

AU - Guo, Defeng

AU - Lin, Chi-Chun

AU - Chuang, Kun-Ta

AU - Wu, Wush Chi-Hsuan

AU - Chen, Ming-Syan

AU - Ren, Kan

AU - Zhang, Weinan

AU - Rong, Yifei

AU - Zhang, Haifeng

AU - Yu, Yong

AU - Wang, Jun

AU - Perlich, Claudia

AU - Dalessandro, Brian

AU - Hook, Rod

AU - Stitelman, Ori

AU - Raeder, Troy

AU - Provost, Foster

AU - Shawe-Taylor, John

AU - Cristianini, Nello

AU - Muthukrishnan, S

AU - Zhang, Weinan

AU - Yuan, Shuai

AU - Wang, Jun

AU - Ren, Kan

AU - Wang, Jun

AU - Zhou, Tianxiong

AU - Wang, Jun

AU - Xu, Jian

AU - Graepel, Thore

AU - Candela, Joaquin Q

AU - Borchert, Thomas

AU - Herbrich, Ralf

AU - He, Xinran

AU - Pan, Junfeng

AU - Jin, Ou

AU - Xu, Tianbing Tao

AU - Liu, Bo

AU - Xu, Tianbing Tao

AU - Shi, Yanxin

AU - Atallah, Antoine

AU - Herbrich, Ralf

AU - Bowers, Stuart

AU - Oentaryo, Richard J

AU - Lim, Ee-Peng

AU - Low, Jia-Wei

AU - Lo, David

AU - Finegold, Michael

AU - McMahan, H Brendan

AU - Holt, Gary

AU - Sculley, David

AU - Young, Michael

AU - Ebner, Dietmar

AU - Grady, Julian

AU - Nie, Lan

AU - Phillips, Todd

AU - Davydov, Eugene

AU - Golovin, Daniel

AU - Richardson, Matthew

AU - Dominowska, Ewa

AU - Ragno, Robert

AU - Cui, Ying

AU - Zhang, Ruofei

AU - Li, Wei Wentong

AU - Mao, Jianchang

AU - Chapelle, Olivier

AU - Zhang, Weinan

AU - Yuan, Shuai

AU - Wang, Jun

AU - Cetintas, Suleyman

AU - Chen, Datong

AU - Si, Luo

AU - Zhang, Weinan

AU - Wang, Jun

AU - Rong, Yifei

AU - Wang, Jun

AU - Zhu, Tianchi

AU - Wang, Xiaofan Xuejian Xiaoling

AU - Rendle, Steffen

AU - Freudenthaler, Christoph

AU - Schmidt-Thieme, Lars

AU - Liao, Hairen

AU - Peng, Lingxiao

AU - Liu, Zhenchuan

AU - Shen, Xuehua

AU - Amin, Kareem

AU - Kearns, Michael

AU - Key, Peter

AU - Schwaighofer, Anton

AU - Google,

AU - Lee, Kuang-Chih

AU - Orten, Burkay

AU - Dasdan, Ali

AU - Li, Wei Wentong

AU - Jalali, Ali

AU - Dasdan, Ali

AU - Ta, Anh-Phuong

AU - Menon, Aditya Krishna

AU - Chitrapura, Krishna-Prasad

AU - Garg, Sachin

AU - Agarwal, Deepak

AU - Kota, Nagaraj

AU - Chen, Bee-Chung

AU - Elango, Pradheep

AU - Agrawal, Rahul

AU - Khanna, Rajiv

AU - Kota, Nagaraj

AU - Zhang, Weinan

AU - Du, Tianming

AU - Wang, Jun

AU - Edelman, Benjamin

AU - Ostrovsky, Michael

AU - Schwarz, Michael

AU - Chen, Ye

AU - Berkhin, Pavel

AU - Anderson, Bo

AU - Devanur, Nikhil R

AU - Li, Xiang

AU - Guan, Devin

AU - Xu, Jian

AU - Shao, Xuhui

AU - Ma, Jianjie

AU - Lee, Kuang-Chih

AU - Qi, Hang

AU - Lu, Quan

PY - 2016

Y1 - 2016

N2 - On the Internet there are publishers (the supply side) who provide free contents (e.g., news) and services (e.g., email) to attract users. Publishers get paid by selling ad displaying opportunities (i.e., impressions) to advertisers. Advertisers then sell products to users who are converted by ads. Better supply side revenue allows more free content and services to be created, thus, benefiting the entire online advertising ecosystem. This thesis addresses several optimisation problems for the supply side. When a publisher creates an ad-supported website, he needs to decide the percentage of ads first. The thesis reports a large-scale empirical study of Internet ad density over past seven years, then presents a model that includes many factors, especially the competition among similar publishers, and gives an optimal dynamic ad density that generates the maximum revenue over time. This study also unveils the tragedy of the commons in online advertising where users’ attention has been overgrazed which results in a global sub-optimum. After deciding the ad density, the publisher retrieves ads from various sources, including contracts, ad networks, and ad exchanges. This forms an explorationexploitation problem when ad sources