Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

The Role of Experience in Information Use and Decision Making by Marketing Managers

Author(s): W. Steven Perkins and Ram C. Rao


Source: Journal of Marketing Research, Vol. 27, No. 1 (Feb., 1990), pp. 1-10
Published by: American Marketing Association
Stable URL: http://www.jstor.org/stable/3172546
Accessed: 21-11-2015 12:30 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

American Marketing Association is collaborating with JSTOR to digitize, preserve and extend access to Journal of Marketing
Research.

http://www.jstor.org

This content downloaded from 130.237.29.138 on Sat, 21 Nov 2015 12:30:22 UTC
All use subject to JSTOR Terms and Conditions
W. STEVENPERKINS
and RAMC. RAO*

Theauthorsreportthe resultsof a studyon the use of informationand decision


makingby practicingmarketingmanagers.Theyexaminethe effect of managerial
experienceand decisionprogrammability on managers'informationuse and deci-
sions. Theresultsindicatethat experienceis an importantdeterminantof managers'
behaviorfor relativelyunprogrammed decisions.Moreover,the effectsof experience
are manifestedin the evaluationand use of "soft" information,amountof infor-
mationused, and the decisionsthemselves.The authorsconcludethat in studying
the effects of experienceon managerialdecisions,one mustexaminethe steps lead-
ing to a decision;theyfindthat experienceaffectsthe priorstep in whichinformation
is acquired.To the extent experienceis a proxy for expertise, the articleaffords
insightsinto the key ingredientsin trainingand modelingexpert managers.

The Role of Experience in Information Use and


Decision Making by Marketing Managers

Understanding decision making by marketing man- decision process itself. In particular, by studying prac-
agers can provide insights that will help in building bet- ticing managers, we can understand better the elements
ter tools for managers to use and in improving marketing that affect real-world decision making. The research we
decisions. Marketing models, for example, do not at- report comprises a formal study of marketing managers
tempt to tell managers what to do; rather, they create a in a single large company. Specifically, our interest cen-
fruitful interaction between managers and models. Ef- ters on the interaction of managerial experience and de-
fective model building in this sense requires the manager cision programmability as it obtains among practicing
to understand and trust the model, and the modeler to managers.
appreciate the manager's problem and take advantage of The few descriptive studies of marketing decision
the manager's knowledge derived from his or her ex- making (Hulbert 1981) have addressed two issues: (1)
perience (Little 1970; Little and Lodish 1981). In this the use of managerial judgments to parameterize deci-
context, a relevant question is: What is the real interplay sion models and (2) the use of marketing research by
between managers' experience and their decision mak- managers. The first issue relates to the use of managers'
ing? Clearly our ability to improve managerial decision judgments to build decision calculus models that may be
making ultimately depends not just on sophisticated difficult to parameterize by other methods (Little 1970).
models, but also on our understandingof the managers' Fudge and Lodish (1977) illustrate the usefulness of this
approach. Chakravarti,Mitchell, and Staelin (1979), in
an important experimental investigation, found the ap-
plicability of this approach to be limited mainly to stable
*W. Steven Perkins is Assistant Professor of Marketing, College markets. (See also Larr6ch6 and Moinpour 1983; Mc-
of Business Administration, Pennsylvania State University. Ram C.
Rao is Professor of Marketing and Founders Professor, School of
Intyre 1982).
Management, The University of Texas at Dallas.
Deshpande and Zaltman (1982, 1984) studied mar-
The authors thank James C. Bartlett and Randall L. Schultz for their keting managers' use of marketing research information.
suggestions in formulating the research, John Wiorkowski, Rick Stae- They found that managers tended to use more informa-
lin, Pam Pickard, and Naufel Vilcassim for valuable discussions, and tion if it was deemed technically adequate, if it con-
Marci Armstrong, Dipankar Chakravarti, Rohit Deshpande, Don Leh-
firmed priorexpectations, and when the managersworked
mann, Harish Sujan, and Mita Sujan for comments on a previous draft.
Any remaining errors are the authors'. closely with the researchers.Lee, Acito, and Day (1987),
in an experimental study of MBA students, found greater
use of marketing research information when the infor-
1
Journal of Marketing Research
Vol. XXVII (February 1990), 1-10

This content downloaded from 130.237.29.138 on Sat, 21 Nov 2015 12:30:22 UTC
All use subject to JSTOR Terms and Conditions
2 JOURNAL
OF MARKETING FEBRUARY
RESEARCH, 1990

