Xin Fu

Bio: Xin Fu is an academic researcher from Facebook. The author has contributed to research in topics: User experience evaluation & Open research. The author has an hindex of 1, co-authored 3 publications receiving 6 citations.

Challenges, Best Practices and Pitfalls in Evaluating Results of Online Controlled Experiments

Abstract: A/B Testing is the gold standard to estimate the causal relationship between a change in a product and its impact on key outcome measures. It is widely used in the industry to test changes ranging from simple copy change or UI change to more complex changes like using machine learning models to personalize user experience. The key aspect of A/B testing is evaluation of experiment results. Designing the right set of metrics - correct outcome measures, data quality indicators, guardrails that prevent harm to business, and a comprehensive set of supporting metrics to understand the "why" behind the key movements is the #1 challenge practitioners face when trying to scale their experimentation program [18, 22]. On the technical side, improving sensitivity of experiment metrics is a hard problem and an active research area, with large practical implications as more and more small and medium size businesses are trying to adopt A/B testing and suffer from insufficient power. In this tutorial we will discuss challenges, best practices, and pitfalls in evaluating experiment results, focusing on both lessons learned and practical guidelines as well as open research questions.

Challenges, Best Practices and Pitfalls in Evaluating Results of Online Controlled Experiments

Abstract: A/B Testing is the gold standard to estimate the causal relationship between a change in a product and its impact on key outcome measures. It is widely used in the industry to test changes ranging from simple copy change or UI change to more complex changes like using machine learning models to personalize user experience. The key aspect of A/B testing is evaluation of experiment results. Designing the right set of metrics - correct outcome measures, data quality indicators, guardrails that prevent harm to business, and a comprehensive set of supporting metrics to understand the "why" behind the key movements is the #1 challenge practitioners face when trying to scale their experimentation program [11, 14]. On the technical side, improving sensitivity of experiment metrics is a hard problem and an active research area, with large practical implications as more and more small and medium size businesses are trying to adopt A/B testing and suffer from insufficient power. In this tutorial we will discuss challenges, best practices, and pitfalls in evaluating experiment results, focusing on both lessons learned and practical guidelines as well as open research questions. A version of this tutorial was also present at KDD 2019 [23]. It was attended by around 150 participants.

Challenges, Best Practices and Pitfalls in Evaluating Results of Online Controlled Experiments

Abstract: A/B Testing is the gold standard to estimate the causal relationship between a change in a product and its impact on key outcome measures. It is widely used in the industry to test changes ranging from simple copy change or UI change to more complex changes like using machine learning models to personalize user experience. The key aspect of A/B testing is evaluation of experiment results. Designing the right set of metrics - correct outcome measures, data quality indicators, guardrails that prevent harm to business, and a comprehensive set of supporting metrics to understand the “why” behind the key movements is the #1 challenge practitioners face when trying to scale their experimentation program 11, 14. On the technical side, improving sensitivity of experiment metrics is a hard problem and an active research area, with large practical implications as more and more small and medium size businesses are trying to adopt A/B testing and suffer from insufficient power. In this tutorial we will discuss challenges, best practices, and pitfalls in evaluating experiment results, focusing on both lessons learned and practical guidelines as well as open research questions. A version of this tutorial was also present at KDD 2019 23. It was attended by around 150 participants. This tutorial has also been accepted for the WSDM 2020 conference.

How to Measure Your App: A Couple of Pitfalls and Remedies in Measuring App Performance in Online Controlled Experiments

Abstract: Effectively measuring, understanding, and improving mobile app performance is of paramount importance for mobile app developers. Across the mobile Internet landscape, companies run online controlled experiments (A/B tests) with thousands of performance metrics in order to understand how app performance causally impacts user retention and to guard against service or app regressions that degrade user experiences. To capture certain characteristics particular to performance metrics, such as enormous observation volume and high skewness in distribution, an industry-standard practice is to construct a performance metric as a quantile over all performance events in control or treatment buckets in A/B tests. In our experience with thousands of A/B tests provided by Snap, we have discovered some pitfalls in this industry-standard way of calculating performance metrics that can lead to unexplained movements in performance metrics and unexpected misalignment with user engagement metrics. In this paper, we discuss two major pitfalls in this industry-standard practice of measuring performance for mobile apps. One arises from strong heterogeneity in both mobile devices and user engagement, and the other arises from self-selection bias caused by post-treatment user engagement changes. To remedy these two pitfalls, we introduce several scalable methods including user-level performance metric calculation and imputation and matching for missing metric values. We have extensively evaluated these methods on both simulation data and real A/B tests, and have deployed them into Snap's in-house experimentation platform.

