Intern(Chow Jun Hoe)

 30/4/2025 I discussed the sales forecast method completed yesterday with Supply Chain Executive, and reported in detail the model principle used, the forecast results and the existing deviation problems. During the communication process, Executive made some suggestions, such as changing the time unit from "week" to "day" for analysis to improve the sensitivity of the model and the precision of the forecast.

 29/4/2025 In order to further improve the accuracy and model diversity of sales forecasts, we actively searched online and explored more advanced forecasting methods, including the Multiple Decomposition model. During the learning process, we gained a deeper understanding of how this method separates trends, seasonality, and residuals in sales data. After mastering the basic principles, we began to try to apply Multiple Decomposition to actual sales data for forecasting, preliminarily tested its performance during holiday peaks and promotion cycles, and compared and analyzed it with previously used methods such as Holt and Holt-Winters.

 28/4/2025 Started to formally learn Holt's Method and deeply understood the principle of introducing trend components on the basis of traditional exponential smoothing. In the process of learning, not only mastered its formula setting and parameter adjustment skills, but also tried to apply Holt's method to actual sales data, conduct forecasting tests, and compare and analyze with previous forecasting results. After mastering Holt's method, further explore the more advanced forecasting technology - Holt-Winters Method, which introduces seasonality factors on the basis of Holt's. Subsequently, use Holt-Winters method to conduct forecasting tests on sales data and observe its performance in capturing seasonal patterns and trend changes.

 25/4/2025 In the process of continuously optimizing the sales forecasting method, we actively searched for and learned new forecasting techniques, and successfully found and understood the principles and application scenarios of the Exponential Smoothing method. Subsequently, we began to try to use this method to conduct forecasting tests on sales data of some products and time periods. After completing the phased forecasting content, we reported the overall progress and test results to the Supply Chain Manager. During the reporting process, the manager affirmed the current efforts and also provided new professional suggestions - guiding us to start researching and applying more advanced forecasting methods, such as Holt’s Trend Method, which adds trend factors to the basic exponential smoothing to better capture the upward or downward trend in sales data and further improve the accuracy of the forecast.

 24/4/2025 The improved forecast data report was submitted to Supply Chain Executive again, and it received positive affirmation and recognition in this report, indicating that the current analysis logic, data presentation method and forecast model have reached a relatively ideal standard. On this basis, we began to prepare the briefing content for the Supply Chain Manager the next day, focusing on the selection basis of the forecast method, actual test results and error evaluation. In addition, based on the established forecast model, we further conducted forecast analysis on this week's sales, selected the most suitable forecast method for the current situation, combined with the latest actual sales data, and tried to generate forecast results that are closer to actual needs.

 23/4/2025 Submit the updated forecast data report to Supply Chain Executive for review again, and explain in detail the adjustments made, the changes in the forecast results, and the data basis behind the optimization. After further comments and suggestions from Executive, optimize and organize the parts of the report that are not clearly structured, the data arrangement is not intuitive, or the charts are not clear.

 22/4/2025 Continue to optimize and modify the forecast data. To improve the efficiency and flexibility of data processing, the relevant sales, forecast values, actual sales, and error indicators are integrated into a unified data table through the VLOOKUP function. This method not only makes the data clearer when consulting and comparing, but also greatly simplifies the subsequent data update and modification process.

 21/4/2025 Systematically organize and summarize the completed forecast data, including the results of various forecasting methods, actual sales comparisons, and error assessment indicators (such as MAPE). After the compilation is completed, report the forecast report to the Supply Chain Executive, explaining the analysis logic and forecast results in detail. After listening to the Executive's feedback, modify and optimize the forecast model and data display method in a timely manner based on his suggestions and opinions.

 18/4/2025 Continue to organize sales forecast data, further refine and standardize the presentation of various forecast results, so as to conduct more systematic analysis and comparison. By setting a unified comparison indicator, MAPE, the performance differences of various forecast methods in different time periods or product types are clearly shown.

 17/4/2025 Continue to organize and improve sales forecast data, focusing on improving the data structure clarity and analysis logic. At 2 pm, I attended the Supply Chain team meeting on time to learn about recent information about ecoshop.