mation confirmed prior beliefs, regardless of technical this section we elaborate on why these factors are im-
quality of the research. portant to examine.
The results of previous research highlight the need for
more descriptive studies of marketing decision making. Experience
They also have some limitations. One has been the ab- Our premise is that with experience managers acquire
sence of an experimental setting that enables subjects to expertise in making decisions. Studies comparing ex-
engage in "meaningful tasks." For example, the use of perts and novices suggest that experts have more highly
MBA studentsreduces our confidence in inferences about developed cognitive structures-the organization of in-
practicing managers. Also, subjects perform experimen- formation in memory and the repertoireof rules for using
tal tasks that differ, often substantially, from the prac- that information-which allow for effective problem
tical decisions they may be accustomed to facing. A more structuring and successful problem solution (Chi, Fel-
valid approach would be to study real-world managers tovich, and Glaser 1981). Effective salespersons, for ex-
making decisions similar to those in their natural envi- ample, have been found to know more customer traits
ronment. (i.e., have more facts in memory) and to know more
Another area not fully addressed by previous research selling strategies (i.e., more rules; Sujan, Sujan, and
is the interaction effect of managers' characteristics and Bettman 1988). According to Harmon and King (1985),
decision characteristicson decision making (Payne 1982). experts use these facts and heuristics to solve problems
In fact, it is this interactionthat defines the "space"within effectively. In fact, experienced subjects are likely to
which decisions are made (Newell and Simon 1972). An search for more information, simultaneously restricting
example in marketing is the idea that effective selling is themselves to relevant and importantinformation(Chiesi,
contingent upon both the salesperson's resources and the Spillich, and Voss 1979). When managers are provided
environmental characteristics (Weitz 1981; Weitz, Su- information, as opposed to acquiring it, they will differ
jan, and Sujan 1986). Finally, in studying decision mak- in their valuation of the information. Specifically, more
ing, it would be useful to examine not just the decision, experienced managers can be expected to place more
but the steps leading up to it. Hogarth (1980) outlines weight on relevant (functional) cues and less on irrele-
the steps as structuringthe decision problem, acquiring vant (peripheral) cues (Brucks 1985). Thus, experience
information, processing it, and then making a decision. determines the amount of information sought and its val-
Our study spans the information acquisition and pro- uation.
cessing steps as well as the decision itself. Differences in information valuation lead to differ-
Our research is aimed at understandingmanagerial de- ences in decisions. Moreover, experienced managers may
cision making by explicitly studying the role of experi- also understand the uncertainties and consequences of
ence when a task characteristic-programmability-is their decisions better than their inexperienced counter-
varied. Programmability of a decision is the extent to parts (Beach 1975; Nisbett et al. 1983). In other words,
which the decision can be made by using relatively rou- experience affects managerialdecisions, and more so when
tine procedures instead of more general problem-solving it influences information valuation. Though the role of
techniques. Subjects for our research are practicing man- expertise and experience in consumer behavior has been
agers who engage in experimental tasks closely resem- studied extensively (Alba and Hutchinson 1987), it has
bling their on-the-job decisions. We obtain insights into not received attention from researchers studying mar-
how experience and programmabilityof decisions affect keting managers' decision making.
(1) managers' use of information to arrive at decisions
and (2) managers' decisions themselves. Briefly stated, Programmability
our overall finding is that more experienced managers A useful classification of decisions is one based on the
regarded more kinds of information as useful and made concept of programmability, proposed by Simon (1960).
more conservative decisions, especially in a relatively In his framework, programmed decisions are character-
unprogrammedsituation. The results are consistent with ized as being routine and structuredwith a well-defined
our hypotheses and prior theory. starting point, a clear goal, and standardized rules for
We next state the research issues and objectives, then reaching the goal. They are repetitive enough to permit
outline the study method. After reporting the data anal- the establishment of definite procedures to process them.
ysis and our results, we discuss our conclusions, our According to Simon, methods for making programmed
study's limitations, and the implications for managerial decisions include habit, standard operating procedures,
practice and future research. and a common set of expectations established by the or-
ganization. In contrast, nonprogrammeddecisions are ill-
RESEARCHISSUES AND OBJECTIVES structuredand have few guidelines. They are novel, not
What is the rationale for choosing experience as the being amenable to processing by a prespecified method
characteristic of managers and programmability as the and often requiring the decision maker to rely on general
task characteristic in studying marketing decisions? One problem-solving abilities. When making unprogrammed
reason is that previous research in marketing decision decisions, managers must exercise judgment and ".
making has not explicitly considered these factors. In this judgment depends, in some undefined way, upon

This content downloaded from 130.237.29.138 on Sat, 21 Nov 2015 12:30:22 UTC
All use subject to JSTOR Terms and Conditions
IN INFORMATION
ROLEOF EXPERIENCE USE 3

experience, insight, and intuition . . ." (Simon 1960, programmed" serving as labels for the extreme cases.
p. 11). We expect, then, that the effects of experience The two marketing decisions in our study do not rep-
will be greater in a relatively less programmed decision. resent these extremes. Rather, they differ on critical pro-
Simon's distinction between programmed and unpro- grammabilitycharacteristics,including repetitiveness and
grammed decisions has provided a basis for studying standardizationof processing method.
managerial decisions in such areas as information sys-
tems (Gorry and Scott Morton 1971) and auditing (Ab- Programmability
dolmohammadiand Wright 1987). The same concept has The consumer promotion decision, for this corpora-
been applied by artificial intelligence researchers in clas- tion, represents a routine task with a fairly standardized
sifying problem types (Steffik et al. 1982). In marketing, set of well-defined information inputs. Though some
there is no absolute measure of programmability of de- managers may not have repeated experience with such
cisions. Gorry and Scott Morton (1971) believe that short- decisions, the firm has developed standard operating
term sales forecasting, for example, could be thought to procedures for making common ones. Having learned
be more programmed than the novel decisions in new these problem-solving procedures through corporate and
product planning. Examples of unstructured marketing personal knowledge, managers can reasonably be ex-
decisions include generating advertising appeals and rec- pected to agree on what information to use and how to
ommending a negotiation strategy, both of which in- use it. As the consumer promotion decision involves rel-
volve incomplete information and a reliance on judg- atively well-understood information and well-defined
mental heuristics (Rangaswamy et al. 1987). Many procedures, we consider it a good exemplar for a pro-
important decisions are in a middle ground of semi- grammed decision in our study. In a recent audit, prod-
structured decisions-for example, the setting of mar- uct managersrated promotionexecution as a routine, less
keting budgets (Keen and Scott Morton 1978). The ex- desirable management task (Quelch, Farris, and Olver
tent of programmabilityof different marketing decisions 1987). The decision's programmed nature presumably
could be distinguished on the basis of how standardized limits the need for individual managerialjudgment. Con-
the decision process is-less standardized for less pro- sequently, we expect individual experience to be a less
grammed-and whether logical and broad information significant factor in this decision.
is needed for the decision as opposed to narrow and New product introduction, in comparison, involves a
arithmetic-the latter being more suitable for pro- more novel decision for which there are fewer formal,
grammed decisions. Hence, we expect the effect of ex- well-developed guidelines. Though the firm has intro-
perience on decisions to be greater in cases in which duced several new products, each carries a different nu-
managerial judgment plays a greater role. In the next ance and in that sense is unique. The decision task has
section, we return to the question of how we operation- less inherent structure. Moreover, individual managers
alized programmability in our study. experience fewer repetitions of this decision. It is rea-
On the basis of the foregoing discussion, we expect sonable to expect that managers often must provide their
to find the following interaction of experience and pro- own structure in this instance. Therefore this decision
grammability: serves as an unprogrammedone in our study. We expect
greater individual differences in this decision, with ex-
-Experience effects will obtainin the decisionsand in the perience being an important explanatory factor.
use andweightingof relevantinformationpriorto making
the decisions. Subjects
-These experienceeffects will be greaterfor the less pro-
grammablethanfor the more programmable decisions. Fifteen managers from a single division of the com-
pany participated. The managers work with different
Thus, we expect experience effects across decisions; we brands, but the firm follows the same marketing process
also expect experience effects in different steps of de- with all the brands. The brands are distributed through
cision making. the same channels, promoted and priced in similar ways,
advertised in the same media, and even produced in the
THE STUDY same plants. The marketing experience of most of the
To investigate our research questions, we conducted managersconsisted of what they had acquiredat this firm.
a study with brand managers in a major Fortune 500 They were about 30 years of age and held MBA degrees
corporation. The corporation serves a national consumer from major universities. The brand management struc-
packaged goods market in which advertising, promo- ture in the firm consists of three staff levels-senior,
tions, and new products are important strategic vari- associate, and assistant manager-and managers spend
ables. The study examined two of its marketing deci- about two years at each level as they move up in the
sions, scheduling consumer promotions and introducing firm. Five subjects at each level participatedin the study.
new products, which represent, respectively, more pro-
grammed and less programmed decisions. Simon, who Experience
coined the term "programmability,"emphasizes that de- We operationalize experience by the months of brand
cisions form a continuum with "programmed"and "un- management experience of a manager (referred to here-

