Eric Marder, Eric Marder Associates, Inc.
Canadian Journal of Marketing Research, Volume 18, 1999 (http://www.ericmarder.com/articles/cjmr1.pdf)
Objectives
- Examines core assumptions of choice modelling, specifically conjoint analysis and SUMM (Single Unit Marketing Model)
- Explores whether, and to what extent, assumptions of conjoint analysis and SUMM are maintained in existing empirical data of available literature
Methods
- draws on empirical evidence that is available in literature
Results and Conclusions
- “There are psychological judgments that are not amenable to modelling of any kind, either by conjoint analysis or by SUMM, because the choices cannot be partitioned. These cases can be studied only in their entirety, by choice experiments in which each respondent is exposed to a single test stimulus embedded in a full competitive frame. There are other cases in which the psychological total is an aggregation of parts — only these cases are amenable to modelling. The direct measurement assumption holds that, in these cases, our nervous system arrives at overall judgments by performing precisely the kind of summing that is simulated in SUMM, and that direct measurement rather than decomposition is more realistic." (p. 7)
- If goal is product development (“What characteristics should I build into my product to increase the probability that people will buy it?” Choice model can tell what characteristics should be of a real, objective product with specific characteristics.
- “SUMM usually provides for measuring each respondent’s beliefs, and then defines the value of a brand for a particular respondent as the sum of the values of those characteristics which that respondent, correctly or incorrectly, believes the brand to possess.”(p. 8)
- choice models must be evaluated on 2 criteria:
1) capacity: SUMM has advantage
* simpler easier and faster to administer
* larger number of characteristics
* can accommodate both special case when desired to study only objective characteristics and general case when “characteristics” are beliefs
2) predictive power: CA has advantage
* greater realism (not self-evident or supported by empirical evidence)
- no compelling reason to give up SUMM’s simplicity and capacity for “realism” in CA
Assumptions
- First Law of Choice = Assumes that choice experiment provides the same competitive frame, accessibility and respondent information as an actual market test situation, choices (market shares) in the test situation are the same as the market
Opinion
The author’s references seem a little dated, but vast including some conference proceedings. Much of his own work regarding SUMM is included which made me question the bias of his conclusion. Ultimately he did not achieve his stated objectives since there was not enough compelling empirical evidence. However core assumptions of choice modelling, specifically conjoint analysis and SUMM (Single Unit Marketing Model) were discussed in depth.
As I mentioned the author’s abrupt conclusion that SUMM was superior to CA made me wonder he has a product that uses SUMM, especially since he is associated with his own private business versus academia.
This article has only been cited by one person and it was listed as a reference under the Wikipedia page for “Conjoint Analysis.”
I will consider SUMM, but look for other methodologies that can answer the question of whether chefs prefer local vs. foreign or domestic import and aquaculture vs. wild. Maybe we need to ask implicitly, but the reality is they are subject to the availability of suppliers, so some type of marketing experiment is required. Perhaps it is CCE.
