ADBUDG :- A decision calculus model for the
advertising budgeting decision. The model assumes that there is a fixed upper
limit of response to saturation advertising, and it also assumes that there is
a fixed lower limit to response under no advertising for an extended period.
Within this range, increases in advertising spending increase response, and
reductions in advertising spending lead to a decay in response over time. The
model's parameters are calculated using subjective responses to a series of
point-estimate questions concerning the likely impact of various advertising
spending decisions (Little 1970). The effectiveness of the model's use has been
discussed by Chakravarthi, Mitchell, and Staelin (1981) and Little and Lodish
(1981).
ASSESSOR:- A model for predicting the market share of a new
frequently purchased product using pretest market information. Perceptions and
preferences of potential customers are measured via interview and a simulated
shopping experience conducted at a central location. The prediction is based on
the sample participants' reaction to advertising (exposure to advertisements
for several brands), estimated level of product trial (based on the simulated
shopping experience), estimated repeat purchase level (via follow-up
interview), and brand preference judgments (Silk and Urban 1978). Evidence on
the model's predictive validity has been reported by Urban and Katz (1983).
Beta Binomial
Model:- A probability mixture model
commonly used to represent patterns of brand choice behavior or media exposure
patterns. The model assumes that each individual's behavior follows a Bernoulli
process. That is, an individual performs some behavior of interest (e.g.,
buying a brand of interest) with probability p on each possible opportunity
(e.g., occasion on which a purchase is made from the product category of
interest). The number of times that the behavior of interest is exhibited, out
of a given number of opportunities, has the binomial distribution for any
individual. The model further assumes that the probability values p vary across
individuals according to a beta distribution (Greene 1982; Massy, Montgomery,
and Morrison 1970). A generalization of this model to represent choice among
more than two items is termed the Dirichlet multinomial model. These models are
used to predict future brand choice or media exposure patterns based on
individuals' past behavior.
BRANDAID:-
A decision support system for determining the marketing mix for a
particular brand. The model has submodels dealing with advertising spending
level, price, and salesperson effort (i.e., dollars per customer per year). Its
parameters can be calibrated by combining historical data (e.g., on sales,
market share, advertising spending, etc.) with structured subjective judgments
(Little 1975).
