Você está na página 1de 6

KATHRYN E. A. VILLANI AND DONALD G.

MORRISON*

Because of inflation and supply shortages, managements Increasingly have


had to consider product reformulation as a means of keeping production
costs down and retail prices competitive. An analytical approach rs presented
for estimating the market sizes for current and proposed nev/ formulations
of a product or brand. The analysis suggests fundamental issues that should
be considered in the design and interpretation of research relevant to the
reformulation decision, such as taste tests, announcement of the formulation
change, and simulated shopping situations.

A Method for Analyzing New Formulation


Decisions

INTRODUCTION If SO, will they like the change? Often taste tests
The decision to change the fortnulation of a product are used to help answer such questions. Ultimately,
or brand is a major otie faced by many marketing however, the decision-maker needs an estimate of the
managers. In particular, today's inflation and supply market size for the new formulation in order to evaluate
shortages have put more pressure than ever on man- its expected profitability in comparison with that of
agements to reformulate as a means of keeping pro- the present product (and its now costly ingredients).
duction costs down and the retail price of their output Cost consciousness is not the only impetus for
competitive. product reformulation decisions. Business often is
Such a situation is common in the food industry confronted with the question of how to make a
where the prices of basic ingredients fluctuate widely currently successful but "old-fashion" product "mod-
as a result of crop yields and alternative uses and ern." For example, publishers must decide when and
demand for grains. For example, for a product to how to update textbooks as new knowledge is ac-
be labelled "mayonnaise," the FDA requires that it cumulated. A new edition which contains some of
contain 65% vegetable oil. Within the vegetable cate- the most recent material in a field might attract a
gory, however, are many different sources of oil such number of present non-users of the text who feel
as corn, saffron, and soybeans. Thus, in 1973-74 when that the content of the current edition is out of date.
soybeans quadrupled in price in one year, manufactur- However, certain changes, particularly any deletions
ers using soybean oil in their formulation of mayon- of material in the present edition, might displease its
naise who wished to keep their costs and prices down current users. To evaluate whether or not it is profit-
had to evaluate new formulations which required less able to introduce a new edition (and alternative ver-
soybean oil and substituted other kinds of relatively sions of a new edition), the publisher must consider
cheaper vegetable oils. these consumer reactions and develop an estimate of
the potential demand for the current and proposed
An important part of such an evaluation is determin- new editions.
ing the impact of a formulation change on demand.
Will consumers perceive a change in the product (e.g., Another typical reformulation decision is whether
taste, texture, color, etc.) with the new formulation? or not to change the packaging of a product, either
temporarily or permanently. For example, during holi-
day periods, liquor is sometimes packaged in elegant,
reusable decanter containers rather than regular bot-
*Kathryn E. A. Viilani is Assistant Professor of Business and tles. The principal reason for the special and costly
Donald G. Morrison is Professor of Business at the Graduate School
of Business, Columbia University. packaging is to increase trial by stimulating the pur-
chase of the brand as a gift for others, some of whom
284

