隨著全球化的浪潮、亞洲國家的經濟發展，時尚與精品產業在亞洲市場的發展已經漸趨成熟，紛紛以設立子公司或合資的方式營運，這也導致以代理國外品牌在台灣銷售的代理商的生存空間受到了壓縮，本研究透過探討在台灣代理時尚與精品品牌的代理商與品牌商國際合作關係時，如何增加績效，一方面降低代理權被取消的風險，加強雙方合作的關係，一方面發展代理商的核心價值，追求成功的商業模式。 在研究方法上採用個案研究法（Case study），透過次集資料的搜集，包括了產業的介紹和回顧了國際進入策略、交易成本理論和代理理論的相關文獻，佐以個案公司訪談，探討如何提升代理商的績效，分析結果後歸納提出命題。 本研究整理了七個命題如下，代理商可以透過以下方法強化績效： 1. 代理商行銷能力越強。 2. 代理商的財務能力及規模越大。 3. 代理商越有優良的代理績效與信用。 4. 代理商與品牌商的非正式“關係”經營的越成功。 5. 代理商發展代理商專屬的通路品牌越成功。 6. 代理商對於品牌VIP顧客關係的經營、維持越成功。 7. 代理商代理品牌數越多，越能強化代理商的績效。 如此一來，在國際代理關係中不再扮演被動角色，而是能主動尋找機會，有效降低經營的風險且強化代理商的核心價值。 / Due to the globalization and the economic development in Asian countries, the market performances of fashion and luxury brands are boosting in recent years. Many brands choose to open their own branches in Asian countries rather than find a local agent. Local agents help oversea manufactures or brands to distribute the products in local markets, but foreign manufactures often viewed agent contracts as a temporary entering strategy. This is becoming a crucial threat for the agents. How agents strength their performance in order to build their own competitive advantage? This study aims to address this issue. I define the fashion and luxury goods industry by introducing the market segmentation and business models within the industry. And then we reviewed the literature based on international entry strategy, transaction cost theory, and agency theory. Based on the literature, we conducted interviews with five Taiwanese agents in fashion and luxury sectors to discover this thesis with seven hypothesizes. By these seven hypothesis agents can improve the performance: 1. The agent should develop the brand building marketing skill. 2. The agent should enhance financial ability and expand capital size. 3. The agent should build up a better credits and reputation. 4. The agent should develop better “Guanxi” in order to strength the performance. 5. The agent should build it’s own distribution channel brand. 6. The agent should foster the relationship with VIP customers. 7. The agent should represent more brands.