Abstract: The demand forecast is one of the most important parts in supply chains. This paper builds both SARIMA and time series neural network models, and forecasts the demand based on historical sales and order quantities from a Chinese medicine enterprise. The result shows us that when we consider data with sales quantities only, time series neural network model has better forecast ability than SARIMA model. However, when we consider those with both sales and order quantities, it has worse forecast ability than that with sales quantities only. Finally, it presents that the upper enterprise in the supply chain should consider not only the sale quantities of the lower enterprise but also sales quantities, and it suggests that the lower enterprise should rationally and scientifically make an order