This content downloaded from 130.237.29.138 on Sat, 21 Nov 2015 12:30:22 UTC
All use subject to JSTOR Terms and Conditions
4 JOURNAL
OF MARKETING FEBRUARY
RESEARCH, 1990

after simply as "experience"), which correlates highly managers make and the information they use. The stim-
with alternative measures used. The correlation between uli were tailored to be as realistic as possible, including
experience and staff level is .82 (p < .01) with seniors, the same kinds of information managers normally would
associates, and assistants having median experience of have available and even specific terminology they em-
46, 26, and 4 months, respectively. We also obtained ployed. Thus the need for extensive instructions was
from the managers their self-reported ability on and fre- eliminated, as the managers were making simulated ver-
quency of participationin each of the two marketing de- sions of the decisions required in their job.
cisions. The correlation of experience with the number The prestudy identified 14 kinds of marketing infor-
of promotion decisions is .78 (p < .01) and with the mation important for the decisions (Table 1). Seven are
number of new product introductions is .52 (p < .05). related primarily to promotion decisions and the other
Experience does not correlate with the measures of self- seven to new product decisions. When managers made
reportedability for promotiondecisions, r = .07 (p n.s.), decisions, they had available only the seven pieces of
but it does for the new product decision, r = .75 (p < information relevant to each decision. For each type of
.01). Hence, though managers made more promotion de- decision, eight scenarios were developed by dichotomiz-
cisions as their tenure increased, their knowledge, as ing the seven pieces of information into high and low
measured by their self-reported ability, did not increase. values according to an orthogonal fractional factorial de-
In contrast, experienced managers had encountered more sign (Cochran and Cox 1957, plan 6A.7, p. 280). These
new product decisions and perceived an increase in their values are listed in Table 1. To check the reasonableness
ability to make those decisions. of the information combinations generated by the de-
Procedure sign, managers assessed the likelihood of encountering
each scenario in practice. The mean likelihood was .7
Each manager participated in an individual two-hour and .6 for promotion and new product decisions, re-
session conducted in private at the corporation. There spectively, indicating that the scenarios represented re-
was no communicationamong the subjects about the study alistic situations. These likelihood values did not differ
during its conduct. The flow of questions was controlled significantly across decisions.
by following a script. Most of the tasks required a verbal
Tasks
response, with responses kept to a minimum. Sessions
were recorded and managers were encouraged to think Managers performed three tasks for each type of de-
aloud, explaining why they used certain information and cision: rating the usefulness of all 14 pieces of infor-
how they made decisions. mation, choosing some or all of seven pieces of infor-
Subjects were assigned randomly to perform either the mation on each of eight trials for each decision type, and
promotion tasks or the new product tasks first. Before making decisions on the eight trials of each type. For
beginning either task they read a one-page statement convenience, we call them information rating, infor-
containing background information related to the deci- mation choice, and decision-making tasks. In the rating
sions. For promotion decisions it read, in part, ". task, managers assessed the usefulness of information
During the annual planning process a consumer pro- for a decision on an 11-point scale ranging from 0 (not
motion was scheduled for the future. The time has come at all useful) to 10 (extremely useful). These data were
to be more specific about the promotion. . . . You must used to examine differences in ratings by experience and
review the current situation and decide if it would be to check on the validity of our choice of what infor-
appropriateto have a consumer promotion for your brand mation to provide for each decision type.
or to wait until later. . . ." It also included information The choice task consisted of selecting information,prior
on the purpose, type, duration, and depth of promotion. to each decision trial, from seven pieces randomly or-
In effect, these aspects were held constant; the manager dered for each scenario. The pieces were presented on
decided only on the scheduling of the promotion. The seven cards, each indicatingthe kind of information(e.g.,
statement for the new product decisions said that the Nielsen data). By turning over the card, the manager
product had been in test market for six months and also would learn whether that variable was better or worse
indicated the product's purpose and the introductory than "expected" (as shown in Table 1). The managers
budget. Specifically, it read, ". . . The time has come were to choose in order of perceived usefulness. They
to decide if the product should be discontinued or rolled were encouraged to choose only as many pieces as they
out nationally. You must review the currentsituation and needed to make a good decision because information in
decide if it would be appropriateto stop marketing the the real world would be costly. Our interest in this task
new product or go ahead with it. .. " centered on studying the number of pieces chosen and
the consistency between the order of choice and the use-
Stimuli fulness ratings. Our rationale is that experienced man-
Stimuli and tasks were constructed during the prestudy agers have well-developed frameworks for decision
phase with the aid of several other managers in the cor- making, which would influence the numberof pieces (i.e.,
poration. Discussions with them spanned the decisions amount) of information sought. Consistency between

