Optimizing cluster-based randomized experiments under a monotonicity assumption

TL;DR

This paper introduces a new two-stage cluster-based experimental design under a monotonicity assumption, optimizing bias reduction in experiments with interference, particularly applied to online advertising auctions.

Contribution

It proposes a novel design framework that identifies the least biased estimator among cluster-based experiments under monotonicity, validated on real auction data.

Findings

The monotonicity condition holds in online advertising auctions.

The proposed design reduces bias compared to traditional methods.

Validation on auction data confirms effectiveness.

Abstract

Cluster-based randomized experiments are popular designs for mitigating the bias of standard estimators when interference is present and classical causal inference and experimental design assumptions (such as SUTVA or ITR) do not hold. Without an exact knowledge of the interference structure, it can be challenging to understand which partitioning of the experimental units is optimal to minimize the estimation bias. In the paper, we introduce a monotonicity condition under which a novel two-stage experimental design allows us to determine which of two cluster-based designs yields the least biased estimator. We then consider the setting of online advertising auctions and show that reserve price experiments verify the monotonicity condition and the proposed framework and methodology applies. We validate our findings on an advertising auction dataset.