Conduct marketing data analytics activities for Philips’ Personal Health products to support business on market opportunities and retention risk. Through machine learning and data mining, find innovative and efficient way to gain business insight.Project: Marketing Mix ModelingBuild statistical solution on marketing and business drivers time series data to estimate various marketing tactics on sales and then forecast the impact of future sets of tactics.

Analyze customer vehicle data to provide answers to improve customer understanding and develop new service. Enables the validation and function design to be based on real quantitative data. Project: • Vehicle Data Collection China Function Owner of massive vehicle data collection and analytic. o Team lead of experts including hardware, software, and QA.o Coordinate with German headquarter to setup vehicle data collection platform including on board and off board. Including legal, telematics service provider, IT, etc.• Customer Data Analytic China Create a comprehensive understanding of vehicle usage patterns and insights into vehicle equipment usage of customers in China. Evaluated research topics based on large dataset. Many measures / potential improvement for new product design. Several patents created.o Data mining: Causal inference, clustering, and variable correlation. Event prediction.

Support factory and design center to improve product quality and reliability consistently. Perform analysis to improve test efficiency when New Product Lunch.Perform data exploration on supplier/manufacture data to find key attribute/parametric correlated to produce defect. Predict product failure rate in field. Use Weibull distribution to estimate product Annual Failure Rate and MTBF, etc. Perform DOE and ALTA (Accelerate Factor) analysis to improve test.Project:• Main contributor of “End User Issue Key Attributes Analysis System”. The system is a centralized webpage to perform data mining for key attributes analysis. It short the first analysis duration from 1-2 days to several Mins.• Machine Learning for Hard Disk Drive Failure Analysis. Decision tree was applied to analysis fail product log and classify failure mode automatically. Its correction rate was ~80%.