To balance inventory cost with diverse demand, an optimal investment decision on necessary process improvement for delayed product differentiation is studied. A two-stage flexible manufacturing system is modeled as a continuous time Markov chain. The first production stage manufactures semi-finished products based on a make-to-stock policy. The second production stage customizes semi-finished products from the first production stage on a make-to-order policy. Various performance measures for this flexible manufacturing system are evaluated by using matrix geometric methods. An optimization model to determine the level of investment on process improvement that minimizes the manufacturer's total cost is established. The results show that, a higher investment level can reduce both the expected customer order fulfillment delay and the expected semi-finished products inventory. When the initial order penetration point is 0. 4, the manufacturer's total cost is reduced by 15. 89％ through process investment. In addition, the optimal investment level increases with the increase in the unit time cost of customer order fulfillment delay, and decreases with the increase in the product value and the initial order penetration point.