Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
166 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
42 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Online Causal Inference for Advertising in Real-Time Bidding Auctions (1908.08600v4)

Published 22 Aug 2019 in cs.LG, cs.GT, econ.EM, and stat.ML

Abstract: Real-time bidding (RTB) systems, which utilize auctions to allocate user impressions to competing advertisers, continue to enjoy success in digital advertising. Assessing the effectiveness of such advertising remains a challenge in research and practice. This paper proposes a new approach to perform causal inference on advertising bought through such mechanisms. Leveraging the economic structure of first- and second-price auctions, we first show that the effects of advertising are identified by the optimal bids. Hence, since these optimal bids are the only objects that need to be recovered, we introduce an adapted Thompson sampling (TS) algorithm to solve a multi-armed bandit problem that succeeds in recovering such bids and, consequently, the effects of advertising while minimizing the costs of experimentation. We derive a regret bound for our algorithm which is order optimal and use data from RTB auctions to show that it outperforms commonly used methods that estimate the effects of advertising.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (72)
  1. Bayesian analysis of binary and polychotomous response data. Journal of the American Statistical Association, 88(422):669–679.
  2. Amemyia, T. (1984). Tobit models: A survey. Journal of Econometrics, 24(1–2):3–61.
  3. Identification of standard auction models. Econometrica, 70(6):2107–2140.
  4. Reserve price optimization at scale. In Zaiane, O. R. and Matwin, S., editors, Proc. 2016 IEEE Internat. Conf. on Data Science and Advanced Analytics, pages 528–536. (IEEE, New York).
  5. Repeated auctions with budgets in ad exchanges: Approximations and design. Management Science, 61(4):864–884.
  6. Contextual bandits with cross-learning. In Wallach, H., Larochelle, H., Beygelzimer, A., d'Alché-Buc, F., Fox, E., and Garnett, R., editors, Advances in Neural Information Processing Systems 32, pages 9679–9688. (Curran Associates, Inc., New York).
  7. Learning in repeated auctions with budgets: Regret minimization and equilibrium. Management Science, 65(9):3952–3968.
  8. Bandits with unobserved confounders: A causal approach. In Cortes, C., Lawrence, N., Lee, D., Sugiyama, M., and Garnett, R., editors, Advances in Neural Information Processing Systems 28, pages 1342–1350. (Curran Associates, Inc., New York).
  9. Online decision-making with high-dimensional covariates. Operations Research, 68(1):276–294.
  10. Consumer heterogeneity and paid search effectiveness: A large-scale field experiment. Econometrica, 83(1):155–174.
  11. The reversed hazard rate function. Probability in the Engineering and Informational Sciences, 12(1):69–90.
  12. Causal models for real time bidding with repeated user interactions. In Zhu, F., Ooi, B. C., and Miao, C., editors, Proc. of the 27th ACM SIGKDD Internat. Conf. on Knowledge Discovery & Data Mining, pages 75–85. (ACM, New York).
  13. Pure exploration in multi-armed bandits problems. In Gavaldà, R., Lugosi, G., Zeugmann, T., and Zilles, S., editors, Internat. Conf. on Algorithmic Learning Theory, pages 23–37. (Springer, Berlin).
  14. Real-time bidding by reinforcement learning in display advertising. In de Rijke, M. and Shokouhi, M., editors, Proc. of the Tenth ACM Internat. Conf. on Web Search and Data Mining, pages 661–670. (ACM, New York).
  15. Chapelle, O. (2014). Modeling delayed feedback in display advertising. In Macskassy, S. and Perlich, C., editors, Proc. of the 21st ACM SIGKDD Internat. Conf. on Knowledge Discovery & Data Mining, pages 1097–1105. (ACM, New York).
  16. A/B testing of auctions. arXiv preprint arXiv:1606.00908.
  17. Bayesian analysis of cross-section and clustered data treatment models. Journal of Econometrics, 97(1):25–50.
  18. Bridging the gap between regret minimization and best arm identification, with application to A/B tests. In Chaudhuri, K. and Sugiyama, M., editors, Proc. of the Twenty-Second Internat. Conf. on Artificial Intelligence and Statistics, pages 1988–1966. (PMLR, Naha).
  19. Can credit increase revenue? In Chen, Y. and Immorlica, N., editors, 9th Internat. Conf. on Web and Internet Economics, pages 121–133. (Springer, Berlin).
