長久以來，「最低標」一直是政府公共工程採購最常採用之決標方式。在合理的產業環境及市場條件下，這種方式固然可以很有效率的節省公帑，但在當前營建景氣低落及不成熟的產業環境下，廠商在維持生存及同業競爭壓力，採用「最低標」發包的結果往往造成低價搶標、圍標及品質不佳等情事發生。 「合格最低標」之決標機制係以業主底價與廠商標單單價之內容做為決策資訊及依據，制訂出合理單價，並以合理單價做為基準，透過統計方法來評定廠商標單單價之合理性，依此給定每一家投標廠商標單之加權分數，透過分數門檻審核之方式決定最後的得標廠商，分數越高代表合理性越佳，反之，則代表廠商所投之單價欠缺合理性，藉由此決標機制及步驟來幫助業主篩選廠商，降低低價搶標與圍標之情事發生。但「合格最低標」機制仍有缺乏模擬案例測試分析、決標機制效能不顯著及缺乏較為嚴謹及科學之方法分析等缺點。 本研究透過模擬案例測試、無母數常態分布檢定、群集分析法及敏感度分析等方法將「合格最低標」決標機制模型加以檢討改良，並分析其於公共工程採購之適用性及，以建立兼顧業主及市場自由競爭獲益的機制，修正改良後之「合格最低標」具有以下功效及特點： 1.將經過常態分布假設及驗證之工程細項單價，作為機制之決策資訊，有助於提升投標廠商深入瞭解工程內容及詳細檢討工程報價內容之意願，藉以篩選外行者或不深入瞭解工程內容得標機會。 2.以群集分析方法合理計算出廠商代表之市場價格與業主底價，客觀計算出「合理單價」，以作為評比廠商單價之基準。 3.藉統計學之平均值及標準差之精密計算方式來防堵圍標或低價搶標等情事發生。 4.利用統計方法分析並評比投標廠商之單價資訊，建立單價合理性評估之立論基礎。 5.可兼顧業主於合理標價中選取最低價格之主觀意願，保障業主權益。 Low bidding is the most commonly used method in Taiwan for the government procurement of public construction projects. However, due to the fierce and irrational market competition, in the past few years the low bid prices frequently were as low as less than 65% of the budget price. Consequently, more than often the contractor bankrupted in the middle of the project and the public owner had to find a new contractor, or the project was finished with unsatisfactory quality. Although an alternative method, the Most Beneficiary method, is used for public procurement as well, so far the low bidding method is still the most dominate one. The Qualified Low Bidding (QLD) method was developed by Huang et. al. in 2001. The method employs a rating mechanism for the rating of tenders’ bids. Only contractors rated in the first 50% are qualified for entering the low bid. This research is a continuous study of the previous one. It examines the effectiveness of the developed QLD method and attempts to improve its mechanism. The two measures for the effectiveness of the QLD method are (1) the capability to prevent the low bidder always being the winner without any examination and (2) the capability to prevent manipulation of the winner by coalition of a group of contractors. Pseudo cases are employed for testing of the QLD method. Conclusions and findings are reported.