Consumer Attitude Metrics For Guiding Marketing Mix Decisions

This article was downloaded by: [128.187.103.
98] On: 10 January 2015, At: 01:53

Publisher: Institute for Operations Research and the Management Sciences (INFORMS)
INFORMS is located in Maryland, USA
Marketing Science
Publication details, including instructions for authors and subscription information:
http://pubsonline.informs.org
Consumer Attitude Metrics for Guiding Marketing Mix

Decisions
Dominique M. Hanssens, Koen H. Pauwels, Shuba Srinivasan, Marc Vanhuele, Gokhan Yildirim
To cite this article:

Dominique M. Hanssens, Koen H. Pauwels, Shuba Srinivasan, Marc Vanhuele, Gokhan Yildirim (2014) Consumer Attitude Metrics
for Guiding Marketing Mix Decisions. Marketing Science 33(4):534-550. http://dx.doi.org/10.1287/mksc.2013.0841
Full terms and conditions of use: http://pubsonline.informs.org/page/terms-and-conditions
This article may be used only for the purposes of research, teaching, and/or private study. Commercial use
or systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisher
approval, unless otherwise noted. For more information, contact permissions@informs.org.
The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitness
for a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, or
inclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, or
support of claims made of that product, publication, or service.
Copyright © 2014, INFORMS
Please scroll down for article—it is on subsequent pages
INFORMS is the largest professional society in the world for professionals in the fields of operations research, management
science, and analytics.
For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org
Vol. 33, No. 4, July–August 2014, pp. 534–550
ISSN 0732-2399 (print) ISSN 1526-548X (online) http://dx.doi.org/10.1287/mksc.2013.0841
© 2014 INFORMS
Consumer Attitude Metrics for Guiding

Marketing Mix Decisions
Downloaded from informs.org by [128.187.103.98] on 10 January 2015, at 01:53 . For personal use only, all rights reserved.
Dominique M. Hanssens
Anderson School of Management, University of California, Los Angeles, Los Angeles, California 90095,
dominique.hanssens@anderson.ucla.edu
Koen H. Pauwels
Özyeğin University, Istanbul, 34794 Turkey, koen.pauwels@ozyegin.edu.tr
Shuba Srinivasan
Boston University School of Management, Boston University, Boston, Massachusetts 02215, ssrini@bu.edu
Marc Vanhuele
HEC Paris, 78351 Jouy-en-Josas, France, vanhuele@hec.fr
Gokhan Yildirim
Management School, Lancaster University, Bailrigg, Lancaster LA1 4YX, United Kingdom,
g.yildirim@lancaster.ac.uk
M arketing managers often use consumer attitude metrics such as awareness, consideration, and preference as
performance indicators because they represent their brand’s health and are readily connected to marketing
activity. However, this does not mean that financially focused executives know how such metrics translate
into sales performance, which would allow them to make beneficial marketing mix decisions. We propose four
criteria—potential, responsiveness, stickiness, and sales conversion—that determine the connection between
marketing actions, attitudinal metrics, and sales outcomes.
We test our approach with a rich data set of four-weekly marketing actions, attitude metrics, and sales
for several consumer brands in four categories over a seven-year period. The results quantify how marketing
actions affect sales performance through their differential impact on attitudinal metrics, as captured by our
proposed criteria. We find that marketing–attitude and attitude–sales relationships are predominantly stable over
time but differ substantially across brands and product categories. We also establish that combining marketing
and attitudinal metrics criteria improves the prediction of brand sales performance, often substantially so.
Based on these insights, we provide specific recommendations on improving the marketing mix for different
brands, and we validate them in a holdout sample. For managers and researchers alike, our criteria offer a
verifiable explanation for differences in marketing elasticities and an actionable connection between marketing
and financial performance metrics.