  20. Estimation considerations in contextual bandits. arXiv preprint arXiv:1711.07077.
  21. Test & roll: Profit-maximizing A/B tests. Marketing Science, 38(6):1038–1058.
  22. Learning to bid without knowing your value. In Tardos, É., editor, Proc. of the 2018 ACM Conf. on Economics and Computation, pages 505–522. (ACM, New York).
  23. Real-time bidding with side information. In Guyon, I., Von Luxburg, U., Bengio, S., Wallach, H., Fergus, R., Vishwanathan, S., and Garnett, R., editors, Advances in Neural Information Processing Systems 30, pages 5163–5173. (Curran Associates, Inc., New York).
  24. Counterfactual data-fusion for online reinforcement learners. In Precup, D. and Teh, Y. W., editors, Proc. of the 34th Internat.l Conf. on Machine Learning, pages 1156–1164. (PMLR, Sidney).
  25. On explore-then-commit strategies. In Lee, D., Sugiyama, M., Luxburg, U., Guyon, I., and Garnett, R., editors, Advances in Neural Information Processing Systems 29, pages 784–792. (Curran Associates, Inc., New York).
  26. Online evaluation of audiences for targeted advertising via bandit experiments. In Stone, P., editor, Proc. of the Thirty-Fourth AAAI Conf. on Artificial Intelligence, pages 13273–13279. (AAAI Press, Palo Alto).
  27. A linear response bandit problem. Stochastic Systems, 3(1):230–261.
  28. Inefficiencies in digital advertising markets. Journal of Marketing, 85(1):7–25.
  29. Confidence intervals for policy evaluation in adaptive experiments. arXiv preprint arXiv:1911.02768.
  30. Learning to bid optimally and efficiently in adversarial first-price auctions. arXiv preprint arXiv:2007.04568.
  31. Optimal no-regret learning in repeated first-price auctions. arXiv preprint arXiv:2003.09795.
  32. Holland, P. W. (1986). Statistics and causal inference. Journal of the American Statistical Association, 81(396):945–960.
  33. Bayesian inference for causal effects in randomized experiments with noncompliance. Annals of Statistics, 25(1):305–327.
  34. Mean field equilibria of dynamic auctions with learning. Management Science, 60(12):2949–2970.
  35. A bandit approach to sequential experimental design with false discovery control. In Bengio, S., Wallach, H., Larochelle, H., Grauman, K., Cesa-Bianchi, N., and Garnett, R., editors, Advances in Neural Information Processing Systems 31, pages 3664–3674. (Curran Associates, Inc., New York).
  36. Real-time bidding with multi-agent reinforcement learning in display advertising. In Cuzzocrea, A., editor, Proc. of the 27th ACM Internat.l Conf. on Information and Knowledge Management, pages 2193–2201. (ACM, New York).
  37. Ghost ads: Improving the economics of measuring online ad effectiveness. Journal of Marketing Research, 54(6):867–884.
  38. A sequential test for selecting the better variant: Online A/B testing, adaptive allocation, and continuous monitoring. In Culpepper, J. S. and Moffat, A., editors, Proc. of the Twelth ACM Internat. Conf. on Web Search and Data Mining, pages 492–500. (ACM, New York).
  39. Kallus, N. (2018). Instrument-armed bandits. In Janoos, F., Mohri, M., and Sridharan, K., editors, Internat. Conf. on Algorithmic Learning Theory, pages 529–546. (PMLR, Lanzarote).
  40. Mechanism design with bandit feedback. arXiv preprint arXiv:2004.08924.
  41. Causal bandits: Learning good interventions via causal inference. In Lee, D., Sugiyama, M., Luxburg, U., Guyon, I., and Garnett, R., editors, Advances in Neural Information Processing Systems 29, pages 1181–1189. (Curran Associates, Inc., New York).
  42. Incrementality bidding & attribution. SSRN:3129350.
  43. Manski, C. F. (1997). Monotone treatment response. Econometrica, 65(6):1311–1334.
  44. A theory of auctions and competitive bidding. Econometrica, 50(5):1089–1122.
  45. Dynamic online pricing with incomplete information using multi-armed bandit experiments. Marketing Science, 38(2):226–252.
  46. Unbiased lift-based bidding system. arXiv preprint arXiv:2007.04002.
  47. Why adaptively collected data have negative bias and how to correct for it. In Storkey, A. and Perez-Cruz, F., editors, Proc. of the Twenty-First Internat. Conf. on Artificial Intelligence and Statistics, pages 1261–1269. (PMLR, Lanzarote).
