台灣的企業分析促銷折扣、行銷資料時，並沒有規劃隨機實驗的習慣，因此會直接以蒐集到的交易資料，即觀察性資料來進行分析。然而在觀察性資料之下，消費者對促銷折扣會有選擇性偏誤，並沒辦法真實評估促銷折扣的效果。若要了解促銷折扣的效果則要透過隨機實驗或是以統計方法調整資料。雖然隨機實驗在國外的商業分析使用頻繁，然而考量到隨機實驗所需要的背景知識以及成本，本研究將以配對方法中的傾向評分配對方法討論是否可以透過此方式了解台灣企業在促銷折扣上的效果。傾向評分配對方法是一種用於改善觀察性資料之選擇性偏誤的統計方法，並且在文獻中，已被醫療、公共衛生、政策評估領域等廣泛應用。在企業實務上使用傾向評分配對方法同樣可以利用現有的交易資料進行分析。
本研究以台灣連鎖零售企業A公司在2021年的交易資料進行資料整理，討論傾向評分配對方法是否能夠讓企業看到行銷的因果關係。然而在實際使用實務資料進行傾向評分配對方法時，會遇到一些變數取得、環境干擾和方法適用性的問題。本研究發現在行銷領域，特別是在台灣的企業中，時間是影響消費特徵、衡量行銷效果的重要因素。在計算消費者接收到行銷之前的消費特徵、計算接收到行銷後的行為改變都需要更精確的時間來衡量。
因此本研究提出一個一時間為基礎的配對方法，希望能改善直接使用傾向評分配對方法所遇到的問題。此方法找出同一天有消費的顧客中，有收到行銷及沒有收到行銷的實驗組與配對組，在三十日內的回購率有何不同。以日期為配對的條件使整個配對過程更為合理且實際。配對過後，一樣可以將資料進行應用，其中包含以邏輯迴歸預估顧客未來的消費行為、進行成本效益分析來挑選顧客，也能夠延伸作為衡量配對好壞的參考。透過本研究的結果，未來台灣企業在分析行銷資料時，可以將日期作為配對的條件，來了解過往所進行的行銷方案所帶來的效果，使分析結果可以有更深入的討論。

In Taiwan's business analysis, companies seldom conduct randomized experiments when analyzing promotional discounts or marketing data. Therefore, analysis is often performed directly on the transaction data or observational data. However, under observational data, there is a potential for self-selection bias, making it difficult to accurately evaluate the effects of promotional discounts. To understand the effects of promotional discounts, randomized experiments or statistical adjustments are necessary. While randomized experiments are commonly used in business analysis abroad, considering the background knowledge and costs associated with conducting such experiments, this study will explore whether the Propensity Score Matching (PSM here after) method, a matching method within the framework of observational data, can be used to understand the effects of promotional discounts in Taiwan's business context. The propensity score matching method is a statistical technique used to remove selection bias in observational data and has been widely applied in various fields such as healthcare, public health, and policy evaluation. Using the propensity score matching method in business practice allows for analysis using existing transaction data.
Using transaction data from Company A, a Taiwanese chain retail enterprise in 2021, this study discusses whether the propensity score matching method enables the company to observe causal effects in one of its marketing instruments. However, when we tried to apply PSM to the data we ran into problems in obtaining matching variables and excluding interference from the other marketing activities. The study finds that “time” is an important factor influencing consumer characteristics and measuring marketing effects in the marketing field, especially in Taiwanese enterprises. Accurate measurement of consumer characteristics prior to marketing exposure and behavioral changes after marketing interferences requires timely assessment.
Therefore, in this study we proposed a time-based matching method that might help to address the limitations encountered when using propensity score matching alone. This method identifies experimental and control groups within customers who made purchases on the same day, distinguishing between those who received marketing and those who did not, and examines the differences in repurchase probabilities within thirty days. Using date as a matching criterion enhances the overall validity and practicality of the matching process. Once the matching is completed, the data can be utilized for various applications, including logistic regression for predicting future consumer behavior, cost-benefit analysis for customer selection, and serving as a reference for assessing the quality of the matching. The results of this study provide insights for Taiwanese businesses in analyzing marketing data, emphasizing the inclusion of date as a matching criterion to gain a deeper understanding of the effectiveness of past marketing campaigns and facilitate more comprehensive discussions in the analysis results.