Keywords: consumer attitude metrics; responsiveness; potential; stickiness; sales conversion; hierarchical linear
model; cross-effects model; empirical generalizations; dynamic programming model; optimal marketing
resource allocation
History: Received: August 14, 2012; accepted: December 10, 2013; Preyas Desai served as the editor-in-chief
and Donald Lehmann served as associate editor for this article. Published online in Articles in Advance
March 7, 2014.
Introduction on financial returns (including the chief financial offi-

Brand managers are urged to compete for the cer (CFO)), but it leaves the deeper reasons for mar-
“hearts and minds” of consumers and often col- keting success or failure unexplored. Inasfar as these
lect brand health indicators such as awareness, lik- reasons change, past sales impact of marketing may
ing, and consideration to this end. These indicators not be the best predictor of its future sales impact.
help understand the state of mind of consumers and Marketers work under the assumption that brand
how marketing affects it. More bottom-line-oriented health indicators are predictive of later marketing and
managers, in contrast, typically assess marketing bottom-line performance but have little guidance on
effectiveness at the observable transaction level, with how a better understanding of this connection can be
measures such as “advertising elasticity” and “return translated into improved decisions on the marketing
on sales.” This practice may satisfy managers focused mix. How actionable is it, for instance, to know that
534
Hanssens et al.: Consumer Attitude Metrics for Guiding Marketing Mix Decisions
Marketing Science 33(4), pp. 534–550, © 2014 INFORMS 535
brand consideration stands at 70% while brand lik- http://dx.doi.org/10.1287/mksc.2013.0841). First, it

ing stands at 40%? Conventional wisdom (e.g., Kotler provides an empirically testable framework on the
and Keller 2012) suggests investing in the “weakest conditions under which consumer attitude move-
link,” i.e., the metric with the most remaining poten- ments result in sales movements. Traditionally, mar-
tial. However, brand liking may have hit its glass keting mix models almost exclusively focus on the
ceiling at 40%, whereas momentum in consideration response of sales to marketing expenditures in order
may still be possible. In addition, consideration could to derive normative implications for marketing bud-
be more responsive to marketing actions than brand get setting. This is not sufficient for the brand man-
liking, and any gains in brand liking may be short- ager interested in quantifying the linkage between
lived because of fickle consumers or tough competi- a firm’s marketing actions, consumer attitude met-
tors; gains in consideration could be longer-lasting. rics, and the brand’s market performance, as con-
As to the end result, consideration gains may con- ceptualized in the brand value chain (Keller and
vert into sales at a higher or lower rate than liking Lehmann 2006). Srinivasan et al. (2010) introduced
gains do. To complicate matters, marketing–attitude mind-set metrics into standard sales response models
and attitude–sales relationships may be generic to the and demonstrated that these metrics indeed improve
category or specific to the brand, indicating competi- sales response models and are advance indicators of
tive (dis)advantage. Finally, these relationships could later sales results. Stahl et al. (2012) made a similar
change over time, obscuring their value and necessi- demonstration with customer lifetime value.
tating their dynamic evaluation to guide future mar- Second, our research objective is normative, aim-
keting mix allocations. ing to use the informational value of these mind-set
In sum, it is no small task for financial and market- variables for improving marketing decision making.
ing managers alike to use consumer attitude informa- As such, our work is fundamentally different from the
tion to guide their marketing strategies and actions. mind-set metric article by Srinivasan et al. (2010).
