2017 AdKDD & TargetAd

Title: Machine Learning and Causal Inference for Advertising Effectiveness

Abstract: This talk will review several recent methodological approaches to measuring the effectiveness of treatments, such as advertising campaigns.  We will discuss methods for dealing with non-random assignment of advertisements to users, as well as methods for estimating optimal policies for assigning advertising to users.

Title: Learning from Logged Interventions

Abstract: Every time a system places an ad, presents a search ranking, or makes a recommendation, we can think about this as an intervention for which we can observe the user's response (e.g. click, dwell time, purchase). Such logged intervention data is actually one of the most plentiful types of data available, as it can be recorded from a variety of systems (e.g., search engines, recommender systems, ad placement) at little cost. However, this data provides only partial-information feedback -- aka ``bandit feedback'' -- limited to the particular intervention chosen by the system. We don't get to see how the user would have responded, if we had chosen a different intervention. This makes learning from logged bandit feedback substantially different from conventional supervised learning, where ``correct'' predictions together with a loss function provide full-information feedback. 

In this talk, I will explore approaches and methods for batch learning from logged bandit feedback (BLBF). Unlike the well-explored problem of online learning with bandit feedback, batch learning with bandit feedback does not require interactive experimental control of the underlying system, but merely exploits logged intervention data collected in the past. The talk presents a new inductive principle for BLBF, new counterfactual risk estimators, and new methods for structured output prediction with BLBF with applications to ad placement.

Title: Users and Time

Abstract: User personalization and modeling is often carried out in a parametric and discretized manner, as a matter of convenience. In this talk I will show that this need not be the case. Quite the opposite - it is possible to design continuous-time nonparametric models that fit the data well, that are easy to implement and that are quite flexible in terms of their ability to accept side information. I will illustrate this for user return time modeling, recommender systems and page views. 

Title: Incrementality Bidding & Attribution

Abstract: The causal effect of showing an ad to a potential customer versus not, commonly referred to as “incrementality,” is the fundamental question of advertising effectiveness. In digital advertising three major puzzle pieces are central to rigorously quantifying advertising incrementality: ad buying/bidding/pricing, attribution, and experimentation. Building on the foundations of machine learning and causal econometrics, we propose a methodology that unifies these three concepts into a computationally viable model of both bidding and attribution which spans randomization, training, cross validation, scoring, and conversion attribution in a causal model of advertising’s effects. Thanks to this method, Netflix has benefited by identifying many cases where traditional models were either overspending or underspending, leading to a significant improvement in the return on investment of advertising.