Experimental Science and Causal Inference Forum – Sessions Overview at DataFun Summit 2022

Background On May 21, 2022, from 9:00 to 13:30, the DataFun Summit 2022 hosts an Experimental Science and Causal Inference forum produced by former Didi data‑science platform and user‑growth lead Dong Yue, inviting six technical experts from Tencent, Kuaishou, Google, and ByteDance to share cutting‑edge practices.

Schedule (illustrated by the included image) outlines the timing of the sessions.

Detailed Introduction

Producer: Dong Yue – Master’s graduate of the University of Florida, former Facebook data scientist, and former head of Didi’s data‑science platform, responsible for large‑scale capabilities such as experimental science, causal inference, and time‑series forecasting.

Speaker 1: Fang Dong (Tencent Interactive Entertainment, Expert Data Scientist) – Background in data‑science applications and operations, Ph.D. from the University of York, former researcher at Bell Labs, and senior data‑science roles at UnitedHealth, NY Melon Bank, and Shopee.

Talk Title: Causal Inference in Games

Outline: Challenges of overlapping game activities, evaluation of external policy impact, repeated user interventions, and presentation of common causal methods (panel DID, propensity‑score matching, double‑robust estimation) applied to game scenarios.

Audience Benefits: Learn how to use causal inference for fine‑grained game operations, obtain robust causal results, and understand its importance for data scientists.

Speaker 2: Cheng Da‑xi (Kuaishou Economist) – B.A. from Peking University, M.S. in Business Analytics from UT Austin, former Ant Group data scientist, now Kuaishou economist focusing on experimental causal analysis for traffic ecosystems.

Talk Title: Complex Experiment Design in Two‑Sided Markets

Outline: Challenges of experiments in two‑sided markets, industry reference solutions, Kuaishou’s approach and its applicability.

Audience Benefits: End‑to‑end two‑sided market experiment design, analysis of traditional method drawbacks.

Speaker 3: Zhu Zhihua (Tencent Data Scientist) – Master’s in Statistics from East China Normal University, experience at eBay and Tencent advertising experiment systems.

Talk Title: Experiment Design for Advertising in Two‑Sided Markets

Outline: Explanation of two‑sided market effects, experimental challenges, and Tencent advertising case studies.

Audience Benefits: Understanding experimental difficulties, common designs, and advertising‑specific solutions.

Speaker 4: Du Jiali (Google Senior Data Scientist) – Experience at Booking.com, Twitter, and Google, focusing on user churn detection and performance optimization.

Talk Title: Quantile Metrics in A/B Testing

Outline: Importance of quantile metrics, statistical challenges, and industry practices for robust testing.

Audience Benefits: Identify difficulties of quantile‑based A/B tests and learn practical implementations.

Speaker 5: Zhou Xiaoyu (Kuaishou Economist Team Tech Lead) – Former SAS Institute developer of panel‑data procedures, now leads causal‑analysis tool development at Kuaishou.

Talk Title: Heterogeneous Causal Effect Modeling with Dual Neural Networks

Outline: Modeling heterogeneous treatment effects for scaling decisions, limitations of existing methods, and a new model for estimating scale effects.

Audience Benefits: Understand limitations of current HTE estimation and how to model scale effects.

Speaker 6: Hei Mengqi (ByteDance Data Scientist) – Former Microsoft Research scientist, contributor to the EconML library, focusing on observational causal inference.

Talk Title: Heterogeneous Causal Models and Strategic Deployment

Outline: From ATE to HTE, overview of HTE methods, and two real‑world deployment cases.

Audience Benefits: Gain insight into the gains of HTE over average treatment effect and practical applications in commercial settings.

Live Registration – QR code images provide free registration; participants are instructed to scan, join the group, and watch the live stream.