Thanks to the fast development of Internet technologies, e-commerce is taking more and more market shares from physical store sales. This emerging retail model also affects the existing supply chain and logistics services, e.g., alongside of the warehouse-to-store shipping in the traditional logistics model, more courier services including warehouse-to-customer and store-to-customer shipping are in compelling need for online sales. Sales volume forecasting is important in supply chain management. Since e-commerce introduces higher workload for logistics services, sales forecasting in e-commerce becomes even crucial. Traditional sales forecasting methods, which focus on physical stores, are no longer effective, because e-commerce brings in more factors such as the Internet capacities that affect the sale volume. In this paper, we aim to design a new causal model for e-commerce sales forecasting. Specifically, we identify the particular factors affecting e-commerce sales and take them into account to build causal-based forecasting models. We use the US e-commerce data in our study, and compare the causal model with the standard time-series analysis model to show the performance difference. We also discuss the trade-off of the two classes of models in utility. Our research can be a guideline for the supply chain management in the e-commerce industry for sales forecasting.