整合聯合分析法與反應時間在顧客偏好預測之應用

The application of Conjoint Analysis integrated with response time on consumer preference estimation

謝文瑋(Wen-Wei Hsieh)

指導教授 ： 任立中

國立臺灣大學/管理學院/國際企業學研究所/碩士(2005年)

https://doi.org/10.6342/NTU.2005.01255

摘要

Kolter(1994)認為現代行銷策略的三部曲－區隔(segmenting)、目標(targeting)及定位(positioning)，一直都是行銷經理人奉為圭臬的行銷核心概念，雖然這種概念的正確與有效性不容置疑，但是消費者行為的複雜性與異質性卻嚴酷地考驗這種做法能否真正有效地執行，因此如何能最準確估計消費者的偏好結構，就是行銷上首要必須克服的難題。
本研究利用估計個人偏好迴歸模式的新技術－適應性聯合分析法(Adaptive Conjoint Analysis)來取代過去傳統紙筆蒐集問卷的方式，適應性聯合分析法最驚人的功能就是能夠精確地量測受測者填答問卷的花費時間，儘管反應時間的概念雖然在學理上已經存在了有一段時間，但是它在行銷上的運用卻還只能稱得上是萌芽時期，故本研究仍從分析個體偏好的角度出發，以計量尺度的估計個人參數方法，融合潛伏在消費者行為底下的反應時間兩者，探討導致個人偏好結構的不穩定性的來源。
首先本研究建構以對數常態模式為反應時間之分析模式，估計每位受測者的真實反應時間，然後將此估計結果投入傳統聯合分析法下無法解釋個體偏好選擇的估計殘差項，建構出估計殘差項裡能被反應時間所解釋的部分，得出影響預測每位受測者偏好準確度的時間效果，最後探討影響時間效果的可能變數，本研究提出三種干擾變數為影響時間效果的主要因素，檢驗干擾變數與時間效果之間的連動性，藉此完成反應時間能夠在個體參數估計值時提供有價的資訊。
上述研究流程分別對兩種研究產品進行分析，此種作法之目的是為了檢測除了研究自變數以外，其他變數是否也有影響反應時間之可能性，故以產品性質為條件之下，研究干擾變數與時間效果之間關係的異同。

關鍵字

聯合分析法 ； 反應時間 ； 顧客偏好預測

並列摘要

The three core concepts of the contemporary marketing strategy－segmentation, targeting and positioning have been regarded the utmost principle to the marketing manager. Despite its accuracy and effectiveness, the complexity and heterogeneity of consumer behavior have challenging the STP strategy whether it can be put into practice or not. Accordingly, how to measure consumer’s preference accurately has become the most urgent problem to overcome.
This study takes advantage of the latest techniques – Adaptive Conjoint Analysis (ACA) for consumer preference measurment; it has been applied to replace the traditional method, paper-and-pencil query, to collect data. The most surprising strength of ACA is able to record the respondents’ response time in the conjoint analysis experiments. Although proposed theories with respect to response time have been existed for a period of time, it can only be seen as rather new application in the field of marketing. Thus we started from analyzing individual preference, understanding the potential source of instability of individual preference by combining the metric scale in estimating individual part-worth and response time underling consumer’s behavior.
Firstly, the analytical model of response time is constructed by lognormal model to estimate respondent’s filtered response time. Then we regress each respondent’s filtered response time at each order with the residual term generating from Hierarchical Bayesian Conjoint Analysis to obtain the time effect which refers to the part that can be explained by response time within the residual term. Following the time effect is regressed by three variables, conflict level, product familiarity and involvement in order to test the connection between them, drawing inference from the results to prove estimating individual part-worth.
This study investigates two extreme product, notebook and instant noodle, which purpose is to verify besides three independent variables in the experiment whether product nature has impacts on the relationship between specified independent variables and time effect.

並列關鍵字

conjoint analysis ； response latency

參考文獻

1.Marinus E. Haaijer, Wagner A. Kamakura, Michel Wedel (1998), “The Information Content of Response Latencies in Conjoint Choice Experiments”

2.Green, P. E. and V. Srinivasan (1978), “Conjoint Analysis in Consumer Research: Issues and Outlook,” Journal of Consumer Marketing, Vol. 43, pp. 83-91

3.Tyzoon T. Tybejee (1979), “Response Time, Conflict, and Involvement in Brand Choice,” Journal of Consumer Marketing, Vol. 6, pp.295

4.Maclachlan, James M. (1977), “Response Latency: A New Measure of Advertising,” New York: Advertising Research Foundation

5.Tyzoon T. Tybejee (1979), “Response Latency: A New Measure for Scaling Brand Preference,” Journal of Marketing Research, Vol. 16, pp.96-101