Yet such guidance is important because managers The objective of this article was to demonstrate that
are charged to allocate marketing resources that pro- the explanatory power of a market response model
vide noticeable and long-lasting improvements in their can be increased by adding mind-set metrics. Vec-
brands’ business performance. Our objective is there- tor autoregressive (VAR) time-series model estimated
fore to provide concrete directions on how the effec- there is descriptive and does not lend itself to norma-
tiveness of marketing mix actions, and therefore the tive inferences such as deriving optimal advertising
allocation of resources, can be improved by exam- spending for a brand. In a VAR model, the marketing
ining attitude metrics. More specifically, we propose (and other) variables are jointly endogenous and their
theory-based criteria on these metrics that identify impact is measured as shock effects, both short term
conditions under which they should be targets of and long term. This approach is useful for descriptive
marketing action. By applying these criteria, man- and forecasting purposes but not for optimal policy
agers with access to the relevant information on the inferences (see Sims 1986 for a detailed discussion in
costs of each marketing instrument can determine an economic policy context). In the spirit of theory
the respective investment appeal of each of these and practice in marketing, our paper builds on these
actions. descriptive inferences to (1) describe criteria for mon-
After a description of our contributions, we begin itoring the evolution of brand mind-set metrics and
by proposing four theory-based criteria for the anal- (2) make brand-specific marketing recommendations
ysis of attitude metrics and show how they can be that are tested in a holdout sample. Because their
operationalized. In the empirical section, we describe objective is fundamentally different, Srinivasan et al.
the data set and demonstrate how the relevant param- do not assess the forecasting accuracy of their models
eters can be estimated. Next, we apply our relevance in a holdout sample.
criteria, first for a diagnostic analysis and then for a Fischer et al. (2011a) proposed a hierarchical deci-
forward-looking analysis using a dynamic program- sion model with mind-set metrics and conversion
ming (DP) analytical model. We conclude with a clas- rates to guide marketing resource allocations in a
sification of brands based on the role that attitude business-to-business (B2B) setting. Their approach is,
metrics play in the connection between marketing unfortunately, not applicable for managers in a typi-
actions and sales. cal business-to-consumer (B2C) context for three rea-
sons. First, it requires individual-level data, with a
Contributions high collection cost (at regular intervals with broad
Our research contributes to the marketing litera- market coverage) for B2C manufacturers. As a result,
ture in four ways (a comparative tabular sum- B2C companies rely on syndicated panel data that
mary with previous work is provided in Web are only available at an aggregate level. Second, their
Appendix C1, available as supplemental material at model assumes a fixed hierarchy-of-effects sequence.
536 Marketing Science 33(4), pp. 534–550, © 2014 INFORMS
Although this is appropriate for their specific B2B Operationalizing the Criteria for
application, where customers follow a standardized Attitude Metrics
purchasing process, such fixed sequences have been Our conceptual framework, displayed in Figure 1,
invalidated in a B2C context (Batra and Vanhonacker contrasts marketing effects that occur through
1988). Third, for their estimations, they develop an changes in attitudinal metrics with those that occur
ad hoc likelihood function without a closed-form without such changes. We denote the former as the
solution. Instead, we use more general econometric
mind-set route and the latter as the transaction route

estimation techniques. Based on our estimates, we in Figure 1. We do not propose that purchases can
develop a general dynamic programming model that occur without the customers’ minds or hearts being
includes sales response (with assumed profit mar- involved (e.g., one needs to be aware of a brand
gins) as the outcome variable and the mind-set and at least right before buying it), but instead that cus-
marketing mix as input variables. We obtain optimal tomers may simply react to a marketing stimulus
marketing resource allocation outcomes and demon- without changing their mind or heart (e.g., the brand
strate their optimality by comparing them to actual was in the consideration set before and remains in
behavior. In these outcomes, we separate marketing the consideration set after a stimulus-induced pur-
effectiveness in a “transaction route” and a “mind-set chase). Our framework therefore accounts for both
route.” generally accepted channels of marketing influence:
A third contribution of the paper is the conceptu- through building the consumer attitudes that consti-
alization of criteria on attitude metrics that identify tute the brand’s health and/or through leveraging the
conditions under which they should be targets of mar- brand’s existing health.
keting action. We offer theoretical foundations and To move from an analysis of attitude metrics to rec-
delineate four key criteria—potential, responsiveness, ommendations for marketing mix decisions, we have
stickiness, and sales conversion—that help us deter- to identify the managerially relevant attitude metrics.