Challenges, Best Practices and Pitfalls in Evaluating Results of Online Controlled Experiments

Abstract: A/B Testing is the gold standard to estimate the causal relationship between a change in a product and its impact on key outcome measures. It is widely used in the industry to test changes ranging from simple copy change or UI change to more complex changes like using machine learning models to personalize user experience. The key aspect of A/B testing is evaluation of experiment results. Designing the right set of metrics - correct outcome measures, data quality indicators, guardrails that prevent harm to business, and a comprehensive set of supporting metrics to understand the "why" behind the key movements is the #1 challenge practitioners face when trying to scale their experimentation program [11, 14]. On the technical side, improving sensitivity of experiment metrics is a hard problem and an active research area, with large practical implications as more and more small and medium size businesses are trying to adopt A/B testing and suffer from insufficient power. In this tutorial we will discuss challenges, best practices, and pitfalls in evaluating experiment results, focusing on both lessons learned and practical guidelines as well as open research questions. A version of this tutorial was also present at KDD 2019 [23]. It was attended by around 150 participants.

Modern Theoretical Tools for Understanding and Designing Next-generation Information Retrieval System

Abstract: In the relatively short history of machine learning, the subtle balance between engineering and theoretical progress has been proved critical at various stages. The most recent wave of AI has brought to the IR community powerful techniques, particularly for pattern recognition. While many benefits from the burst of ideas as numerous tasks become algorithmically feasible, the balance is tilting toward the application side. The existing theoretical tools in IR can no longer explain, guide, and justify the newly-established methodologies. With no choices, we have to bet our design on black-box mechanisms that we only empirically understand. The consequences can be suffering: in stark contrast to how the IR industry has envisioned modern AI making life easier, many are experiencing increased confusion and costs in data manipulation, model selection, monitoring, censoring, and decision making. This reality is not surprising: without handy theoretical tools, we often lack principled knowledge of the pattern recognition model's expressivity, optimization property, generalization guarantee, and our decision-making process has to rely on over-simplified assumptions and human judgments from time to time. Facing all the challenges, we started researching advanced theoretical tools emerging from various domains that can potentially resolve modern IR problems. We encountered many impactful ideas and made several independent publications emphasizing different pieces. Time is now to bring the community a systematic tutorial on how we successfully adapt those tools and make significant progress in understanding, designing, and eventually productionize impactful IR systems. We emphasize systematicity because IR is a comprehensive discipline that touches upon particular aspects of learning, causal inference analysis, interactive (online) decision-making, etc. It thus requires systematic calibrations to render the actual usefulness of the imported theoretical tools to serve IR problems, as they usually exhibit unique structures and definitions. Therefore, we plan this tutorial to systematically demonstrate our learning and successful experience of using advanced theoretical tools for understanding and designing IR systems.

How to Measure Your App: A Couple of Pitfalls and Remedies in Measuring App Performance in Online Controlled Experiments

Abstract: Effectively measuring, understanding, and improving mobile app performance is of paramount importance for mobile app developers. Across the mobile Internet landscape, companies run online controlled experiments (A/B tests) with thousands of performance metrics in order to understand how app performance causally impacts user retention and to guard against service or app regressions that degrade user experiences. To capture certain characteristics particular to performance metrics, such as enormous observation volume and high skewness in distribution, an industry-standard practice is to construct a performance metric as a quantile over all performance events in control or treatment buckets in A/B tests. In our experience with thousands of A/B tests provided by Snap, we have discovered some pitfalls in this industry-standard way of calculating performance metrics that can lead to unexplained movements in performance metrics and unexpected misalignment with user engagement metrics. In this paper, we discuss two major pitfalls in this industry-standard practice of measuring performance for mobile apps. One arises from strong heterogeneity in both mobile devices and user engagement, and the other arises from self-selection bias caused by post-treatment user engagement changes. To remedy these two pitfalls, we introduce several scalable methods including user-level performance metric calculation and imputation and matching for missing metric values. We have extensively evaluated these methods on both simulation data and real A/B tests, and have deployed them into Snap's in-house experimentation platform.

User Sentiment as a Success Metric: Persistent Biases Under Full Randomization

Abstract: We study user sentiment (reported via optional surveys) as a metric for fully randomized A/B tests. Both user-level covariates and treatment assignment can impact response propensity. We show that a simple mean comparison produces biased population level estimates and propose a set of consistent estimators for the average and local treatment effects on treated and respondent users. We show that our problem can be mapped onto the intersection of the missing data problem and observational causal inference, and we identify conditions under which consistent estimators exist. Finally, we evaluate the performance of estimators and find that more complicated models do not necessarily provide superior performance as long as models satisfy consistency criteria.