Multi-Metric Adaptive Experimental Design under Fixed Budget with Validation

TL;DR

This paper introduces a fixed-budget multi-metric adaptive experimental design framework with a two-phase process, improving statistical inference and treatment selection in online experiments with multiple metrics and heterogeneous variances.

Contribution

It proposes SHRVar, a novel multi-metric adaptive exploration method that generalizes sequential halving with variance-based sampling and elimination, providing exponential error reduction.

Findings

SHRVar outperforms existing methods in numerical experiments.

The framework effectively identifies the best treatment with high confidence.

The approach handles multiple metrics and heterogeneous variances efficiently.

Abstract

Standard A/B tests in online experiments face statistical power challenges when testing multiple candidates simultaneously, while adaptive experimental designs (AED) alone fall short in inferring experiment statistics such as the average treatment effect, especially with many metrics (e.g., revenue, safety) and heterogeneous variances. This paper proposes a fixed-budget multi-metric AED framework with a two-phase structure: an adaptive exploration phase to identify the best treatment, and a validation phase with an A/B test to verify the treatment's quality and infer statistics. We propose SHRVar, which generalizes sequential halving (SH) (Karnin et al., 2013) with a novel relative-variance-based sampling and an elimination strategy built on reward z-values. It achieves a provable error probability that decreases exponentially, where the exponent generalizes the complexity measure for SH (Karnin et al., 2013) and SHVar (Lalitha et al., 2023) with homogeneous and heterogeneous variances, respectively. Numerical experiments verify our analysis and demonstrate the superior performance of this new framework.