Journal of Marketing Research


Vol. XIII (August 1976). 284-8
METHOD FOR ANALYZING NEW FORMULATION DECISIONS 285

are not current users of the brand. The new packaging If no explicit announcement is made about the
probably will not substantially affect the overall vol- formulation change, one would expect the trial proba-
ume bought by current users (although it might affect bility for the new formulation to be about 1.0 for
the timing of one of their purchases to take advantage the current users, because they have no reason to
of the special packaging deal). Therefore, the profit- change their present behavior. For the same reason,
ability of the packaging change depends mainly on one also would expect the trial rate of the current
whether the increase in trial by present non-users non-users to be the same for the new formulation
(compared with the trial with the regular bottle) and as it was for the current brand.
the repeat purchase generated from such a trial increase An explicit announcement of the formulation
more than offset the extra packaging costs. change, however, might provide an incentive for
The purpose of this report is to provide an analytical current non-users to try the product, making the trial
approach for estimating the market sizes for current rate by this segment higher than it would be in the
and new formulations of a product or brand. First, "unannounced" situation. In contrast, an announce-
those factors which appear to have an important ment of a change in the formulation might decrease
influence on the demand for alternative formulations the trial rate and product acceptance by current users.
are discussed. Then a decision tree approach is devel- In the latter case, announcing the formulation change
oped for integrating the influence of these factors causes the current user to evaluate the change con-
on demand in a systematic and meaningful manner. sciously, and forces him to "actually like" rather than
Finally, the implications of this analysis for the design "not dislike" the new formulation.
and interpretation of relevant research, such as taste Finally, if the formulation change is announced,
tests, are discussed. the degree to which it affects the behavior of members
of both segments will depend partly on the advertising
CONSUMER BEHAVIOR AND effort behind the announcement. In general, one would
REFORMULATION expect that the greater the advertising effort, the
Current Users and Non-Users Differ greater the impact the announcement will have, espe-
cially on the trial rates and possibly on the liking
If a new formulation is presented to consumers, of the new formulation.
in lieu of the current product, the total sales volume To predict the level of trial and liking of a new
will come from (1) current users who try and like formulation, therefore, one must consider the market-
the new formulation and (2) new users who try and ing mix that will be used if the new formulation is
like the new formulation. The important issue is that introduced. Also, any research, such as taste tests,
the expected responses to the new formulation from undertaken to examine consumer reaction to a for-
the present users of the current product probably differ mulation change should simulate the conditions under
substantially from those of the current non-users. which the change will be communicated.
From a behavioral point of view, one would expect
that the current users probably will try the new DECISION TREE ANALYSIS
formulation, and continue using it as long as they
do not dislike it. In contrast, the current non-users Construction of Tree
must be convinced to frythe new formulation (although One way of integrating the aforementioned con-
this might be less difficult with the new formulation sumer behavior assumptions to estimate demand in
than with the current one). Also, the current non-users a systematic manner is by building a decision tree.
probably will require a higher level of satisfaction The one in Figure 1 captures the essence of the
to continue using the product. Thus, in most cases consumer behavior crucial to the new formulation
one would expect current users and non-users to have adoption decision. It assumes:
potentially significant differences in what is needed
to induce them to try or to continue using alternative 1. potential consumers for the new formulation are in
one of two categories—users or non-users of the
formulations. As a result, any researcher attempting current formulation,
to forecast the market size for a new formulation 2. potential consumers will either try or not try the new
first must examine the demand for each of these formulation, and
segments separately. 3. potential consumers will either like or dislike the new
formulation.
Announcement of Change Matters
A similar decision tree analysis pertains to demand
Two factors which will affect the level of trial and for the current or old formulation of the product and
liking of a new formulation by the potential consumers it also is shown in Figure 1.
are (1) whether or not an explicit announcement of In the simple classification of "users" versus "non-
the formulation change is made, and (2) if it is users" of the current formulation, such categories
announced, the level of advertising effort put behind as heavy versus light users, loyal versus brand-switch-
selling it. ing prone users, and so on, are ignored. Also, the
286 JOURNAL OF MARKETING RESEARCH, AUGUST 1976

Figure 1 erence for one version of a product rather than another


NEW FORMULATION DECISION TREE are taste tests and home-use tests (monadic or paired
comparison design). Field experiments and lab tests
such as those conducted In simulated supermarkets
can be used to study the trial rates for the different
formulations.
Market Size Estimates
One now can derive the following formulae to
estimate the expected total market sizes for the for-
mulations by the decision tree in Figure I.

where;
Mne^. = The expected number of consumers who will use
the new formulation.
M^u = The expected number of consumers who will use
the current formulation.

non-users consist of consumers with some potential The estimated total market size for the current
for using the current or proposed new formulation. formulation (M^,^) is the sum of the number of present
Trial refers to tasting or using the product formulation buyers (NJ weighted by the probability that they
once. Finally, the categorization of "like" versus like the current formulation (P^,) plus the expected
"dislike" ignores degrees of liking, percentage of number of non-buyers who will try and like this
category purchases to this brand, and so on. Although formulation {N^P^,P^,). The estimated total market
all these assumptions are somewhat oversimplified, size for the new formulation (M^^^,) is obtained by
the essence of the problem is retained. In fact, the estimating the number of people in each buying cate-
simplifying behavioral assumptions allow the analyti- gory for the current product and weighting these by
cal framework to focus more clearly on the major the probabilities that they will try and like the new
issues. formulation.
A Numerical Example
Information Needs
Assume the following taste test situation. In a paired
As indicated in Figure 1, in order to estimate market
comparison test of a current and a new formulation,
size by the decision tree approach, the following data
respondents are asked to indicate the degree to which
are required.
they "like" each alternative on a 6-point scale and
Ny — The number of current users of the current for- later are forced to choose which formulation they
mulation. prefer over the other. Hypothetical results are shown
N^ — The number of potential consumers who are non- in following table.
users of the current formulation.
Pu, — The probability that a current user will try the
current formulation (by definition it is 1.0). HYPOTHETICAL TASTE TEST RESULTS
Pn, — The probability that a non-user will try the current Preferred choice Rated "liked"
formulation.
P^i — The probability that a current user who tries the Old New Old New
current formulation will like it. Sample Number form form form form
Pn, — The probability that a current non-user who tries Users 100 80 20 95 50
the current formulation will like it. Non-users 400 120 280 205 350
Total 500 200 300 300 400
Qn,
— Definitions analogous to those used for the P's,
Qu but with respect to the new formulation. The first column indicates there are 100 users and
Q
400 non-users in the sample. This breakdown is what
one would expect by chance for a random sample
Because most marketers have some measure of their for a product with a 20% market share. The users
market share and the size of their potential market, are assumed to prefer the old formulation by an 80
estimates of N^ and N^ can be obtained relatively to 20 spread, which might be considered excessive
easily. Traditional marketing research approaches can for some products but is used to dramatize the point
be used to help estimate the trial and liking probabil- in question. The non-users prefer the new formulation
ities. Typical methods for examining consumers' pref- by slightly more than 2 to 1. Finally, in the last two
METHOD FOR ANALYZING NEW FORMULATION DECISIONS 287

