便利商店新設店址評估與來客數預測之研究

Abstract

連鎖式便利店之展店是各連鎖通路爭取市場佔有規模的重要戰術之一，以往各業者均自有一套評估技術，但經市場環境變遷及消費習性轉變，昔日判斷經驗、客源影響變數及分析演算方法得到的預估值與營運實績之乖離率日趨擴大。有鑑於此，本研究重新修正分析流程，對問題背景再透析、資料清洗與預處理、特徵萃取與選擇、統計模型與訓練測試、模型結果解析、驗證與優化、完成展店決策新模型與策略。本研究以某便利商店 (統稱A超商) 歷年展店數值為基礎，運用一般迴歸、決策樹與多項式迴歸等方法進行特徵因子篩選與來客數預測建模，並以108年新開店來客數進行模型驗證，最後針對預測結果，提出未來A超商在展店流程中可以改善之要點，包括1) GIS系統中缺少一套能夠主動搜尋、演算潛力商圈之系統；2) 開發專員須持續不斷教育有效商圈踩點之定義；3) 開發專員促進基盤物件成熟後才做踩點並呈報主管，開發效率較低；4) 踩點表之模型易失真 (人為操作)，開店成功率無法持續上昇。

The expansion of chain convenience stores is one essential tactic to gain direct market share, and each corporation has its convention for the site assessment. Due to the rapid evolutions in the market and consumer habits, the customer flow estimated by the empirical experience and customer-based factors deviates more and more from the practical numbers. In this regard, this thesis renovates the analytical procedures conducted on the data collected from the site assessment and investigation. Consequently, the strategy for evaluating feasible store expansions is analyzed and renewed. Generalized linear models and decision trees are employed to filter the critical factors and fit the interactive model. The model is further validated with the real data of stores opened in 2019. Four significant findings are derived after the case study: 1) the Geographic Information System (GIS) currently in use shall be integrated with the ability to search and calculate the potential districts actively; 2) the staffs of site assessment and investigation should be reevaluated with their jobs contents; 3) the staffs often wait for the sufficient collection of site information and thus miss the expanding opportunity; 4) the model of customer flow prediction is often intentionally manipulated, and the success rate of opening a store is not sustainable.