Be responsible for all data analysis work in 【Mengtui】:(1) From 0 to 1: build KPI system, analytic system, BI system. Built ‘Order Attribution Model’ and ‘Customer Portrait System’(2) End to end solution: data monitoring and strategy evaluation of business, push strategy optimizationProject experience and achievement1. Customer Portrait SystemBuild 300+ tags of customers, incl. Demographic / Transaction / in-APP Action / Preference / Sense of Price tagsHow to use: (1) build “AIPL model” to help optimal operation, ~12% lift of ARPU; (2) personalized push and recall strategy, ~14% lift of recall rate2. ROI Lift Model for User Acquisition(1) Build Growth funnel analytic system (Ad Exposure → Ad Click → User Active → CPA → Purchase → Retention&Repurchase), get insight of new user, e.g. Wechat users prefer snack, Tiktok users prefer FMCG, Kuaishou users easy to purchase but hard to retention. Transform insight to multi-strategy paid traffic operation, ~8% lift of ROI(2) Base on the insight, build machine learning model. Using new users’ tag of Ad source, Ad form, IP, phone brand, phone system … to show different landing page and different goods when new user register. ~12% lift of ROI3. Repurchase Analysis, which changes top strategy of Mengtui in 2020Background: the biggest issue of Mengtui is low repurchase rate, low repurchase rate means low LTV and not profitable.Analysis method: break down repurchase rate to retention rate and conversion rate, and find the key influence factor of them – for retention rate, customer experience, like logistics time and goods quality, is the key; for conversion rate, goods richness and user-goods matching are the key.Solution: the older repurchase strategy is using deep discount for conversion. Accepting the insight, top management team change strategy to optimize logistic, customer service, BD more supplier to extend our goods richness and quality.Performance: 5% lift of retention rate, 7% lift of repurchase rate

In-APP User Behavior Analytic to optimize conversion (from DAU to Purchase) by optimizing user path design. Insight and solution:1. Users’ actual core-paths are chaos and not similar with APP designed core-paths, the reason is some unnecessary page will confuse users, and make them “back-and-forth” in APP. Solution: kill the unnecessary page and embed its function into searching function.2. Majority of users will click “nearby” button when they search. Solution: directly highlight “nearby” on the Home-page to shorten users’ path.Achievement: After optimization, 20% lift of conversion rate and the shorten step length of conversion from 12 steps to 8 steps

Role & Responsibilities: Direct report to CEO, leader of BI team. (1) design KPI of the whole business, lead daily/weekly/monthly report; (2) lead building BI system, cooperate with R&D; (3) business analysis and review; (4) data modeling project; Project experience and achievement:1. Commodity Life Cycle Management ProjectProject key value: daily detect commodities that at the end of their life cycle, automatically alert by email, and help business make quick decision to do clearance salesMethod: build model via Hive and Python, daily email alertAchievement: reduce low-sell-through commodity rate 12%, increase shelf turnover rate 25%2. Shelf Dynamic Stock ProjectProject key value: Before this project we manage shelf commodity stock in city level, the problem is, for one single SKU, it would be out-of-stock at high sales shelf, but would be expire at low sales shelf. This project help business to manage shelf commodity stock in shelf level for delicacy managementMethod: build model via Hive and Python, can adjust commodity stock in shelf level for every dayAchievement: reduce 21% of shelf stock, increase shelf turnover rate 25%

Role & Responsibilities: Direct report to region sales director. (1) build and deliver data reporting system and data visualization dashboard, based on business understanding; (2) lead regular business review, in order to dig out key business opportunities and deliver actionable insight, via multi-source data analysis, including: product distribution and sell through, the first moment of truth, promotion and pricing, category management. Project experience and achievement:Y customer sales forecasting model (GBDT + RandomForest)Use Hive and KNIME for ETL of Y customer daily POS data (data granularity: total one billion rows), then use R for forecasting modeling via Model Ensembling with GBDT and RandomForest. We got 91% forecasting accuracy.Achievement: Sales team optimize promotion mechanism by leveraging the forecasting model. And solve the overstock issue since inaccuracy sales forecasting

In EJ, I work in Marketing for 3 years. Seeing that Real Estate would not keep its booming in future 20 years, I resigned and came back to school for a big data related master degree in CUHK, then I started my career path of data analytic