  48. Reserve prices in internet advertising auctions: A field experiment. Working paper, Stanford University.
  49. Pearl, J. (2009). Causality. Cambridge University Press.
  50. Optimizing cluster-based randomized experiments under monotonicity. In Guo, Y. and Farooq, F., editors, Proc. of the 24th ACM SIGKDD Internat. Conf. on Knowledge Discovery & Data Mining, pages 2090–2099. (ACM, New York).
  51. Optimizing reserve prices for publishers in online ad auctions. In Ishibuchi, H. and Zhao, D., editors, 2019 IEEE Conf. on Computational Intelligence for Financial Engineering & Economics, pages 1–8. (IEEE, New York).
  52. Russo, D. (2020). Simple Bayesian algorithms for best arm identification. Operations Research, 68(6):1625–1647.
  53. A tutorial on Thompson sampling. Foundations and Trends® in Machine Learning, 11(1):1–96.
  54. An experimental investigation of the effects of retargeted advertising: The role of frequency and timing. Journal of Marketing Research, 56(3):401–418.
  55. Customer acquisition via display advertising using multi-armed bandit experiments. Marketing Science, 36(4):500–522.
  56. Scott, S. L. (2015). Multi-armed bandit experiments in the online service economy. Applied Stochastic Models in Business and Industry, 31(1):37–45.
  57. Predicting ad click-through rates via feature-based fully coupled interaction tensor factorization. Electronic Commerce Research and Applications, 16:268–282.
  58. Competition and crowd-out for brand keywords in sponsored search. Marketing Science, 37(2):200–215.
  59. Thompson, W. R. (1933). On the likelihood that one unknown probability exceeds another in view of the evidence of two samples. Biometrika, 25(3/4):285–294.
  60. A near-optimal bidding strategy for real-time display advertising auctions. Journal of Marketing Research, 58(1):1–21.
  61. Vijverberg, W. P. M. (1993). Measuring the unidentified parameter of the extended Roy model of selectivity. Journal of Econometrics, 57(1–3):69–89.
  62. Multi-armed bandit models for the optimal design of clinical trials: Benefits and challenges. Statistical Science, 30(2):199–215.
  63. Online inference for advertising auctions. arXiv preprint arXiv:1908.08600.
  64. Adaptive experimentation with delayed binary feedback. In WWW ’22: Proceedings of the ACM Web Conference 2022, page 2247–2255.
  65. Online learning in repeated auctions. In Feldman, V., Rakhlin, A., and Shamir, O., editors, 29th Annual Conf. on Learning Theory, pages 1562–1583. (PMLR, New York).
  66. A multi-agent reinforcement learning method for impression allocation in online display advertising. arXiv preprint arXiv:1809.03152.
  67. Predicting winning price in real time bidding with censored data. In Cao, L. and Zhang, C., editors, Proc. of the 21st ACM SIGKDD Internat. Conf. on Knowledge Discovery & Data Mining, pages 1305–1314. (ACM, New York).
  68. Lift-based bidding in ad selection. In Stone, P., editor, Proc. of the Thirtieth AAAI Conf. on Artificial Intelligence, pages 651–657. (AAAI Press, Palo Alto).
  69. Estimation bias in multi-armed bandit algorithms for search advertising. In Burges, C. J. C., Bottou, L., Welling, M., Ghahramani, Z., and Weinberger, K. Q., editors, Advances in Neural Information Processing Systems 26, pages 2400–2408. (Curran Associates, Inc., New York).
  70. A framework for Multi-A(rmed)/B(andit) testing with online FDR control. In Guyon, I., Luxburg, U. V., Bengio, S., Wallach, H., Fergus, R., Vishwanathan, S., and Garnett, R., editors, Advances in Neural Information Processing Systems 30, pages 5957–5966. (Curran Associates, Inc., New York).
  71. Real-time bidding benchmarking with iPinYou dataset. arXiv preprint arXiv:1407.7073.
  72. Bid-aware gradient descent for unbiased learning with censored data in display advertising. In Krishnapuram, B. and Shah, M., editors, Proc. of the 22nd ACM SIGKDD Internat. Conf. on Knowledge Discovery & Data Mining, pages 665–674. (ACM, New York).
Citations (8)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now