16/4/2025 Organize and optimize sales forecast results, present data in a more detailed and clear manner, and improve the readability and interpretability of the report.

 15/4/2025 After completing the previous sales analysis, the Stock on Hand (SOH) data was further integrated into the existing analysis table through VLOOKUP to more comprehensively evaluate the relationship between inventory levels and sales of each sub-category at different time points. Then, continued to deepen the research on sales forecasting methods. This round of research attempts to separate national holidays from non-national holidays, using data from the past two years and only using data from the past year to conduct comparative tests to observe the impact of different data cycles on forecasting results.

 14/4/2025 After completing the comparative analysis of sales data before Hari Raya for two years, the analysis level was further refined by screening and filtering the number of SKUs under each sub-category to ensure the accuracy and reference of the calculation results of the Sales Growth Rate of each sub-category. Specifically, products with insufficient SKUs or too low sales in each sub-category were excluded from the analysis to avoid deviations caused by too few samples, ensuring that only representative SKUs with sufficient sales records during the analysis period were retained. This can more realistically reflect the overall growth trend of the sub-category.

 11/4/2025 Compare the sales data and out-of-stock situations (OOS) of the 13 weeks before Hari Raya and the week before Hari Raya in the past two years to identify the difference between the sales trend before the festival and the sales performance in the last week, and analyze the possible reasons for the change in sales. First, the average sales and average out-of-stock quantity of the 13 weeks before Raya were sorted and calculated separately as a benchmark reference for normal sales trends before the festival. Then, the sales and OOS data of the week before Raya were analyzed separately to observe whether there was a significant decline or abnormal fluctuation.

 10/4/2015 Analyze the average sales and out of stock (OOS) of Week 13 and Week 14, and compare them with the overall average sales and OOS rate of the 13 weeks before Week 13 (Week 1 to Week 13) to identify changes in sales trends and inventory performance.

 9/4/2025 We will continue to study sales forecasting methods in depth. This time, we will focus on distinguishing national holidays from non-national holidays (non-national holidays, such as regional holidays or specific store promotion days) to analyze the actual impact of different types of holidays on sales in a more detailed manner. After completing the holiday classification and impact assessment, we will build a classification forecasting model based on the obtained data and apply it to sales forecasts for national and non-national holidays respectively.

 8/4/2025 According to the feedback from the Supply Chain Manager yesterday, the existing sales forecast model was optimized and improved. After improving the model, the MAPE (Mean Absolute Percentage Error) indicator was used to measure the accuracy of different forecasting methods. Through multiple tests and comparisons, the performance of various forecasting strategies in different time periods was analyzed.

 7/4/2025 Discussed the data analysis results from Week 7 to Week 15 completed last Friday with the Supply Chain Forecast and Promotion Executive, focusing on the relationship between the sales trend (Sales Trend) and out of stock (OOS) in each week. Afterwards, participated in a meeting on the arrangement of goods for the upcoming promotion, and discussed the current inventory level, replenishment plan and promotion strategy with the relevant team. After the meeting, continued to deepen the research on the sales forecasting method, and then discussed the results of this analysis and forecast with the Supply Chain Manager.

 4/4/2025 Continue to analyze the sales data (Sales Quantity) and out of stock (OOS) of Week 7 and Week 13 in depth to further explore the correlation between the two. After completing the comparative analysis of Week 7 and Week 13, based on the forecasting method learned last week, today only the data of the last six months is used for sales forecasting to test the impact of different data ranges on forecast accuracy.

 3/4/2025 Export the sales quantity and out of stock (OOS) data for Week 7 and Week 13 from Qlik Sense, and perform an in-depth comparative analysis of the two weeks’ data.

 2/4/2025 Based on the sales forecasting method learned and tested last week, today we focus on using the most recent year's data for in-depth data analysis to evaluate the impact of the data range on forecast accuracy. By comparing the forecast results obtained using 1 year of data and 2 years of data, we analyze which data range can more effectively reflect the current market trend and improve the accuracy of sales forecasts.

 1/4/2025 Public Holiday