columns are the numbers of respondents who indicate approach is needed to weight correctly the combined
they "like" the new and/or old formulations by using effects of trial and liking by the users and non-users
the top two points of the 6-point scale. Note that on the total demand estimates.
the rows sum to more than the total number of users Equally dramatic differences between a naive
and non-users in the sample because some respondents aggregate interpretation and the decision tree inter-
indicate "liking" both formulations. pretation could have been obtained by many combina-
On the basis of the aggregate (total) results of the tions of factors. In the example the taste preferences
taste test, the new formulation would be determined for the two groups differed greatly. Less extreme taste
the "winner." The total direct paired comparison differences coupled with lower non-user trial and/or
results indicate that the new formulation is preferred a lower percentage of users in the sample could have
by 60% of the sample whereas 40% favor the old generated similar results. The decision tree approach
product. Similarly, the ratings indicate that 80% of or "model" is really nothing more than an accounting
the sample respondents like the new formulation but equation. Nevertheless, the striking numerical exam-
only 60% like the old. The aggregate analysis, there- ple presented, combined with the fact that many other
fore, would lead one to conclude that the expected plausible combinations of factors can produce similarly
demand for the new formulation is greater than that misleading aggregate results, implies that all of these
for the old formulation. factors should be considered and estimated explicitly
However, what demand estimates are obtained by before a reformulation decision is made.
use of the decision tree approach in analyzing the
taste test results? Assume the product market contains IMPLICATIONS FOR RESEARCH
5,000,000 users and that the test brand has a 20% Marketing research commonly is used to develop
share or 1,000,000 users. The trial rate for non-users estimates of the probabilities of trial or liking of new
is 0.05 and it is assumed that all current users will and current formulations. The decision tree analysis
try. The formulation change will not be announced points out the following fundamental issues which
because it is considered mainly a cost saving device; should be considered in the design and interpretation
therefore the trial rates for the old and new formula- of tests.
tions are expected to be the same, hence P^, = (}„,
= 1.0 and P„, = Q„, = 0.05. Assuming the "like" Sampling
ratings determine the probabilities of actually liking In any tests designed to estimate trial or liking
the formulations, the expected demand estimates for probabilities, it is essential that these probabilities be
the formulations are as follows. estimated separately for the current users and non-
users. This statement implies that current brand users
M =NPP+NPP and target non-users must be defined and identified
in the sample selection process, and that the sampling
= 1.000,000 (1.0)(0.95) + 4,OOO.OO0(0.O5)(O.5125)
procedure must be adjusted so that each group is
= 1,052,500. "represented" adequately.
In many cases a random sample of respondents
selected from the total target market would not provide
= 1,000.000(1.0(0.50) + 4,000,000 (0.05)(0.875) adequate samples for estimating the probabilities for
both the current users and the non-users separately.
= 675,000. For example, if the current brand has a low usage
On the basis of the decision tree analysis results, incidence (i.e. small market share), a random sample
which are contrary to the aggregate analysis, one would selected from the total market would result in only
conclude that the demand for the o/d formulation will a very small part of the sample being composed of
be substantially greater than the demand for the new current users. And unless a very large total sample
formulation. were selected, it would be impossible to get a good
Why do the two analytic approaches lead to different estimate of the probabilities for the current users
conclusions? Which analysis should guide the deci- because they are not represented adequately in the
sion-maker? The aggregate analysis of the taste test sample.
is misleading. It gives equal weight to the taste It should be pointed out that in many taste tests
preferences of the present users and non-users. users and non-users are not identified and only overall
However, even though the non-users have a relatively probability estimates for the total sample are derived,
high probability of liking the new formulation once in the case of the very low incidence brand and a
they try it. they have a low probability of trying the typical random sample selection procedure, the overall
brand and thus their actual impact on demand for results of the taste test for all practical purposes
the old or new formulation is greatly limited. Of course, compare only the probabilities that current non-users
mangement could try to raise this trial rate by promo- will like the current formulation (P,,,) or the new
tions and advertising. In any case, the decision tree formulation (Qn,). Similarly, if the current brand has
288 JOURNAL OF AAARKETING RESEARCH, AUGUST 1976

