How to design a dynamic pricing strategy in smart grid is an important but challenging problem due to the uncertainty of customers' demand and their reaction to the price. Most of the existing pricing strategies depend on the prediction of customers' demand and their reaction to the price, thus are not robust enough due to the prediction which may cause a high cost to the supplier. In this paper, we propose a closed-loop pricing algorithm (CPA), which can largely decrease the cost and result in both costing savings for customers and higher profit for the supplier, and thus achieves a win-win solution. We provide sufficient conditions to guarantee the win-win solution for customers and the supplier. We also prove that CPA can obtain a stable price. Extensive simulations confirm with our analysis and show that the combined cost saving and profit gain of both customers and the supplier can be 76% by the proposed algorithm compared with the prediction-based pricing algorithm.