In this study we consider a CONWIP system studied in Park and Lee [1] in which the processing times at each station follow a Coxian distribution and the demands for the finished products arrive according to a compound Poisson process. The demands that are not satisfied Immediately are either backordered or lost according to the number of demands that exist at their arrival instants. For this system using the results in [1] we develop an approximation method to calculate order based performance measures such as the mean time of fulfilling a customer order and the mean number of customer orders. To test the accuracy of the approximation method, the results obtained from the approximation method are compared with those obtained by simulation. Comparisons with simulation have shown that the approximate method provides fairly good results.