Problem Description: Due to the long lead time in the supply chain, Nivea Taiwan faces issues in accurately predicting their sales in advance and adequate the inventory level for their online e-commerce platforms in China. Therefore, our mission with the study from the business perspective is “to optimize the efficiency of the downstream supply chain in Cross-Border E-commerce Department of Nivea Taiwan”. On the other hand, from the forecast perspective, the study focuses on forecasting sales quantity of hero products of Nivea Tmall1 flagship store.
Data Description: Weekly sales data is generated from Tmall on every Monday. There is a total of 118 SKU’s on shelf and 15 hero SKUs contributing the most are selected to be the forecasting time series. The time series contain 60-week data from November 1, 2017 to December 23, 2018.
Data Preparation: Data has been rearranged from the original sales report to another source file for R programming. Dummy variables are created to indicate promotional periods. Averaging is applied to the weeks of “Double 11” to remove outliners occurred in big promotion periods. Sales quantity and dummy variables are aggregated from a weekly basis to 8-week basis.
Forecasting Solution: Naïve forecasts are used as the benchmark. Different forecasting methods are tested, including Simple Exponential Smoothing, Linear regression with External Information, Linear Regression, Moving Average, ARIMA, and ENSEMBLE. Considering the desirable over-forecasted results and the performance measures, ENSEMBLE model, averaging the forecasts among SES, MA and ARIMA, is used as the final forecasting solution.
Future Forecasts: Based on the ENSEMBLE model, forecasted sum of 8-week sales quantity
from week 52 of 2018 to week 7 of 2019 for “80105”, “81288”, “83807”, “83921” and “83922”
2 are 1550, 218, 3209, 1199 and 482 pcs respectively.
Conclusions: This forecasting project not only provides indicators of an appropriate timing to place an order to distributor, but also allows managers to adjust marketing strategies base on supply and demand. Based on the assumption of sufficient inventory in distributor’s warehouse, managers should keep an eye on the upstream supply chain before adopting this forecast. It is recommended to maintain sales data with high quality for further studies.