mine and understand the connection between market- Market research firms provide many possible survey-
ing actions, attitudinal metrics, and sales outcomes. based consumer attitudes metrics. However, not all
By applying these criteria, we can determine the mar- of those can be expected to be relevant for marketing
keting investment appeal of each marketing instru- planning for a given brand at a given time. We must
ment. For example, if the sales conversion is the same therefore provide specific relevance criteria for these
for two brands but one of them obtains sales conver- metrics. We propose that relevant attitude metrics
sion with less advertising (i.e., it has higher respon- have potential for growth, are sticky and resistant
siveness), it is possible for that brand to obtain a to competitive erosion, respond to marketing stimuli,
competitive advantage. For managers and researchers and convert into sales.
alike, our criteria, more generally, offer a verifiable Potential as a driver of marketing impact has long
explanation for changing marketing elasticities and an been appreciated and used, especially in the con-
actionable connection between marketing and finan- text of market potential (e.g., Fourt and Woodlock
cial performance metrics. 1960). The central premise is that of diminishing
Finally, our mixed-effects response models of the returns; i.e., the larger the remaining distance to the
link between marketing actions, attitudes, and sales maximum or ceiling, the higher the impact poten-
are well suited for a decision support focus. In partic- tial.1 Fourt and Woodlock (1960) applied this prin-
ular, cross-effects models establish the extent to which ciple to new product penetration forecasting and
the four criteria connecting attitudes to behavior vary found that penetration evolves as a constant frac-
over time and across brands. They also indicate what tion of the remaining distance to the ceiling. Thus
matters more: brand or time variation. In addition, if awareness affects new product trial, then, all else
longitudinal hierarchical linear models (HLMs) examine equal, marketing spending aimed at awareness build-
how marketing–attitude and attitude–sales relations ing will have more impact potential if the beginning
vary by brand. Using our approach, we demonstrate awareness is 20% as opposed to 70%. Not account-
superior results, in terms of both forecast accuracy ing for potential ignores diminishing return effects,
and business performance evolution, from using a resulting in possible overspending with consequently
combined transaction and mind-set approach com-
pared with using only attitudinal (mind-set) or mar- 1
As noted by Haley and Case (1979), over a large range of atti-
keting mix (transaction) models. tude scores, their brand share effect can display increasing returns,
Our work is the first to provide criteria for the at least in self-stated consumer interviews (there are no observed
sales or market share data in their study). However, our study con-
decomposition of sales effects through attitudinal siders only the leading brands in each category, so we do not have
metrics and to offer mind-set-specific guidelines for any observations at the lower end of the attitude scales. Therefore,
improving marketing mix decisions. we do not expect to observe increasing returns to attitude changes.
Figure 1 Conceptual Framework
-ARKETING "RANDHEALTH
ACTIONS INDICATORS &IRM
PERFORMANCE
!WARENESS
!DVERTISING
3TICKINESS
0ROMOTIONS 2ESPONSIVENESS #ONVERSION
#ONSIDERATION
"RANDSALES
$ISTRIBUTION
3TICKINESS
,IKING
0RICE
3TICKINESS
0OTENTIAL
-IND SETROUTE -IND SETROUTE
4RANSACTIONSROUTE
lower returns. It can also result in missed opportu- marketing, as in the decay rate of advertising in
nities on metrics with high potential. The theoret- the Koyck model (Hanssens et al. 2001). Stickiness
ical foundation for “potential” comes from market is a property of the attitudinal metric: if the met-
size considerations, where the “market” is defined ric changes for any reason, how fast does it revert
in attitude space. For example, if 100 individuals back to its mean? The reasons for such changes may
comprise a regional market, and 20 of those have include marketing, competitive marketing, external
prior awareness of brand A, then any marketing cam- shocks, etc. For example, if consumer memory for the
paign aimed at improving brand awareness has the brands in a category is long-lasting, it will take lit-
potential of affecting 80 prospects. As the attitude tle or no reminder advertising for a brand to sus-
score increases and approaches 100%, the “untapped tain a recently gained increase in brand awareness.
