Conducting a Basic Self-Explicated Conjoint Analysis Online With Qualtrics®

Conducting a Basic Self-Explicated Conjoint Analysis Online With Qualtrics®

DOI: 10.4018/978-1-5225-8563-3.ch004

Abstract

A recent feature in the Qualtrics® Research Core Platform 2018 (or Qualtrics Research Suite) is a basic self-explicated conjoint analysis, which is a research method to understand respondent preferences in a real-world context with limited available features and selection tradeoffs at respective price points. This chapter will introduce the basic self-explicated conjoint analysis tool and how to design questions for this, how to deploy the conjoint analysis (as either part of a larger survey or as a stand-alone survey), and how to analyze and use the resulting data. This chapter will describe the assertability of the findings based on the back-end factorial statistical analysis and suggest ways to explore beyond the initial conjoint analysis.
Chapter Preview
Top

Introduction

On a typical day, a person may make hundreds of decisions based on his or her conscious, subconscious and unconscious preferences. These decisions may be mundane, but they may also be surprisingly persistent; taken together, individual decisions may have personal impacts, including larger-scale emergent ones. Individuals and organizations have an interest in eliciting client preferences, so they may provide the products and services that others want. There are a certain class of survey instruments that enable choice experiments, through “choice modeling,” a term of art that collectively refers to “choice experiments, contingent ranking, contingent rating, (and) paired comparisons” (Hanley, Mourato, & Wright, 2001, p. 438), among others. “Choice experiments” require respondents to “choose between two or more alternatives (where one is the status quo)”; “contingent ranking” research requires respondents to “rank a series of alternatives”; “contingent rating” research requires respondents to “score alternative scenarios on a scale of 1 – 10”; “paired comparisons” require respondents to “score pairs of scenarios on similar scale” (Hanley, Mourato, & Wright, 2001, p. 438).

One basic and popular form of choice experiments research is the “conjoint analysis” (CA), in which respondents make selections of particular attributes or packages of attributes based on their own preferences about a particular product, service, or decision space (in real-world and practical contexts and sometimes in theorized ones). “Conjoint analysis” is a collective term, “covering both the theory and methods of a variety of different paradigms that can be used to design, implement and analyse (sic) judgment data experiments” …or “evaluative rankings or ratings of a set of multi-attribute alternatives” (Louviere, 1988, pp. 94 – 95). The sets of multi-attribute alternatives may be constructed using experimental or quasi-experimental design techniques.

Conjoint analyses come in various types. These variants are…

… based on the way preference scores are elicited (e.g., ratings, rankings, self-explicated, constant sum, choice), the type of designs used (e.g., full factorial, fractional factorial, adaptive), the type of models estimated (e.g., regression, logit, probit, hierarchical Bayes), and the estimation procedures employed to make inferneces (e.g., maximum likelihood, Markov Chain Monte Carlo) (Ding, Grewal, & Liechty, 2004, p. 1).

One classification of types is by those in a conjoint analysis suite in Qualtrics, which include the following: self-explicated conjoint analysis, choice-based / discrete choice, adaptive choice-based, menu-based, and MaxDiff (“What is a conjoint…?” 2018). Other types differentiate between classic conjoint analyses and adaptive ones, which are computerized and adapt to the feedback from the survey participants. “Conjointedness” refers to the “combining of all (factors) involved,” so preferences in a constrained practical environment become clearer.

A basic self-explicated conjoint analysis in Qualtrics enables the building of a conjoint module that presents various specific attributes (features, dimensions) of a product, service, or choice space; the attributes are optimally “orthogonal,” with no overlap with other features (so the analysis may be discriminative between attributes and features) and sufficiently sensitive), and survey respondents select their best-worst options among the subsets, and their in-between preferences (between their defined best-worst preferences) in a stepwise approach. This research method helps explicitly decompose the respective variables that affect people’s decision-making in selecting particular products, services, or policies. One research team describes conjoint analysis as a decompositional method:

CA is a decompositional method which asks for general judgments on alternatives (stimuli, products) which are decomposed into part-worths for single attribute levels. In any version of CA, the DM (note: decompositional method) has to evaluate tradeoffs between different levels of several attributes in order to decide whether an alternative A is better, equivalent or worse than another alternative B and sometimes additionally how strong this preference is. (Scholl, Manthey, Helm, & Steiner, 2005, p. 766)

Key Terms in this Chapter

Attribute: A feature.

Self-Explicated Conjoint Analysis: A basic form of choice experiment in which respondents explicitly define their preferences from lists of attributes that comprise a particular product or service or real-world and/or theoretical choice-space.

Part-Worth (Attribute Importance Scores, Level Values, Utility Score, Part Utility): A utility measurement (“util”) that shows a weighted preference for a particular attribute (or factor or dimension) in a product, service, or choice-space.

Conjoint analysis: A type of choice-based experiment in which respondents indicate their preferences among a selection of attributes related to a particular choice-space (product, service, or other real-world and/or theoretical decision space).

Conjointedness: The combination of all factors involved.

MaxDiff (Maximum Difference Scaling, Best-Worst Scaling): A paired comparison method in which respondents identify their best and worst attributes among a set, which reveals a number of other ranked preferences among paired sets.

Complete Chapter List

Search this Book:
Reset