Lorentz Yeung
Worked with Microsoft as Data Analyst / AI Certificate (Harvard University) / MSc in Digital Marketing / GC in Art History / Bachelor in Economics /
Back To Front PageSales Forecast For a Fast-Moving-Fashion Online Shop
Nov 2023
It was a fast-moving-fashion online shop, and I developed a comprehensive sales forecasting model which will forecast 365 days into the future. It was incorporated with the regions festive events, bank holidays, and marketing strategies, and thus has high predictive power. In the demo is a scattered graph. The scattered dots are the actual, while the blue line is the forecast. The shadow areas are the 5% confidence interval.
Optimising the Storage Strategy for the Fast-Moving-Fashion Online Shop
Dec 2023
By integrating with the forecasting model, my storage efficiency analysis significantly enhanced the stocking strategy, leading to a substantial reduction in overstocking costs by 94.6%.
In the demo graph, the red line is the suggested storage level according to the model, and the blue line below is the suggested min stocking level. Actionable insights: The PIC is recommended to keep the stock level within the vertical range.
LLM Classification Model
During my tenure in Microsoft
The model was developed to solve client challenges, notably in managing an extensive collection of over 20,000 keywords. The classifier was designed to organize these keywords effectively, enhancing marketing and sales strategies.
Skills: BERT, Python, Pandas, NLP, data wrangling.
Built Forecasting Models in Microsoft
During my tenure in Microsoft
Built forecasting models for forecasting the SRPVs, clicks, CPC, etc. I shared the experience in a world renowned data science publication after the projects.
Sales Strategy Optimization for Merck
2020
I focused on understanding my client's objectives then I proceeded with: 1] RFM analysis,
2] a K-means clustering model,
3] a classification model to classify the returning purchase date.
In the actionable insights demo, the next purchase date propensity column can be found on the left of the chart.