This content downloaded from 130.237.29.138 on Sat, 21 Nov 2015 12:30:22 UTC
All use subject to JSTOR Terms and Conditions
ROLEOF EXPERIENCE
IN INFORMATION
USE 5

Table 1
INFORMATION
AVAILABLE
FORDECISIONS

Values
Information type +1 -1
Consumer promotion
1. Weekly sales report
Actual and targeted sales by product and region Above target Below target
2. Trade promotion
Brand sales promotion to stores Promotion on Promotion off
3. Ad schedule
Ad pulse or flighting schedule Flighting on Flighting off
4. Competitor's promotion
Major competitor sales promotion to stores Promotion on Promotion off
5. Responsiveness to consumer promotions
Past impact of this type of promotion Strong response Weak response
6. Consumer promotion budget
Budget available for consumer promotion Above expected Below expected
7. Season of sales year
Seasonal pattern of sales through 13 annual periods Higher than average Lower than average
New product introduction
1. Nielsen store audit data
Bimonthly sales data based on store inventories Sales above average Sales below average
2. Diary panel data
Detailed record of product use by panel families Use above average Use below average
3. Headquarters support
Strength of corporate support for new product Above average Below average
4. Peer opinion
Subjective input from your fellow product managers Positive, do it Negative, don't do it
5. Competitor's new product introduction
Major competitor has launched new product similar to yours Yes No
6. Consumer research
Burke research on consumer attitudes from focus groups Positive results Negative results
7. Merchandiser word of mouth
Subjective feedback from key route personnel Positive feedback Negative feedback

stated ratings and behavior (i.e., actual selection) was promotion decisions and the other seven more relevant
expected to be more likely with better developed struc- for new product decisions. To test this assumption we
tures. performed a fully saturated repeated-measures analysis
Finally, the managers stated the probability of action of variance over managers' ratings of all 14 pieces of
given the information. For promotions they indicated the information and both decision tasks. Let R denote the
probability of carrying out the promotion and for new usefulness rating of information, I a 14-level information
products they gave the probability that they would in- factor, D a (2-level) decision factor, and S a 15-level
troduce the new product. They stated their decision first subject factor. Then the ANOVA model is
on the basis of the information they selected. They were
then given all the information and again asked to state (1) R= + I + D + S + (Ix D) + (I x S)
their decision. In our analysis we use the second prob- + (D xS) + (Ix D x S).
ability statement to preserve the orthogonality of the in-
dependent variables. As a check, the first and second In model 1 we examine the interaction I x D and find
statements correlated .95 for promotions and .94 for new that ratings differ by decision type: F(13,182) = 9.84
products. Managers made one practice decision before (p < .01). Table 2 gives the results of univariate tests
beginning their eight trials. of differences of mean ratings of each piece of infor-
mation for the two decisions. Twelve of 14 pieces of
ANALYSIS
AND RESULTS information are valued differently in the expected way
with six being statistically significant. One piece of in-
Information Rating formation has the wrong sign and is significant. These
The first issue to address is whether or not managers findings generally confirm that our information manip-
valued the information differently by decision type. Our ulation is valid for the two decisions.
study was designed on the assumption that, of the 14 The next question is what, if any, effect experience
pieces of information, seven would be more relevant for has on these ratings. We find both quantitativeand qual-

This content downloaded from 130.237.29.138 on Sat, 21 Nov 2015 12:30:22 UTC
All use subject to JSTOR Terms and Conditions
6 OF MARKETING
JOURNAL FEBRUARY
RESEARCH, 1990

Table2
MEANUSEFULNESS
MANAGERS' RATINGS FORINFORMATION
(standarddeviationsin parentheses)

Marketing decision
Probability rating differs
Information Promotion New product (t-test)
Consumer promotion
1. Weekly sales report 4.9 8.3 .01
(2.5) (2.9)
2. Trade promotion 9.3 6.6 .01
(.9) (2.5)
3. Ad schedule 8.1 6.9 .10
(1.7) (2.4)
4. Competitor's promotion 6.3 6.0 .71
(2.8) (2.6)
5. Promotion responsiveness 9.5 5.8 .01
(1.1) (2.7)
6. Consumer promotion budget 9.7 6.7 .01
(.5) (2.6)
7. Season of sales year 8.3 6.3 .03
(1.7) (2.9)
New product introduction
1. Nielsen data 4.6 8.3 .01
(2.2) (1.8)
2. Diary panel data 5.1 8.6 .01
(3.3) (1.7)
3. Headquarters support 6.8 8.1 .13
(2.3) (2.4)
4. Peer opinion 5.7 4.4 .13
(2.0) (2.4)
5. Competitor new product 6.7 8.4 .06
(3.0) (1.4)
6. Consumer research 6.7 9.2 .01
(2.5) (1.7)
7. Merchandiser word of mouth 6.1 7.1 .83
(3.1) (2.4)

