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

2025-10-22

AI Summary

Online platform marketing optimization faces dual challenges: misalignment between prediction and decision objectives, and the trade-off between bias in observational data and high variance in experimental data. To address these, we propose Bi-DFCL, a bilevel decision-focused causal learning framework. It constructs an unbiased decision-quality estimator from experimental data and jointly optimizes observational and experimental data via implicit differentiation—thereby theoretically breaking the bias-variance trade-off. Crucially, it backpropagates gradients from the downstream decision objective into the causal effect prediction model, enabling end-to-end, decision-performance-driven learning. The method integrates causal inference, bilevel optimization, and surrogate loss design. Extensive evaluations on public benchmarks, industrial datasets, and large-scale online A/B tests at Meituan demonstrate significant improvements over state-of-the-art methods. Bi-DFCL has been deployed in production, yielding statistically significant gains in user retention and platform revenue.

Technology Category

• Machine Learning: Causal Learning
• Search and Optimization: Learning to Search
• Reasoning under Uncertainty: Stochastic Optimization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Economics and fairness of platforms and recommendation systems
• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising

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.

Problem

• Addresses prediction-decision misalignment in marketing optimization
• Resolves bias-variance dilemma between observational and experimental data
• Develops joint optimization framework for improved marketing decisions

Innovation

• Bi-level optimization framework integrates observational and experimental data
• Unbiased estimator guides machine learning via surrogate loss functions
• Implicit differentiation solves joint optimization for bias-variance tradeoff