a large share of its target market and a random sample be useful to experiment with this variable in the taste
is chosen, the probabilities of liking the current and tests in order to measure its influence.
new formulations reflect mainly those of the current
users (P^,and Q,,,) in the typical taste test. In general, CONCLUSION
therefore, if the taste test does not identify users and A decision tree approach is described for estimating
non-users of the current formulation, the results in- the market sizes for current and new formulations
clude an unknown weighting of the probability that of a product. For clear estimates of demand, the
users like the old formulation (P^,) and the probability probabilities of trying and liking the formulations must
that non-users like the old formulation (P^,) versus be determined separately for current users and non-
the probabilities that users and non-users like the new users. A typical situation for a brand with strong loyalty
formulation {Q„, and (?„,, respectively). Obviously, among its current user group (even if the group is
such results are not very useful and may even be very small) might be that the probability of their trying
very misleading as shown by the numerical example. the current formulation iP^,) is essentially 1.0, but
The practical problem in sampling for a low inci- the probability of their liking the new product formula-
dence brand is that current users are difficult and tion (Q^,) most likely will be lower. Will the trial
expensive to find. Therefore, budget considerations rate ((?„,) and liking (Q„,) of the new formulation
may preclude having enough users in the sample. In among the current non-users make up for this loss
contrast, the choice of strong markets as test areas of loyal current users? An answer can be obtained
or "convenience" samples (employees) is likely to only if separate estimates for the current users and
bias the sample in the direction of current users. These non-users are determined.
sample selection practices can be extremely mislead- The decision tree analysis also suggests the following
ing. guidelines for designing and interpreting research un-
Thus, the question arises of whether the cost of dertaken to estimate trial and liking probabilities for
getting better estimates (through research and proper the formulations.
sampling procedures) of the trial and/or liking proba- 1. Current users and non-users must be defined and
bilities for the current users is worthwhile when the quantified.
number of current users is very small. The answer 2. Sampling procedures should provide enough current
depends on the sensitivity of the market size estimates users and non-users in the sample to allow reasonable
(M^^^, and M,,,^) to the probability estimates. This estimates to be made for each segment separately.
can be examined by using optimistic and pessimistic 3. Estimates of each segment's probability of trying the
estimates of the market sizes in the decision tree new formulation should incorporate the influence of
formulae. the strategy that will be used lo introduce il, i.e.,
whether or not the formulation change will be an-
Test Under Simulated Market Conditions nounced explicity and tbe level of advertising.
4. Research done to estimate the probabilities of liking
As was discussed, the decision as to whether or the current and new formulations should consider
not to announce the formulation change explicitly explicitly the effect of the formulation change being
probably will affect both current users and non-users. announced or unannounced.
If no announcement of a change is made, the effect 5. The decision tree approach facilitates analysis of the
of any advertising on trial is the same for the current sensitivity of market size estimates to probability
and new formulations. However, if an explicit an- estimates, thereby providing a vehicle for evaluating
nouncement of the formulation change is made, one whether the cost of doing research to get better trial
would expect the trial rate by current non-users (Q,,,) and/or liking probability estimates is worthwhile.
to be higher than with an unannounced change. And Whether they involve changing the ingredients in
the degree to which the announcement affects trial a food product, or changing a package design, or
will depend largely on the advertising budget and effort changing some of the chapters in a new edition of
behind the announcement. Thus, it is essential that a book, formulation change situations are similar from
research done to provide estimates of the probability an analytical standpoint. They require that the effect
of trial for the new formulation by non-users ((?„,) on buyer behavior for both the current brand and
incorporate the effect of the marketing budget that the new formulation be forecast so that the decision-
will be used to induce trial. maker can evaluate which alternative best meets his
Because it is also probable that explicit announce- selection criteria. They also provide opportunities for
ment of the change will affect the degree to which consumer research such as taste tests, which if prop-
a new formulation is liked, especially by present users, erly designed and interpreted contribute valuable in-
it is important that the conditions under which prefer- formation concerning the probable acceptance and
ence is tested (e.g., taste tests) reflect the communi- success of the alternative formulations. The decision
cation that will accompany the introduction of a new tree approach presented herein provides a helpful
formulation. Also, if the possible effect on preference framework for both these tasks of testing and fore-
of communicating the change is unknown, it might casting.

Você também pode gostar