are typically unknown before trail. This problem is modelled using Partially Observable Markov Decision Process (POMDP), and the exploration efficiency is increased by utilising the correlation of ads. The proposed method reports 23.4% better than the best performing baseline in the real-world data based experiments. Since some ad networks allow (or expect) an input of keywords, the thesis also presents an adaptive keyword extraction system using BM25F algorithm and the multi-armed bandits model. This system has been tested by a domain service provider in crowdsourcing based experiments. If the publisher selects a Real-Time Bidding (RTB) ad source, he can use reserve price to manipulate auctions for better payoff. This thesis proposes a simplified game model that considers the competition between seller and buyer to be one-shot instead of repeated and gives heuristics that can be easily implemented. The model has been evaluated in a production environment and reported 12.3% average increase of revenue. The documentation of a prototype system for reserve price optimisation is also presented in the appendix of the thesis.

AB - On the Internet there are publishers (the supply side) who provide free contents (e.g., news) and services (e.g., email) to attract users. Publishers get paid by selling ad displaying opportunities (i.e., impressions) to advertisers. Advertisers then sell products to users who are converted by ads. Better supply side revenue allows more free content and services to be created, thus, benefiting the entire online advertising ecosystem. This thesis addresses several optimisation problems for the supply side. When a publisher creates an ad-supported website, he needs to decide the percentage of ads first. The thesis reports a large-scale empirical study of Internet ad density over past seven years, then presents a model that includes many factors, especially the competition among similar publishers, and gives an optimal dynamic ad density that generates the maximum revenue over time. This study also unveils the tragedy of the commons in online advertising where users’ attention has been overgrazed which results in a global sub-optimum. After deciding the ad density, the publisher retrieves ads from various sources, including contracts, ad networks, and ad exchanges. This forms an explorationexploitation problem when ad sources are typically unknown before trail. This problem is modelled using Partially Observable Markov Decision Process (POMDP), and the exploration efficiency is increased by utilising the correlation of ads. The proposed method reports 23.4% better than the best performing baseline in the real-world data based experiments. Since some ad networks allow (or expect) an input of keywords, the thesis also presents an adaptive keyword extraction system using BM25F algorithm and the multi-armed bandits model. This system has been tested by a domain service provider in crowdsourcing based experiments. If the publisher selects a Real-Time Bidding (RTB) ad source, he can use reserve price to manipulate auctions for better payoff. This thesis proposes a simplified game model that considers the competition between seller and buyer to be one-shot instead of repeated and gives heuristics that can be easily implemented. The model has been evaluated in a production environment and reported 12.3% average increase of revenue. The documentation of a prototype system for reserve price optimisation is also presented in the appendix of the thesis.

UR - https://www.mendeley.com/catalogue/4a40f0b1-99ce-3694-867d-8fc4f5f6c65c/

M3 - Article

VL - 56

SP - 1

EP - 6

JO - Kdd

JF - Kdd

IS - 3

ER -

Supply side optimisation in online display advertising (Chapter 3)

Abstract

Other files and links

Cite this