本研究探討國防部福利總處下轄福利品供應站之經營效率與標竿學習測度，研究方 法擴展原有傳統資料包絡分析（Data Envelopment Analysis, DEA)方法，並整合關鍵(投 入、產出3測度及背景依賴（context-dependent) DEA模式。研究結果顯示：（1)福利站之 總體技術無效率主要歸因於純技術無效率，未來應優先於規模效率前作提升改善；（2)福 利站純技術效率値與服務滿意度指數呈顯著高相關；（3)「產品成本」及「硬體設施滿意 度指數」分居關鍵(投入、產出）因子優先改善目標；（4)在不同層級背景値評估水準下， 吸引力測度均顯示「新營」福利站最具競爭力，而進步測度顯示「台東」福利站績效最 待改善；（5)背景依賴DEA模式確可建構無效率福利站逐步改善之標竿學習策略，並定 義最佳競爭力的福利站；本研究提供DEA模式輔助軍方決策參考之另一例證。

This paper aims to explore the operating efficiency and the benchmark-learning roadmap of outlets for the Taiwan's General Welfare Service Ministry. This study also presents an extension to Data Envelopment Analysis (DEA) by incorporating traditional DEA, critical input / output measure, and context-dependent DEA. Several empirical results are shown: (1) the overall technical inefficiencies of outlets are primarily due to the pure technical inefficiencies rather than the scale inefficiencies. This also suggests that managers should focus on removing the pure technical inefficiency of outlets, before improving their scale efficiencies; (2) the service-satisfaction levels do have a very significant influence upon a outlet's performance; (3) the critical input / output measures indicate that the 'cost of products' and 'facilities-satisfaction index' are the critical input measure and the critical output measure respectively for most inefficient outlets; (4) the attractiveness measure shows that the Hsinying is the most attractive outlet, i.e. global leader, no matter which evaluation context is chosen, and the progress measure shows that Taitung outlet is the worst; (5) the context-dependent DEA successfully draws the outlet' benchmark-learning roadmap to improve the inefficient outlets progressively and it can identify the best outlet. The potential applications and strengths of DEA in assessing the military outlets are highlighted.