market size” shrinks, and holding constant the qual- Similarly, if consumers in a category exhibit strong
ity of the marketing effort, its anticipated impact will habits and routinely choose among the same subset of
lessen accordingly. four brands, then the consideration metric for any of
Potential (POTt ) is operationalized as the remaining these four brands may be sticky. Overall, if a market-
distance to the maximum, preferably expressed as a ing effort increases a brand’s score on a sticky atti-
ratio in light of the multiplicative nature of market tudinal metric, then all else equal, that effort is more
response. For example, if maximum awareness (MAX) likely to have higher returns. The theoretical foun-
is 100% and previous awareness At−1 is 30%, then dation for “stickiness” comes from memory theory
and habit formation theory (Bagozzi and Silk 1983).
POTt = 6MAX − At−1 7/MAX = 0070 (1) The results of experiments in social cognition (e.g.,
Lingle and Ostrom 1979) and consumer behavior (e.g.,
Most consumer attitude metrics are expressed in per- Kardes 1986) suggest that consumers who process
cent (MAX = 100%) or in Likert scales (e.g., 1 to 7, brand attribute information to make evaluative judg-
where MAX = 7), both of which readily accommodate ments will base subsequent brand evaluations primar-
our proposed definition of potential. ily on the recalled judgment rather than the original
Stickiness refers to the staying power as a result of factual information. In the popular associate network
the inertia or lock-in of a change in the attitudi- model of the brain, information is “encoded in long-
nal metric, in the presence of competitive market- term memory as a pattern of linkages between con-
ing. It captures more than the carryover effect of cept nodes” (Burke and Srull 1988, p. 56). However,
brand-related memories may fade over time and as a where the response parameters i now indicate either
result of additional learning (e.g., Baumgartner et al. a concave (i < 1) or an S-shaped (i > 1) response
1983, Belch 1982, Bettman 1979, Schank 1982). Man- curve. The resulting response elasticity i is now con-
agers need to know the extent of stickiness in the tingent on the attitude metric’s potential as follows:
metrics they use; not accounting for stickiness may
result in myopic decision making and possibly waste- i = i × POTt 0 (4)
ful marketing spending.
Responsiveness refers to marketing’s ability to “move For example, in an awareness-to-advertising relation-

the needle” on the attitude metric. In this context, ship with a response elasticity 0.2 at zero initial
different marketing actions will likely have different awareness, the response elasticity will decline to 002 ×
responsiveness. For example, advertising is known to 006 = 0012 when awareness reaches 40%.
be better at inducing trial purchases than repeat pur- Sales conversion indicates that changes in an atti-
chases (Deighton et al. 1994), so an awareness metric tudinal metric convert into sales performance. Sales
may be more responsive to it than a preference met- conversion can be expected to vary in different stages
ric. The theoretical foundation for “responsiveness” of the purchase funnel; e.g., the lower the funnel
comes from utility theory. If a commercial message or stage, the higher the sales conversion. This follows in
an offer provides a net addition to a consumer’s per- general from the hierarchy-of-effects model (even if
ceived utility from purchasing the associated brand, the exact hierarchy may not be fixed; see Batra and
then a response effect is expected. Note that “net” Vanhonacker 1988). For example, a 10% increase in
addition here refers to a comparison of marginal util- advertising awareness may increase sales by only 3%,
ity versus marginal cost. Marginal cost could be price whereas a 10% increase in brand liking may increase
related (e.g., the willingness to pay more for an adver- sales by 6% (Srinivasan et al. 2010). Not account-
tised brand) or habit related (e.g., the inherent cost ing for sales conversion runs the risk of silo mar-
associated with taking a risk and switching to a pre- keting practice, i.e., attitude metrics are viewed as
viously unknown brand). the ultimate performance indicator for marketing, but
Responsiveness is operationalized as the short-term financial executives have no evidence of marketing’s
response of the attitude metric with respect to a mar- impact on cash flows.