論文審定書 i
致謝 ii
中文摘要 iii
英文摘要 iv
目錄 vi
圖目錄 viii
表目錄 ix
第一章 緒論 1
第一節 研究動機 1
第二節 研究目的及架構 1
一、研究範圍 2
第二章 文獻探討 1
第一節 因果關係 1
第二節 傾向評分配對方法 4
第三章 研究方法 1
第一節 傾向評分配對使用方式 1
第二節 資料來源與處理 2
第三節 資料探索性分析 2
第四節 配對方法的處理 6
第四章 研究結果 9
第一節 傾向評分配對方法的處理 9
第二節 傾向評分配對方法的使用限制 14
第三節 以日期為配對基礎的DMATCH配對方法 16
第四節 改善後的配對處理 21
一、比對特徵差異 21
二、改善後的配對結果 24
第五節 配對資料的應用 25
ㄧ、建立預測模型 25
二、成本效益分析 31
三、比較兩種配對模型 33
第五章 結論與建議 39
第一節 研究結果 39
第二節 研究限制與未來研究方法 40
參考文獻 42

Almus, M., & Czarnitzki, D. (2003). The effects of public R&D subsidies on firms' innovation activities: the case of Eastern Germany. Journal of Business & Economic Statistics, 21(2), 226-236.
Austin, P. C., Yu, A. Y. X., Vyas, M. V., & Kapral, M. K. (2021). Applying propensity score methods in clinical research in neurology. Neurology, 97(18), 856-863.
Barajas, J., Kwon, J., Akella, R., Flores, A., Holtan, M., & Andrei, V. (2012). Marketing campaign evaluation in targeted display advertising. In Proceedings of the Sixth International Workshop on Data Mining for Online Advertising and Internet Economy (pp. 1-7).
Burk, S. (2006). A Better Statistical Method for A/B Testing in Marketing Campaigns. Marketing Bulletin, 17.
Cunningham, S. (2021). Causal inference: The mixtape. Yale university press.
Dehejia, R. H., & Wahba, S. (2002). Propensity score-matching methods for nonexperimental causal studies. Review of Economics and statistics, 84(1), 151-161.
Gutierrez, P., & Gérardy, J. Y. (2017). Causal inference and uplift modelling: A review of the literature. In International conference on predictive applications and APIs (pp. 1-13). PMLR.
Heckman, J. J., Ichimura, H., & Todd, P. (1998). Matching as an econometric evaluation estimator. The review of economic studies, 65(2), 261-294.
Heinrich, C., Maffioli, A., & Vazquez, G. (2010). A primer for applying propensity- score matching. Inter-American Development Bank
Kohavi, R., & Longbotham, R. (2017). Online Controlled Experiments and A/B Testing. Encyclopedia of machine learning and data mining, 7(8), 922-929.
Lechner, M. (1999). The effects of enterprise-related training in East Germany on individual employment and earnings. Annales d'Economie et de Statistique, 97-128.
Moser, P. (2005). How do patent laws influence innovation? Evidence from nineteenth-century world's fairs. American economic review, 95(4), 1214-1236.
Pearl, J., & Mackenzie, D. (2018). The book of why: the new science of cause and effect. Basic books.
Randolph, J. J., & Falbe, K. (2014). A step-by-step guide to propensity score matching in R. Practical Assessment, Research & Evaluation, 19.
Rosenbaum, P. R., & Rubin, D. B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70(1), 41-55.
Rubin, D. B. (1973). Matching to remove bias in observational studies. Biometrics, 159-183.
Rubin, D. B., & Waterman, R. P. (2006). Estimating the causal effects of marketing interventions using propensity score methodology. Statistical Science, 206-222.
Smith, J. A., & Todd, P. E. (2005). Does matching overcome LaLonde's critique of nonexperimental estimators?. Journal of econometrics, 125(1-2), 305-353.
Trujillo, A. J., Portillo, J. E., & Vernon, J. A. (2005). The impact of subsidized health insurance for the poor: evaluating the Colombian experience using propensity score matching. International journal of health care finance and economics, 211-239.
Varian, H. R. (2016). Causal inference in economics and marketing. Proceedings of the National Academy of Sciences, 113(27), 7310-7315.
Wei, J. T., Lin, S. Y., & Wu, H. H. (2010). A review of the application of RFM model. African Journal of Business Management, 4(19), 4199.
Wooldridge, J. M. (2006). Introductory Econometrics-A Modern Approach. Thomson South-Western. International Student Edition.
Xu, Y., Chen, N., Fernandez, A., Sinno, O., & Bhasin, A. (2015). From infrastructure to culture: A/B testing challenges in large scale social networks.
In Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (pp. 2227-2236).
Huang, J., & Ling, C. X. (2005). Using AUC and accuracy in evaluating learning algorithms. IEEE Transactions on knowledge and Data Engineering, 17(3), 299-310.
關秉寅, & 李敦義. (2008). 補習數學有用嗎? 一個 [反事實] 的分析. 臺灣社會學 刊, 41, 97-148.
黃玉麗, 張元, & 沈中華. (2012). 不請自來的銀行信用評等有向下偏誤的現象 嗎?-配對方法. 經濟論文叢刊, 40(2), 149-188.