itative differences in the effect of experience. First, we all pairs of managers (Hays 1988). This correlation av-
performed an analysis of variance over ratings of all 14 erages .40 for promotion decisions but is significantly
pieces of information, and both tasks, treating experi- lower, .21, for new products: t(208) = 4.88 (p < .01).
ence as a covariate replacing the subject factor in equa-
tion 1. Across all ratings and both decisions, experience Information Choice
is significant, F(1,364) = 5.75 (p < .05). The inter- To understandmore fully the managers' use of infor-
action of experience and information also is significant, mation, we examined their actual choice of information
F(13,364) = 1.74 (p < .05), as is the three-way in- when faced with the decision task. Recall that managers
teraction of experience with information and decision, chose to look at some or all of the seven pieces of in-
F(13,364) = 1.71 (p < .10). Hence, experience is seen formation before stating their decision. Given this rank
to affect the perceived usefulness of information and this ordering of information, we examined three issues: the
effect varies by decision programmability. Moreover, by consistencybetween statedratingsand rankorderof choice
examiningthe qualitativedifferencesacross decision types, of a piece of information in each task, the convergence
we see that the more experienced managers rated several among managers on the rank ordering, and the average
of the "promotioninformation"pieces as useful even when number of information pieces chosen before making a
making new product decisions. Ratings of five of these decision.
information pieces are correlated positively with expe- Consistency for each individual was determined by
rience, suggesting that experiencedmanagersvalue a wide correlating the ratings and rank ordering of information
variety of information for new product decisions. averaged across trials. As a lower rank order would im-
A second analysis confirms that differences across ply a greater usefulness, this correlation should be neg-
managers are greater in the case of new product deci- ative. For the promotion decisions the correlation aver-
sions than in the case of promotions. Convergence among ages -.55 (n = 7 per subject) across all managers. In
managers was examined by computing the average the same analysis for the new product decisions, the cor-
Spearman correlation between the information ratings of relation averages -.59 (n = 7 per subject) across all

This content downloaded from 130.237.29.138 on Sat, 21 Nov 2015 12:30:22 UTC
All use subject to JSTOR Terms and Conditions
IN INFORMATION
ROLEOF EXPERIENCE USE 7

managers. These correlations do not vary significantly Let Ej denote the months of brand management experi-
by experience for either decision.1 ence of individualj. We develop only the promotionmodel
Convergence among managers was determined by in detail, as the new product model is identical. A linear
computing the average Spearman correlation between all model of the promotion decisions, then, is
pairs of managers' rank orderings. This correlation av-
erages .46 for promotion decisions; for new product de- (2) Y,= ao + yE E Pk,Eki + kXikE, + Zi
cisions it is marginally lower, .38. Managers thus gen- k k

erally agreed more on how valuable the informationpieces where 4iiis the errorterm.3 The parametershave the fol-
were for promotion decisions. This confirms the findings
based on ratings. lowing interpretation:a0 can be thought of as the prob-
ability of promotion by a manager with no experience
Finally, for the number of pieces of information cho- when all the information is as expected, y measures the
sen, we find that managers selected 5.1 pieces on av- incremental probability of promotion per month of ex-
erage for promotion decisions and 5.5 pieces for new
perience, 3k(-3k) measures the effect of being above
product decisions. In a paired t-test these means differ, (below) expected on information k, and 5k measures the
t(14) = 1.53 (p < .10). We find no differences by ex- interaction of experience and information.
perience in the number of information pieces chosen.2
Consumer Promotion Decisions
Marketing Decisions
Table 3 reports the results of estimating equation 2 for
Recall that in our study a decision is the probability the promotion decisions by ordinary least squares.4 We
of acting given information. The two decisions are the find that all the experience-related variables are insig-
probability of a consumer promotion given information nificant. In particular, ^ and 8k, for all k, are not sig-
and the probability of new product introduction given
nificantly different from zero, indicating that experience
information. Each manager provides his or her proba- does not affect the promotion decisions across managers.
bility for each of eight trials on each of the two deci- When all the variables are as expected (i.e., neither be-
sions. The trials represent a fractional replication of the low nor above expected), the probability of promotion
factorial design of a high and low value on each of seven is given by ^o = 44.25%. Of the information variables,
pieces of information. We are interested in two aspects promotion response is the most heavily weighted: if the
of these decisions: the weight placed on each informa-
responsiveness of the brand to promotion is strong, the
tion piece in making a decision and the effect of expe-
probability of promotion increases by (s = 19.39%.
rience on these weights. We estimated the weights using
Similarly, with an above-expected promotion budget, the
a dummy variable regression model, as is often done in
probability of promotion increases by 36 = 9.73%. If
the analysis of profiles from fractional factorial designs trade promotion is on, the probability of promotion in-
in conjoint analysis (Green and Srinivasan 1978). We creases by 32 = 6.87%. Of the remaining information
introduced a second set of coefficients in the regression
variables, weekly report has a marginally significant (p
model to capturethe experienceeffect (Green and DeSarbo < .10) negative coefficient, S1,indicating that if the lat-
1979; Holbrook 1981). We estimated separate models est weekly report shows actual sales to be above target,
for the promotion and new product decisions. the probability of a promotion decreases. No coefficients
Let Yj represent the decision on trial i, i = 1, ..., 8, have implausible signs. The regression has an R2 of .55,
by individual j, j = 1, ..., 15, for the promotion deci- indicating good overall explanatory power of the model.5
sions. Denote Xik to be the code of information k, k =
1, ..., 7, on trial i. Recall that information takes on one
of two values,
- 1-1 if informationis below "expected" 3The error C, is likely to be correlated across trials i, suggesting a
repeated-measures analysis of variance. Unfortunately, given our
Xk if informationis above "expected." fractional factorial design, such an analysis can only speak to the ex-
perience x trial interaction and not the experience x information in-
teraction of interest here. For completeness, we performed a repeated-
measures ANCOVA for the two decisions. It revealed that neither
'To explicate this issue further, we looked at individuals in the ex- experience by itself nor the interaction experience x trial was signif-
tremes of staff levels by collecting the five assistant managers in one icant for the promotion decisions, but both were significant for the
group and the five senior managers in another group. The correlation new product decisions.
between rating and rank order for promotion decisions did not differ 4To account for the correlation across trials, one might estimate the
across the groups. In contrast, for the new product decisions, the cor- model using seemingly unrelated regressions (Zellner 1962). This ap-
relations turned out to be -.39 for the assistants and -.80 for the proach does not help in our particular case, however, as the X matrix
seniors. Thus, senior managers appear to be more consistent in their is identical across trials (Theil 1971, p. 309-310).
ratings and choice of information for the less programmed decision. 'Qualitatively, on the basis of sign (for the nonsignificant param-
2Again, we examined the average number of information pieces eters), we find that experienced managers promote less often (j < 0),
chosen by managers at the extremes of staff levels. We found that for are not influenced as positively by trade promotion and promotion
the promotion decisions senior managers chose fewer pieces than as- budget (82, 6 < 0), weight promotion response higher (8, > 0), and
sistants, but for the new product decisions they chose more than the are more likely to promote when advertising flighting is on (83 > 0)
assistants. or if seasonal sales are higher (87 > 0).

