B2C E-commerce Product Sales Forecasting In The Presence Of Promotions

Accurate sales forecasting is a critical prerequisite for B2 C e-commerce platforms to make successful operation plans and conduct efficient supply chain management.With the increasingly fierce competition among B2 C e-commerce platforms,promotions have gradually become a powerful marketing tool for companies to seize more market share.Thus,great importance should be attached to the impact of promotions on product sales.Moreover,the predicted values of promotional sales made by B2 C e-commerce platforms often deviate from the actual values since the product sales trajectory in the presence of promotions are,in fact,unstable,complex and changeable.If the predicted value of promotional sales is less than the actual value,it will increase the delivery time of products and cause unnecessary losses such as out-of-stock costs;if the predicted value of promotional sales is greater than the actual value,higher inventory costs and waste of resources will occur.Therefore,the key problem that B2 C e-commerce platforms urgently need to solve is that how to improve the accuracy of product sales forecasting in the presence of promotions.There are three difficulties that need to be overcome in promotional sales forecasting,which are derived from the analysis of the characteristics of promotions and promotional sales trajectory in B2 C e-commerce.The three difficulties are as follows:(1)B2C e-commerce platforms have a large number of products with different characteristics,and different products respond differently even to the same promotion.Thus,B2 C e-commerce platforms need to ensure the efficiency of the forecasting method while considering the potential impact of the difference between products on the forecast accuracy.(2)It is very common that many products are first-promoted and do not have historical data of promotions.The reason is that products on the B2 C e-commerce platform updates and iterations frequently and there are naturally some old products that have never been on promotion before.It is a huge challenge to forecast promotional sales without historical data of promotions.(3)Promotions in B2 C e-commerce become more and more diversified in form as well as more and more routine in frequency.In this case,products sales trajectories in the presence of promotions often show intermittent,instability,and variability.Therefore,the third difficulty of solving the problem is how to measure the influence of diversified promotional mechanisms on the product sales in the presence of promotions.In terms of the mentioned characteristics and difficulties,this paper mainly carried out the following three aspects of work:(1)An analysis of B2 C e-commerce promotional sales forecasting problems.First,we analyze the characteristics of promotions and promotional sales in B2 C e-commerce.Secondly,we conclude the difficulties of promotional sales forecasting in B2 C e-commerce caused by these characteristics.Finally,a forecasting method is proposed for B2 C e-commerce platforms to forecast product sales in the presence of promotions.(2)A research on the promotional sales forecasting method in B2 C e-commerce.Drawing on the idea of classification modeling,we present a forecasting method that is a combination of classification methods based on promotion sensitivity as well as semantic similarity and a forecasting model based on the LightGBM algorithm.Finally,this paper gives a complete flow chart of the forecasting method.(3)Case study and experiments.First,we verify the effectiveness of the proposed promotional sales forecasting method for B2 C e-commerce by conducting the experiments on the data provided by one of China's largest B2 C e-commerce platforms.According to the results,the proposed forecasting method is more accurate compared with the benchmark forecasting methods.Besides,we also give some corresponding management suggestions through further analysis of the results.The method proposed in this paper can improve the accuracy of B2 C e-commerce product sales forecasting in the presence of promotions,and can provide support for B2 C e-commerce platforms to make scientific operational decisions.The proposed method also meets the actual operational needs of B2 C e-commerce platforms with relatively low costs of data collection,and has a wider range of applicable products.In addition,the product classification method based on promotion sensitivity provides scientific strategies and new ideas for solving sales forecasting problems and demand forecasting problems in other fields.