Lecture 2 Flashcards
Digital Experiments:
The process of manipulating one or more independent variables and measuring their effect on one or more dependent variables, while controlling for extraneous variables, that are facilitated by a digital environment.
Why focus on digital experiments?
Digital = fast, real-time, comprehensive, scalable
Jeff Bezos: “Our success at Amazon is a function of how many experiments we do per year, per month, per week, per day.”
Conjoint Analysis
› Experimental method for studying consumer preferences (“utilities”) for products
› Products are perceived as attribute bundles (“attributes CONsidered JOINTly”)
Conjoint methods
- Rating-based, Ranking-based
* Choice-based
Limitations of Rating-based Conjoint
› Can consumers quantify their utility?
› What does a rating of “7” mean compared to a rating of “8”?
› How likely are consumers to order a burger at all?
› Validity: How often do you rate products in real life?
› Alternative: Choices as dependent variable
Choice-based Conjoint (CBC)
› Presents a selection of stimuli and asks for the most preferred option
› Choices as dependent variable
› Choices are natural manifestations of preference; free from interpretation
› Possibility to include a no-choice (“None”) option
› Model allows choice predictions
Stages of the Conjoint Analysis Process
- Study design: Attributes and levels
- Experimental design & choice design
- Implementation
- Model specification and estimation
- Assessment of model fit
- Interpretation of the results & simulations
Rules for Selecting Attributes and Levels
› Try to represent the product (or market) with the selected attributes
› Do not include too many attributes: Usually not more than 6
› Attributes are independent: levels can combine freely with one another
› Do not include too many levels: Usually not more than 5 per attribute
› Levels should have concrete/unambiguous meaning E.g., “expensive” is up to interpretation; €500.- is not
Experimental Designs (Conjoint)
› Experimental design:
• The attribute level combinations shown to respondents
• The independent variable matrix
› Full factorial: All possible attribute level combinations
› Fractional factorial: A subset of the full factorial
Efficient (Conjoint) designs are:
• Balanced – Each level is displayed an equal number of times
• Orthogonal – Each level combination appears an equal number
of times; no correlation between attributes
Number of Alternatives and Choice Sets
› Number of alternatives per choice set, usually 2 to 5
› Randomly allocated, with minimal overlap
› Number of choice sets, usually 8 to 16
› Number of choice sets depends on fatigue effects
› Fatigue depends on complexity of the choice situation
› Need to motivate consumers when using more than 12 choice sets
Random Utility Model
› Choices are based on overall utilities of alternatives
› Utility of consumer n for product i:
U = V + E
with
› V = systematic utility component, rational utility
› ε = stochastic utility component, error term
Utility Function
› Assumption: Goods and services are combinations of attributes
› Consumers attach part-worth utilities to each attribute
› Systematic utility of consumer n for product i is sum of part- worth utilities:
V = sum(B * X)
with
› k = (1, …, K) number of attributes
› x = dummy indicating the specific attribute level of product i
› β = part-worth utility of consumer n for attribute k
Vector Model (Numeric, linear)
Linear relation between increase in level and utility
Ideal Point (Quadratic) model
Two parameters, 1 linear and 1 for the increasing or decreasing marginal returns. Thus creating a ideal point