This content downloaded from 130.237.29.138 on Sat, 21 Nov 2015 12:30:22 UTC
All use subject to JSTOR Terms and Conditions
8 JOURNAL
OF MARKETING FEBRUARY
RESEARCH, 1990

Table 3 are as expected, as j < 0. In particular, for a manager


OF PROMOTION
ESTIMATES DECISIONSMODEL with five years (60 months) of experience, this proba-
in
(t-values parentheses) bility decreases by 10%. Headquarters support is more
valued by experienced managers (83 > 0), whereas mer-
Parameter estimates chandiser word-of-mouth support is not valued as much
No experience Experience
(8s < 0). Hence, experiencedmanagersdiffer more on
Variable interaction interaction the valuation of the subjective information-headquar-
ters support and merchandiser word of mouth. All man-
Intercept do = 44.25a = -.05
(15.74) (-.53) agers weight the "hard"data heavily-Nielsen store au-
Weekly report = -5.09 b, = .00 dit data (p, = 9.49), diary panel data (32 = 11.11), and
(-1.81) (.02) consumer research (36 = 11.41). These coefficients are
Trade promotion = 6.87b
P2 2= -*.09 all highly significant. All managers view a competitor's
(2.45) (-1.02) new product as a reason for not introducing a new prod-
Ad schedule P3 = -1.22 83 = .14
(-.43) (1.57) uct (P3 = -6.02). Of the "soft" information, only peer
Competitorpromotion 4 = -1.49 84 = .00 opinion does not have a significant weight, though it is
(-.53) (.02) positively viewed (P4 > 0). Inexperienced managers
Promotion response = 19.39' 8, = .03
P5
(6.90) (.38) weight merchandiserword of mouth positively (37 > 0),
Promotion budget 6 = 9.73a 6 = -.10
but for experienced managers it is insignificant. Head-
(3.46) (-1.15) quarters support is viewed positively (33 > 0), but is
Season 7 = -.47 7 = .10 insignificant for inexperienced managers. The model has
(-.17) (1.14) an R2 of .58, indicating good explanatory power.6
Adjusted R2 = .55, F(15,104) = 10.72 We see that more information is valued by managers
ap < .01, two-tailtest. for the new product decisions than for the promotion de-
bp < .05, two-tail test.
cisions, experienced managers differ significantly from
inexperienced ones in the new product decisions but not
in the promotion decisions, and these differences are in
New Product Decisions the use of "soft" information.
Table 4 gives the results for the new product decisions
from estimating a model similar to equation 2. For the DISCUSSION AND CONCLUSIONS
experience-related variables, we find that three coeffi- Our study largely confirms our hypotheses that man-
cients are significant. Experiencedmanagersare less likely agerial experience and decision programmabilityinteract
to introduce new products when all information variables in managerialdecision making. Specifically, we find that:
-the effects of experienceare morepronouncedin the less
programmed (new product)decisionthanin the morepro-
Table4 grammed(promotion)decision,
ESTIMATES
OF NEW PRODUCT DECISIONS
MODEL -the effect of experienceis manifestednot only in the de-
cisions themselves,but also in what informationis used
in
(t-values parentheses) to make the decisions, and
-the moreexperiencedmanagersdifferfromtheirless ex-
Parameter estimates periencedcolleagueson the new productdecisionin sev-
No experience Experience
eral importantways: they ratedmoreinformationas use-
Variable interaction interaction ful, weightedthe "soft" informationdifferentlyin their
do = 47.93' j = -. 17 decisions, and made more conservativedecisionson av-
Intercept
(18.92) (-2.18) erage.
Nielsen data = 949 ~ = .02 As Simon (1960) noted almost 30 years ago, pro-
(3.74) (.22)
Diarypaneldata 2 = 11.1a 2 = -.08 grammed and unprogrammeddecisions require different
(4.38) (-.97) decision-making techniques. Routine, repetitive deci-
Headquarters support P3 = 4.05 83 = .18b
sions yield to habit or specialized procedures. However,
(1.60) (2.36)
Peer opinion P4 = 1.89 = .01
84
many importantmarketing decisions are closer to the un-
(.74) (.10) programmed end of the spectrum, requiring managerial
Competitor's new P5 = -6.02b 5 = -.01
product (-2.38) (-.12)
Consumer research 36
= 11.41' -.08
8•=
(4.50) (-1.04) 6Qualitatively,on the basis of sign (for the nonsignificantparam-
Merchandiser word of 37 = 11.82' 87 = -.17b
eters),we find thatexperiencedmanagersvaluemoretheirpeeropin-
mouth (4.66) (-2.12) ion (84 > 0) and Nielsen store audit data (8, > 0), weight less the
Adjusted R2 = .58, F(15,104) = 12.07 informationfromdiarypaneldata(82 < 0) andconsumerresearch(86
ap < .01, two-tailtest. < 0), andaremorenegativelyaffectedby a competitor'snew product
bp < .05, two-tail test. (86 < 0).

