Bi-Level Decision-Focused Causal Learning for Large-Scale Marketing Optimization: Bridging Observational and Experimental Data

TL;DR

Bi-DFCL tackles marketing optimization under budget by coupling a bi-level optimization framework with a Bridge Network to jointly exploit observational and experimental data. It introduces an unbiased decision-loss estimator from RCTs and two surrogate losses to enable end-to-end training that aligns prediction with downstream decisions, solved via implicit differentiation. The approach yields superior decision quality and robust bias-variance tradeoffs, demonstrated across offline benchmarks and large-scale online A/B tests, and has been deployed at Meituan. The work advances scalable, data-efficient marketing optimization under real-world budget constraints with practical impact on revenue and user engagement.

Abstract

Online Internet platforms require sophisticated marketing strategies to optimize user retention and platform revenue -- a classical resource allocation problem. Traditional solutions adopt a two-stage pipeline: machine learning (ML) for predicting individual treatment effects to marketing actions, followed by operations research (OR) optimization for decision-making. This paradigm presents two fundamental technical challenges. First, the prediction-decision misalignment: Conventional ML methods focus solely on prediction accuracy without considering downstream optimization objectives, leading to improved predictive metrics that fail to translate to better decisions. Second, the bias-variance dilemma: Observational data suffers from multiple biases (e.g., selection bias, position bias), while experimental data (e.g., randomized controlled trials), though unbiased, is typically scarce and costly -- resulting in high-variance estimates. We propose Bi-level Decision-Focused Causal Learning (Bi-DFCL) that systematically addresses these challenges. First, we develop an unbiased estimator of OR decision quality using experimental data, which guides ML model training through surrogate loss functions that bridge discrete optimization gradients. Second, we establish a bi-level optimization framework that jointly leverages observational and experimental data, solved via implicit differentiation. This novel formulation enables our unbiased OR estimator to correct learning directions from biased observational data, achieving optimal bias-variance tradeoff. Extensive evaluations on public benchmarks, industrial marketing datasets, and large-scale online A/B tests demonstrate the effectiveness of Bi-DFCL, showing statistically significant improvements over state-of-the-art. Currently, Bi-DFCL has been deployed at Meituan, one of the largest online food delivery platforms in the world.

Figures (3)

• Figure 1: The primal (left) and dual (right) formulations of the MTBAP.
• Figure 2: Overview of the Bi-Level Decision-Focused Causal Learning (Bi-DFCL) Framework.
• Figure 3: Robustness of Bi-DFCL under multiple budget values $B$ on Marketing Data I and II.

Theorems & Definitions (1)

• proof