Useful Information - Choice Experiments
* STEP
_ each page of a booklet devoted to different brand (seafood company) shows: [picture, price, brief statement summarizing brand's principal benefits
_ participant attaches one of 10 adhesive labels (saying what?) to each brand
_ 2 or more randomly equivalent groups receive booklet, but "test" brand is systematically different
_ Requirements: 1) respondent exposed to all brands in frame, 2) test brand is not singled out, 3) each respondent is exposed to single test stimulus, and 4) respondent makes some choice among brands of competitive frame
- Choice Modelling
* Answers change in market share if variables or characteristics changed
* Not an end in itself, but rather a screening device and should be confirmed by choice experiment
* used because number of variables is too large for choice experiment
* Criteria for assessing choice model: 1) capacity or flexibility in allowing study of large number of variables, 2) predictive power of actual choice experiment if one were possible
* Major assumption of choice modelling (and conjoint analysis and SUMM) = value of a product is aggregation of values of characteristics
- Conjoint Analysis (CA) vs. Single Unit Marketing Model (SUMM)
* both begin with "map" of characteristics which includes characteristics likely to influence a respondent's choice
* difference in nomenclature (ex. CA has attributes and attribute levels where as SUMM has SUMM topics and attributes, referred to here as dimensions and characteristics)
* both use individual analysis (versus aggregate level)
* value of each respondent of each characteristic in the map is measured or computed => each brand is credited with an aggregation of values of characteristics
* end result is brand that is most valuable to a respondent allowing market share for each brand to be known
* CA measurement
_ major focus is estimating value to each respondent of various characteristics
_ "profiles" are constructed each representing one possible combination of characteristics
_ respondents rank, rate or choose among pairs or profiles
^ ful-profile method: rate all possible profiles (2^5 = 32 for 5 dimensional model), works well if less than six dimensions)
* SUMM measurement
_ measures characteristics directly as opposed to using overal ratings of products
_ weighs pros and cons (Ben Franklin in 1772 coined it "moral or prudential algebra (Dawes & Corrigan, 1974)) by assigning weights and using the sum to evaluate options
_ also referred to as multi-attribute utility model, subjective evaluation model and multi-criteria model (Huber, 1974)
_unbounded write-in scale allows respondent to write unlimitied amount of Ls or Ds to report strength of likes or dislikes (Marder 1997) allowing for measurement of importance weights across dimensions in a single step
^ original "self-explicated" approach: respondents first rated all characteristics of a dimension, assigning +10 to most preferred characteristics and then rated all other characteristics in relation to this; after this 50 labels were assigned to top characteristics yielding an importance weight; value obtained by multiplying rating by importance weight of dimension
^ value of a brand defined by sum of values of characteristics respondent believe a brand possesses, used by changing respondent beliefs in computer to determine resulting share gains or losses
- Conflicting assumptions between CA and SUMM
* Decomposition versus direct measurement
_ CA assumes people cannot directly report relative value of different product characteristics (decomposition)
_ SUMM assumes people can directly report relative value provided measurements are made properly (direct measurement)
* objective reality versus beliefs
_CA assumes products (brands) can be described using objective characteristics regardless of beliefs of individual (objective reality
_ SUMM assumes that a product’s (brand’s) depend on individual beliefs of brand about subjective and objective characteristics
- Interaction effects: Green (1984) shows empirical evidence that models that include interaction terms have lower predictive validity and provide negligible benefit despite increased realism
Useful References
Johnson, R.M. (1974) Trade-off analysis of consumer values. Journal of Marketing Research, 11 (May): 121-7.
Anderson, J.C. & N. Donthu (1988) A proximate assessment of the external validity of conjoint analysis. AMA Educators’ Proceedings, G. Frazier, et al., eds. Series 54, Chicago: American Marketing Association: 87-91.
Green, P.E. (1984) Hybrid models for conjoint analysis: An expository review. Journal of Marketing Research, 21 (May): 155-9.
Leigh, T.W., D. MacKay, & J.O. Summers (1984) Reliability and validity of conjoint analysis and self-explicated weights: A comparison. Journal of Market Research, 21 (Nov): 456-62.
Lenk, P.J., W.S. Desarbo, P.E. Green, & MR Young (1996) Hierarchical Bayes conjoint analysis: Recovery of part-worth heterogeneity from reduced experimental designs. Marketing Science, 15(2), 173-91.
Srinivasan, V. (1996) Conjoint analysis and the robust performace of simpler models and methods. American marketing Associaion’s Annual Marketing Research Conference, September.
Huber, G.P. (1974) Multiattribute utility models: A review of field and field-like studies. Management Science, 20: 1393-402.
Bass, F.M., E.A. Pessemier & D.R. Lehmann (1972) An experimental study of relationships between attitudes, brand preference, and choice. Behavior science, 17 (Nov.): 532-41.
Wilkie, W.L. & E.A. Pessemier (1973) Issues in marketing's use of multi-attribute attitude models. Journal of Marketing Research, 10: 428-41.
No comments:
Post a Comment