This content downloaded from 130.237.29.138 on Sat, 21 Nov 2015 12:30:22 UTC
All use subject to JSTOR Terms and Conditions
IN INFORMATION
ROLEOF EXPERIENCE USE 9

judgment and general problem-solving abilities. Thus, perience have been shown in several domains ranging
improvements in marketing management can be effected from physics (Larkin 1983) to auditing (Ettenson 1984).
through a better understanding of managerial decision At a more substantive level, we believe future re-
making. In particular, unprogrammed decision making search should extend our idea of examining not just the
can be improved by training managers and modeling decision, but the steps leading up to it (e.g., Soelberg
managers. In terms of training, knowledge differences 1967). We stepped back to look at information choice
due to experience can be exploited to accelerate and di- and weighting. As an extension, subjects could be asked
rect novices' learning (Weitz, Sujan, and Sujan 1986). to interpret raw data as opposed to being told that the
For instance, the fact that senior managers valued more information is better or worse than expected. Another
sources of information as well as "soft" information for possible extension might be to step fartherback, to study
new product introductions should be codified and passed the way managers structure their problem and how that
on to novices. In contrast to traditional training, part of influences their actions toward raw data, informationand,
what managers learn through experience is tacit knowl- finally, decisions.
edge, not explicitly expressed or taught, but often a ma- The formal study of decision making by practicing
jor building block of practical, real-world intelligence managers has enormous potential because surprisingly
(Wagner and Sternberg 1985). For example, the fact that little is known. Extant work indicates that both research
senior managers paid close attention to internal corporate (Deshpande and Zaltman 1983) and models (Little 1970,
information and discounted the word-of-mouth infor- 1986) are often underutilized by managers. The question
mation received from sales personnel may have to be is: How, then, do they make decisions? For now, on the
learned by on-the-job training or learning by doing (An- basis of our findings, we can say that for unprogrammed
zai and Simon 1979). decisions, experienced managers differ from novices in
In terms of modeling managers, our findings suggest both their use of information and in the decisions they
that the input of managerial knowledge will have the make.
greatest impact on models of decision making in less
programmed domains. Because experienced managers REFERENCES
differ from novices on unprogrammed decisions, cap-
turing their knowledge could have a significant effect on Abdolmohammadi, MohammedandArnoldWright(1987), "An
decision making. Expert systems, of course, attempt to Examinationof the Effects of Experienceand Task Com-
do just that. Examples of such unprogrammeddecisions plexity on Audit Judgments,"AccountingReview, 62 (1),
in marketing include new product design, launch, and 1-13.
positioning (Rangaswamy et al. 1987). However, trans- Alba, J. W. and J. Wesley Hutchinson(1987), "Dimensions
of Consumer Expertise," Journal of Consumer Research, 13
ferring experts' knowledge from human to machine has
been much more time consuming and difficult than ex- (March),411-54.
Anzai, Yuichiroand HerbertA. Simon (1979), "TheTheory
pected by early advocates of expert systems (Hayes-Roth, of Learningby Doing,"PsychologicalReview,86 (2), 124-
Waterman, and Lenat 1983). Research such as ours may 40.
speed their development by explicating differences be- Beach, BarbaraH. (1975), "ExpertJudgmentAbout Uncer-
tween more and less experienced decision makers. tainty:BayesianDecisionMakingin RealisticSettings,"Or-
Future research should address some of the opera- ganizational Behavior and Human Performance, 14 (1), 10-
tional limitations of our study. First, the measurement 59.
of marketing experience could be improved upon to cap- Brucks,Merrie(1985), "TheEffectsof ProductClass Knowl-
ture marketing "expertise," rather than merely length of edge on InformationSearch Behavior,"Journal of Con-
sumer Research, 12 (June), 1-16.
experience. For example, comparing the decisions with
an objective criterion for an optimal, or even a good, Chakravarti,Dipankar,AndrewA. Mitchell, and RichardS.
Staelin(1979), "JudgmentBased DecisionModels:An Ex-
decision would provide a measure of decision quality as
perimental Investigation of the Decision Calculus Ap-
well as increase the generalizability of the results. Next,
proach," Management Science, 25 (3), 251-63.
the operationalization of decision programmabilitycould Chi, Michelene,PaulJ. Feltovich,and RobertGlaser(1981),
involve developing experimental tasks that manipulate "Categorizationand Representationof PhysicsProblemsby
the level of programmability. Third, other specific hy- Experts and Novices," Cognitive Science, 5 (2), 121-52.
potheses about the effect of experience on the use of Chiesi, HarryL., George J. Spillich, and James F. Voss (1979),
information and decision making may be in order. For "Acquisition of Domain-Related Information in Relation to
example, our decision-making task does not consider the High and Low Domain Knowledge," Journal of Verbal
configural or nonlinear use of information, which may Learning and Verbal Behavior, 18 (3), 257-73.
increase with experience (Slovic 1969). Finally, one Cochran,William G. and GertrudeM. Cox (1957), Experi-
mental Designs, 2nd ed. New York: John Wiley & Sons,
strengthof our study, the use of practicingmanagersfrom Inc.
a single corporation, is also a drawback: it limits the Deshpande,Rohit and GeraldZaltman(1982), "FactorsAf-
sample size. Some of the specific results may not be fecting the Use of Marketing Research Information: A Path
generalizable outside the corporation we studied; how- Analysis," Journal of Marketing Research, 19 (February),
ever, the conclusions related to programmabilityand ex- 14-31.