keting stimulus. Furthermore, stickiness is the carry- The theoretical foundation for “conversion” comes
over magnitude in the same response equation. We from attitude behavior theory. Although consumers
propose to use a well-established, robust response may have awareness of and attitudes toward several
function to estimate responsiveness. The standard brands in a category, they will not necessarily pur-
multiplicative response model has the advantage of chase all these brands. Since the behavior (purchas-
producing elasticities as responsiveness metrics: ing) is costly, but the attitude is not, attitude behavior

At = cAt−1 X1t1 X2t2 X3t3 etu 1 (2) conversion will only occur when an attitude change
is sufficiently strong. For example, a consumer may
where A is the attitudinal metric of awareness as report increased ad awareness, consideration, and/or
an example and Xi (i = 11 21 3) are marketing instru- liking for a brand (e.g., because of a new ad cam-
ments. Not only does a multiplicative response model paign) but fail to purchase it in the supermarket for a
provide readily interpretable results, it has also been host of reasons, including a high price or habit inertia
shown to outperform more complex specifications for the previously purchased brand.
in forecasting product trials for consumer packaged Conversion rates are typically well below unity;
goods (CPG) (e.g., Hardie et al. 1998). Stickiness is for example, Jamieson and Bass (1989) reported ratios
measured through the carryover parameter . For of actual versus stated consumer trial in 10 prod-
example, if = 006, this means that 60% of any change uct categories ranging from 0.009 to 0.896, averaging
in At is carried over to the next period. We expect around 0.5. When historical data are available, conver-
to be less than 1 because of memory decay effects sion metrics may be estimated from a “funnel” model,
that are well documented in psychology (Baddeley with metrics such as awareness and preference or lik-
et al. 2009).
ing. Logically, one might expect that there is a hierar-
Note that responsiveness may be related to poten-
chy among mind-set metrics and that awareness, for
tial as follows: the closer the attitude metric is to
instance, leads to consideration. Research on the hier-
its ceiling value, the more difficult it will be to
archy of effects, however, shows that evidence on the
register further increases through marketing. That
exact sequence of effects is mixed (Franses and Vriens
phenomenon is readily incorporated in (2) by express-
2004, Vakratsas and Ambler 1999). As such, we do not
ing the dependent variable as an odds ratio (e.g.,
want to impose a hierarchy of effects, because there is
Johansson 1979):
little support for such fixed unidirectional hierarchies
0
A0t = At /4MAX − At 5 = cAt−1 X1t1 X2t2 X3t3 etu 1 (3) (e.g., Batra and Vanhonacker 1988, Norris et al. 2012).
Instead, we allow for a multiplicative funnel model marketing mix, consumer attitude metrics (based on
that can be applied across conditions. For example, 8,000 households in France), and performance metrics
with intermediate attitudinal metrics awareness (A), across brands in each category on a four-week basis.
consideration (C), and liking (L), a multiplicative fun- For the period between January 1999 and May 2006,
nel model for sales revenue (S) would be we analyze data for the six major brands in each

St = cSt−1 At−1
1
Ct−1
2
Lt−1
3
etu 0 (5) of these four categories: bottled juice, bottled water,
cereal, and shampoo. The broad nature of our data set
Conversion models such as (5) can be tested either allows us to investigate whether the extent to which
with longitudinal or with mixed cross-sectional time- attitude metrics affect sales varies across brands and
series data.2 products. Specifically, as a first validation of our
How do the proposed criteria relate to traditional model, we verify whether sales conversion from atti-
notions of short-term and long-term marketing elas- tudes into purchase behavior differs between higher
ticity? Short-term marketing-sales elasticity is a combina- versus lower involvement purchase situations within
tion of the marketing responsiveness and the potential the studied fast-moving consumer goods. Nelson
and sales conversion of each metric. Our decomposi- (1970) developed an economic perspective classifying
tion allows managers to assess whether, for instance, a brand purchase decision as either low involvement,
low short-term elasticity is due to low marketing
where trial is sufficient, or high involvement, where
responsiveness versus low inherent potential versus
information search and conviction are required prior
low sales conversion of a metric. Stickiness corre-
to purchase. When product involvement is high, a
sponds to the carryover of marketing effects, so adding
this to the other criteria constitutes long-term market- brand needs to change consumers’ hearts and minds
ing elasticity. As a special case, permanent marketing- to overcome consumers’ reluctance to change their
sales effects (Dekimpe and Hanssens 1999) arise when purchase behavior (Bauer 1967, Peter and Tarpey
