Abstract: Customer satisfaction evaluation in third party logistics (3PL) enterprises is a dynamic, multiple-input, and nonlinear problem. To solve such a problem, factors influencing customer satisfaction are analyzed in the perspective of the 3PL services. Then, to evaluate customer satisfaction in 3PL, a neural networks ensemble is trained and tested with the data of CRC Logistics and used to construct the plan to improve customer satisfaction. Results show that, with this model, the prediction is accurate and good at generalization ability. The model, with a high practical value, can be used to study the main factors influencing enterprise's customer satisfaction and provide the basis for improving customer satisfaction.