This content downloaded from 130.237.29.138 on Sat, 21 Nov 2015 12:30:22 UTC
All use subject to JSTOR Terms and Conditions
10 JOURNAL
OF MARKETING FEBRUARY
RESEARCH, 1990

and - (1983), "Patternsof Research Use in Pri- (1986), "Research Opportunities in the Decision and
vate and Public Sectors," Knowledge: Creation, Diffusion, Management Sciences," Management Science, 32 (Janu-
Utilization, 4 (June), 561-75. ary), 1-13.
and - (1984), "A Comparison of Factors Af- and Leonard M. Lodish (1981), "Commentary on
fecting Researcher and Manager Perceptions of Market Re- 'Judgment Based Decision Models,'" Journal of Marketing,
search Use," Journal of Marketing Research, 21 (Febru- 45 (Fall), 24-9.
ary), 32-9. McIntyre, Shelby H. (1982), "An Experimental Study of the
Ettenson, Richard T. (1984), "A Schematic Approach to the Impact of Judgment-Based Marketing Models," Manage-
Examination of the Search for and Use of Information in ment Science, 28 (January), 17-33.
Expert Decision Making," unpublished dissertation, Kansas Newell, Allen and Herbert A. Simon (1972), Human Problem
State University. Solving. Englewood Cliffs, NJ: Prentice-Hall, Inc.
Fudge, William K. and Leonard M. Lodish (1977), "Evalu- Nisbett, Richard, David H. Krantz, Christopher Jepson, and
ation of the Effectiveness of a Model Based Salesman's Ziva Kunda (1983), "The Use of Statistical Heuristics in
Planning System by Field Experimentation," Interfaces, 8 Everyday Inductive Reasoning," Psychological Review, 90
(1), Part 2, 97-106. (4), 339-63.
Gorry, G. Anthony and Michael S. Scott Morton (1971), "A Payne, John W. (1982), "Contingent Decision Behavior,"
Framework for Management Information Systems," Sloan Psychological Bulletin, 92 (2), 382-402.
Management Review, 13 (1), 55-70. Quelch, John A., Paul W. Farris, and James M. Olver (1987),
Green, Paul E. and Wayne S. DeSarbo (1979), "Componential "The Product Management Audit," Harvard Business Re-
Segmentation in the Analysis of Consumer Trade-offs," view, 65 (2), 30-6.
Journal of Marketing, 43 (Fall), 83-91. Rangaswamy, Arvind, Raymond Burke, Jerry Wind, and Je-
- and V. Srinivasan (1978), "Conjoint Analysis in Con- hoshua Eliashberg (1987), Expert Systems for Marketing.
sumer Research: Issues and Outlook," Journal of Consumer Cambridge, MA: Marketing Science Institute.
Research, 5 (September), 103-23. Simon, Herbert A. (1960), The New Science of Management
Harmon, Paul and David King (1985), Expert Systems: Arti- Decision. New York: Harper & Row Publishers, Inc.
ficial Intelligence in Business. New York: John Wiley & Slovic, Paul (1969), "Analyzing the Expert Judge: A Descrip-
Sons, Inc. tive Study of a Stockbroker's Decision Processes," Journal
Hayes-Roth, Frederick, Donald A. Waterman, and Douglas B. of Applied Psychology, 53 (August), 255-63.
Lenat, eds. (1983), Building Expert Systems. Reading, MA: Soelberg, Peer (1967), "UnprogrammnedDecision Making,"
Addison-Wesley Publishing Company. in Research Toward the Development of Management To-
Hays, William L. (1988), Statistics, 4th ed. New York: Holt, day, Michael P. Hottenstein and R. William Millman, eds.,
Reinhart and Winston. Papers and Proceedings of the 26th Annual Meeting of the
Hogarth, Robin (1980), Judgmentand Choice. New York: John Academy of Management, 3-16.
Wiley & Sons, Inc. Steffik, Mark, Jan Atkins, Robert Blazer, John Benoit, Law-
Holbrook, Morris B. (1981), "Integrating Compositional and rence Birnbaum, Frederick Hayes-Roth, and Earl Sacerdoti
Decompositional Analyses to Represent the Intervening Role (1982), "The Organization of Expert Systems, A Tutorial,"
of Perceptions in Evaluative Judgments," Journal of Mar- Artificial Intelligence, 18 (1), 135-73.
keting Research, 18 (February), 13-28. Sujan, Harish, Mita Sujan, and James R. Bettman (1988),
Hulbert, James M. (1981), "Descriptive Models of Marketing "Knowledge Structure Differences Between More Effective
Decisions," in Marketing Decision Models, R. L. Schultz and Less Effective Salespeople," Journal of Marketing Re-
and A. A. Zoltners, eds. New York: North-Holland Pub- search, 25 (February), 81-6.
lishing Company. Theil, Henri (1971), Principles of Econometrics. New York:
Keen, Peter G. W. and Michael S. Scott Morton (1978), De- John Wiley & Sons, Inc.
cision Support Systems: An Organizational Perspective. Wagner, Richard K. and Robert J. Sternberg (1985), "Prac-
Reading, MA: Addison-Wesley Publishing Company. tical Intelligence in Real-World Pursuits: The Role of Tacit
Larkin, Jill H. (1983), "The Role of Problem Representation Knowledge," Journal of Personality and Social Psychology,
in Physics," in Mental Models, Dedre Gentner and Albert 49 (2), 436-58.
L. Stevens, eds. Hillsdale, NJ: Lawrence Erlbaum Associ- Weitz, Barton A. (1981), "Effectiveness in Sales Interactions:
ates. A Contingency Framework," Journal of Marketing, 45
Larr6ch6, Jean-Claude and Reza Moinpour (1983), "Mana- (Winter), 85-103.
gerial Judgment in Marketing: The Concept of Expertise," -- , Harish Sujan, and Mita Sujan (1986), "Knowledge,
Journal of Marketing Research, 20 (May), 110-21. Motivation, and Adaptive Behavior: A Framework for Im-
Lee, Hanjoon, Frank Acito, and Ralph L. Day (1987), "Eval- proving Selling Effectiveness," Journal of Marketing, 50
uation and Use of Marketing Research by Decision Makers: (October), 174-91.
A Behavioral Simulation," Journal of Marketing Research, Zellner, Arnold (1962), "An Efficient Method of Estimating
24 (May), 187-97. Seemingly Unrelated Regression Equations," Journal of the
Little, John D. C. (1970), "Models and Managers: The Con- American Statistical Association, 57 (June), 348-68.
cept of a Decision Calculus," Management Science, 16
(April), 466-85. Repnnt No. JMR271100

This content downloaded from 130.237.29.138 on Sat, 21 Nov 2015 12:30:22 UTC
All use subject to JSTOR Terms and Conditions

You might also like