Uplift modeling for location-based online advertising

Kawanaka, Shogo; Moriwaki, Daisuke

doi:10.1145/3356994.3365505

Cited by 12 publications

(9 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…This information is also effective for marketing strategies, such as coupons (Molitor et al, 2020). In addition, the most recent papers in this topic link the location advertisement to the causal inference machine learning to assess the effect of given advertisements (Kawanaka & Moriwaki, 2019;Moriwaki et al, 2020).…”

Section: Location Base Advertisement and Commercial Areamentioning

confidence: 99%

“…Our interest in this task is to sort the customers by descending order of i , and distribute the ad following that order to maximize the aggregated effect of i . In the literature of computer science, this i is called as lift effect and the prediction models of lift effect are called as uplift model (Gutierrez & Gérardy, 2017;Kawanaka & Moriwaki, 2019;Rzepakowski & Jaroszewicz, 2012;Zaniewicz & Jaroszewicz, 2013).…”

Section: The Aim Of Uplift Modelmentioning

confidence: 99%

See 1 more Smart Citation

Online-to-offline advertisements as field experiments

Matsui

Moriwaki

2021

JER

Self Cite

View full text Add to dashboard Cite

Online advertisements have become one of today’s most widely used tools for enhancing businesses partly because of their compatibility with A/B testing. A/B testing allows sellers to find effective advertisement strategie,s such as ad creatives or segmentations. Even though several studies propose a technique to maximize the effect of an advertisement, there is insufficient comprehension of the customers’ offline shopping behavior invited by the online advertisements. Herein, we study the difference in offline behavior between customers who received online advertisements and regular customers (i.e., the customers visits the target shop voluntary), and the duration of this difference. We analyze approximately three thousand users’ offline behavior with their 23.5 million location records through 31 A/B testings. We first demonstrate the externality that customers with advertisements traverse larger areas than those without advertisements, and this spatial difference lasts several days after their shopping day. We then find a long-run effect of this externality of advertising that a certain portion of the customers invited to the offline shops revisit these shops. Finally, based on this revisit effect findings, we utilize a causal machine learning model to propose a marketing strategy to maximize the revisit ratio. Our results suggest that advertisements draw customers who have different behavior traits from regular customers. This study demonstrates that a simple analysis may underrate the effects of advertisements on businesses, and an analysis considering externality can attract potentially valuable customers.

show abstract

Section: Location Base Advertisement and Commercial Areamentioning

confidence: 99%

Section: The Aim Of Uplift Modelmentioning

confidence: 99%

Online-to-offline advertisements as field experiments

Matsui

Moriwaki

2021

JER

Self Cite

View full text Add to dashboard Cite

show abstract

“…The attempts to incorporate this lift-effect in bidding strategy are lift-based bidding [11], incrementality bidding [4], or more broadly uplift modeling [5,[7][8][9]12]. We advance this line of literature and address the inherent bias in impression log data in a theoretically grounded and tractable manner.…”

Section: Background and Related Workmentioning

confidence: 99%

“…( 5). In the A/B testing, we randomly assigned the two systems to users on an online ad campaign that aims to promotes the app released by a major consumer electronics retailer in Japan 5 . The primary aim of the campaign is to increase the number of app users and visitors to the real stores located across Japan.…”

Section: Online Experiments 61 Experimental Designmentioning

confidence: 99%

Unbiased Lift-based Bidding System

Moriwaki,

Hayakawa,

Munemasa

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Conventional bidding strategies for online display ad auction heavily relies on observed performance indicators such as clicks or conversions. A bidding strategy naively pursuing these easily observable metrics, however, fails to optimize the profitability of the advertisers. Rather, the bidding strategy that leads to the maximum revenue is a strategy pursuing the performance lift of showing ads to a specific user. Therefore, it is essential to predict the lift-effect of showing ads to each user on their target variables from observed log data. However, there is a difficulty in predicting the lift-effect, as the training data gathered by a past bidding strategy may have a strong bias towards the winning impressions. In this study, we develop Unbiased Lift-based Bidding System, which maximizes the advertisers' profit by accurately predicting the lift-effect from biased log data. Our system is the first to enable high-performing lift-based bidding strategy by theoretically alleviating the inherent bias in the log. Real-world, large-scale A/B testing successfully demonstrates the superiority and practicability of the proposed system. CCS CONCEPTS• Information systems → Display advertising.

show abstract

“…Uplift modeling Gutierrez & Gérardy (2017); Hu et al (2020); Knaus et al (2020); Radcliffe & Surry aims to identify individuals who benefit the most as a result of receiving a certain intervention. This has a wide range of applications, including political or marketing campaigns Rzepakowski & Jaroszewicz (2012b), policy-making Guelman et al (2012), ad targeting Kawanaka & Moriwaki (2019), personalized medicine Jaskowski & Jaroszewicz (2012), and much more. Under Rubin's framework of causal inference Rubin (1974), uplift modeling amounts to estimating the Conditional Average Treatment Effect (a.k.a.…”

Section: Introductionmentioning

confidence: 99%

Generalized Causal Tree for Uplift Modeling

Nandy¹,

Yu²,

Liu³

et al. 2022

Preprint

View full text Add to dashboard Cite

Uplift modeling is crucial in various applications ranging from marketing and policymaking to personalized recommendations. The main objective is to learn optimal treatment allocations for a heterogeneous population. A primary line of existing work modifies the loss function of the decision tree algorithm to identify cohorts with heterogeneous treatment effects. Another line of work estimates the individual treatment effects separately for the treatment group and the control group using off-the-shelf supervised learning algorithms. The former approach that directly models the heterogeneous treatment effect is known to outperform the latter in practice. However, the existing tree-based methods are mostly limited to a single treatment and a single control use case, except for a handful of extensions to multiple discrete treatments. In this paper, we fill this gap in the literature by proposing a generalization to the tree-based approaches to tackle multiple discrete and continuous-valued treatments. We focus on a generalization of the well-known causal tree algorithm due to its desirable statistical properties, but our generalization technique can be applied to other tree-based approaches as well. We perform extensive experiments to showcase the efficacy of our method when compared to other methods.

show abstract

Uplift modeling for location-based online advertising

Cited by 12 publications

References 7 publications

Online-to-offline advertisements as field experiments

Online-to-offline advertisements as field experiments

Unbiased Lift-based Bidding System

Generalized Causal Tree for Uplift Modeling

Contact Info

Product

Resources

About