a marketing action succeeds in increasing a sales- 1975). In such cases, we expect movements in atti-
converting metric that has a stickiness of 1. Finally, tudinal metrics to be strongly associated with sales
our decomposition across metrics allows managers to (i.e., there is sales conversion). In contrast, when prod-
assess whether a given marketing sales elasticity is uct involvement is low, consumers may choose a
driven by the mind-set route through awareness, con- brand simply because it is available or promoted,
sideration, or liking. without having fundamentally changed their opin-
In conclusion, marketing may influence consumer ion about it. This low involvement path is compatible
attitudes, and this, in turn, may improve the brand’s with Ehrenberg’s (1974) awareness-trial-reinforcement
business performance. The degree to which this will model. In such cases, we expect low sales conversion.
occur depends on the nature of the category (for Marketing actions may have a direct impact on
example, low versus high consumer involvement) sales without affecting the attitudinal metrics; this
and on the potential, stickiness, responsiveness, and is called the transaction route in Figure 1. In our
conversion of the attitude metrics. Each of these cri- data set, involvement is measured at the category
teria is supported by either consumer behavior or level through several items, including “product cat-
market response theory. By combining these scores, a egory X is important; you have to be careful when
brand may obtain an a priori indication of how effec- choosing a product.” The results show that shampoo
tive different marketing campaigns are likely to be. (37.8%) is more involving than the food and bever-
In what follows, we apply our framework for differ- age categories juice (29.8%), cereal (28.4%), and bot-
ent brands in multiple categories varying in consumer
tled water (28.2%),3 which have minimal variation.
involvement level.
The focal brand performance measure is sales vol-
ume4 aggregated across all product forms of each
Product Categories, Data, and brand (in milliliters or grams). The marketing mix
Modeling data include average price paid, value-weighted dis-
The data come from a brand performance tracker tribution coverage, promotion, and total spending on
developed by Kantar Worldpanel, which reports the advertising media.
2
To avoid correlated errors stemming from the use of mind-set
3
variables both as dependent variables (Equation (3)) and as predic- We show the percentge agreeing that a buyer has to pay close
tors, we used the first lagged mind-set variables as predictors in the attention to the product chosen.
4
“sales conversion” model (Equation (5)). This approach is intuitive, Although the actual measure of brand performance is purchases,
as attitudes logically precede choices; i.e., consumers enter the pur- as registered by consumers, and not sales, as registered by stores,
chase occasion with a preexisting attitude, then combine that with we use the word “sales” in the remainder of the paper. Future
the prevailing marketing mix conditions (prices, promotion, etc.) research should include actual market-level sales data as a depen-
and make a decision. dent variable, particularly if the emphasis is on resource allocation.
After a discussion with the data provider, we Econometric Modeling

selected the following three measures from the avail- Our empirical setting covers multiple brands in four
able attitudinal metrics: advertising awareness,5 inclu- different categories over time. Thus we face some
sion in the consideration set, and brand liking. This critical questions about the stability and specificity
selection aimed at covering the three main stages of of the relationships we seek to estimate. In particu-
the purchase funnel. The first two measures refer to lar, we need to test whether attitude stickiness and
the cognitive status of a brand in the consumer’s sales conversion are stable over time or are idiosyn-
mind, whereas brand liking obviously refers to the cratic to certain time periods. In addition, we need
affect status. Two other available measures were not to establish whether different brands experience dif-
included for lack of variation, aided brand aware- ferent marketing–attitude response effects or if the
ness (which exhibited ceiling effects) and collinearity effects are generic to the product category. These dis-
(intention to purchase correlated highly with the con- tinctions are not only econometrically important, they
sideration set, and the data provider considered the also have different strategic implications. For exam-
latter to be managerially more useful). ple, if the attitude-to-sales conversion parameters,
For advertising awareness, survey respondents including competitive actions, are found to be simi-
indicated, in a list of all brands present on the market, lar across brands, then no single brand can claim a
those for which they “remember having seen or heard competitive sales advantage from lifting an attitude
advertising in the past two months.” Our measure metric, though a brand can achieve an advantage if it
gives the percentage of respondents who were aware. can lift an attitude metric efficiently.
