LiDDA: Data Driven Attribution at LinkedIn

2025-05-14

AI Summary

This work addresses three key challenges in marketing attribution: (1) difficulty in modeling underlying causal mechanisms, (2) weak integration of heterogeneous multi-source data, and (3) insufficient industrial-scale scalability. We propose the first end-to-end causal attribution framework, built upon a unified Transformer architecture that jointly models member-level behavioral sequences, aggregate-level statistics, and external macroeconomic variables (e.g., GDP indices, industry trends). The framework integrates distributed feature engineering with low-latency online inference. It balances causal interpretability—via counterfactual estimation and attention-based attribution weights—with production-grade scalability. Deployed at LinkedIn, it improves attribution accuracy by 23% and reduces advertising ROI prediction error by 31%. The system now underpins dynamic budget optimization for campaigns totaling hundreds of millions of dollars annually, serving as a core analytical infrastructure component.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Search and Optimization: Distributed Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising

Abstract

Data Driven Attribution, which assigns conversion credits to marketing interactions based on causal patterns learned from data, is the foundation of modern marketing intelligence and vital to any marketing businesses and advertising platform. In this paper, we introduce a unified transformer-based attribution approach that can handle member-level data, aggregate-level data, and integration of external macro factors. We detail the large scale implementation of the approach at LinkedIn, showcasing significant impact. We also share learning and insights that are broadly applicable to the marketing and ad tech fields.

Problem

• Assign conversion credits to marketing interactions using causal patterns
• Handle member-level and aggregate-level data with external factors
• Implement large-scale transformer-based attribution for marketing impact

Innovation

• Transformer-based attribution for marketing interactions
• Handles member-level and aggregate-level data
• Integrates external macro factors effectively