A Regression Approach to Penalty Analysis to Assess the Relative Importance of JAR Attributes

• Activity Number: 357
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Consulting
• Author(s): Jason Parcon*+
• Companies: PepsiCo
• Keywords: Regression ; Just-About-Right Scales ; Penalty Analysis ; Consumer Research ; Sensory Science ; Multiplicity

Abstract

An important aspect of product development is to verify if a prototype is perceived by consumers as having a "just-about-right" (or JAR) level of an important attribute or if it has "too much" or "not enough" of it. It is assumed that target consumers penalize a product for not achieving JAR, and the penalty is in terms of a drop in overall liking. In particular, it is the difference between the mean liking scores from those who perceived the prototype as JAR versus having "too much" or "not enough" of an attribute. A t-test is then conducted to assess the statistical significance of the penalty. In a typical consumer test, JAR assessments are made on several attributes, and penalty analysis is performed on each attribute. The individual analyses pose two problems: (1) the possibility of multiplicity and (2) the relative importance of the various JAR attributes on liking is not measured. This paper proposes a regression procedure that addresses the above-mentioned issues. Examples from real data showed the current version of penalty analysis leads product developers to focus on attributes that may be relatively unimportant. The proposed method provides a clear differentiation between important and unimportant JAR attributes.