A pragmatic policy learning approach to account for users' fatigue in repeated auctions - Archive ouverte HAL Accéder directement au contenu
Pré-Publication, Document De Travail Année : 2024

A pragmatic policy learning approach to account for users' fatigue in repeated auctions

Résumé

Online advertising banners are sold in real-time through auctions. Typically, the more banners a user is shown, the smaller the marginal value of the next banner for this user is. This fact can be detected by basic ML models, that can be used to predict how previously won auctions decrease the current opportunity value. However, learning is not enough to produce a bid that correctly accounts for how winning the current auction impacts the future values. Indeed, a policy that uses this prediction to maximize the expected payoff of the current auction could be dubbed impatient because such policy does not fully account for the repeated nature of the auctions. Under this perspective, it seems that most bidders in the literature are impatient. Unsurprisingly, impatience induces a cost. We provide two empirical arguments for the importance of this cost of impatience. First, an offline counterfactual analysis and, second, a notable business metrics improvement by mitigating the cost of impatience with policy learning
Fichier principal
Vignette du fichier
article.pdf (468.25 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04646638 , version 1 (12-07-2024)

Licence

Copyright (Tous droits réservés)

Identifiants

  • HAL Id : hal-04646638 , version 1

Citer

Benjamin Heymann, Rémi Chan--Renous-Legoubin, Alexandre Gilotte. A pragmatic policy learning approach to account for users' fatigue in repeated auctions. 2024. ⟨hal-04646638⟩
0 Consultations
0 Téléchargements

Partager

Gmail Mastodon Facebook X LinkedIn More