相對於傳統的單層連續抽樣計畫，CSP-T是一個較經濟且較嚴謹的多層連續抽樣計畫。對於CSP-T的研究，傳統上是透過馬可夫鏈(Markov Chain)之技術來探討各種績效測度。然而，該方法的評估過程相較於圖形評核術(GERT)而言是較冗長且複雜。因此，本研究採用圖形評核術來處理具退貨成本的CSP-T。首先，我們採用圖形評核術來重建CSP-T的績效測度。其次，更新過程(Renewal Process)的概念被使用來獲得檢驗政策的平均單位成本。透過最小的平均單位檢驗成本，我們決定了最佳的檢驗政策（包括檢驗數與抽樣頻率）。解析的結果顯示，CSP-T有三個可能的最佳檢驗政策（不檢驗、全部檢驗、(i, ƒ)的任何組合），且檢驗政策受到重做成本、退貨成本、檢驗成本與不良率等四個參數的影響。當我們限制重做成本與抽樣頻率時，我們的結果支持戴明博士(Dr. Deming)的結論：假如單位檢驗成本對單位退貨成本的比率小於不良率時，則要全部檢查，否則就不要檢驗。

In contrast to single-level continuous sampling plans, CSP-T, a multilevel continuous sampling procedure, is a more economic and tightened scheme. Traditional performance measures for CSP-T plan had been developed via the technique of Markov chain which is more tedious and complex than the Graphical Evaluation and Review Technique (GERT). The main purpose of this paper is to deal with the problem of designing a CSP-T with return cost by employing GERT. First, the traditional performance measures for CSP-T are re-built using GERT. Second, the concept of a renewal process is employed to obtain the long run average unit cost of the inspection policy. Then, by minimizing the long run average unit cost, we determine the optimal inspection policy (including the optimal clearance number and sampling frequency). Overall, the analytical results indicate that depending on four parameters (the unit rework cost, the unit return cost, the unit inspection cost, and the defective percentage), there are three possible optimal inspection policies for CSP-T: "Do not inspect", "Do 100% inspection", or any setting for (i, f). An interesting result is that by suspending the restriction cost of rework and imposing the sampling fraction, we can easily obtain Dr. Deming's statement: "Do 100% inspection" if the ratio of the unit inspection cost to the unit return cost is less than the defective percentage and "Do not inspect" otherwise.