For the consideration set, respondents were asked to Our response models focus on brand-level response
indicate “the brands that you would consider buying” to marketing and attitudinal variables. Naturally,
from a list of all brands in the market. We use the
these actions and attitudes take place in a competi-
percentage of respondents who consider buying as
tive environment. Our parameter estimates implicitly
the relevant measure. Liking is measured on a seven-
control for competitive effects because the dependent
point scale (from “like enormously” to “not at all”),
variables are real-world observations. Thus, brand
and the measure we use is the average rating.6 More
sales lifts are implicitly bounded by total category
details on these data sources are described in Srini-
demand and consumer loyalty for competing brands.
vasan et al. (2010).
We choose not to include formal competitive mod-
With a time sample of more than seven years, the
presence of different players with different strategies els (e.g., equations reflecting individual competitive
in different product categories, and wide coverage behavior and/or market share models) because the
of the marketing mix as well as consumer attitudi- sectors under study (shampoo, cereal, fruit juice,
nal metrics, these data are uniquely suited to address and bottled water) are mature, with several impor-
our research questions. The country of investigation is tant brands competing in each time period. There is
France, which is more homogeneous than large mul- always at least one brand promoting or advertising;
ticultural markets such as the United States in terms thus competition is stable at the category level. Fur-
of consumer behavior and retail industry structure. thermore, research in similar markets (see Steenkamp
et al. 2005 for a study on more than 1,200 Euro-
5
Whereas awareness typically means “brand awareness” in mar- pean CPG brands) has shown that the predominant
keting theory, recent empirical studies (Lautman and Pauwels 2009, competitive reaction in advertising or promotion is
Pauwels et al. 2013, Srinivasan et al. 2010) have shown that adver- no reaction at all. We leave the formal inclusion of
tising awareness is a key driver of sales across different indus- brand-level competition as an area for future research.
tries (e.g., drugs, food, drinks, health, beauty) and countries (e.g.,
Table 1 contains an overview of the econometrics
the United States, the United Kingdom, France, Brazil). Intuitively,
advertising awareness is important because consumers are exposed models used in estimation.
to hundreds of ads daily and can recall only a fraction of them These models are a combination of attitude and
(Burke and Srull 1988). Moreover, previous studies found that sales response and are estimated as mixed-effects
advertising awareness also increased with factors other than adver- models. This allows us to combine fixed and ran-
tising (Table 6 in Srinivasan et al. 2010). Thus, advertising aware-
dom effects to separate and investigate how each
ness may be a proxy for brand salience (Tulving and Pearlstone
1966). Because our data do not contain measures for brand salience, level affects the attitude criteria (see Web Appendix A
we leave its possible connection with advertising awareness as a for a technical explanation and model specification
promising area for future research. choices). First, cross-effects (CRE) models allow ran-
6
The asymmetry in the brand liking versus brand consideration dom effects to vary both by brand and over time
and ad awareness scales is typical for commercially available atti- (Baltagi 2005). A typical operationalization is within
tude measurement. However, it may partially explain high average
a cross-sectional or dynamic panel where the cross-
scores for brand liking (as the typical respondent answers the ques-
tion for each brand) compared with ad awareness and considera- sectional dimension (brand, market, etc.) is crossed
tion (because the respondent needs to put the effort into clicking with the dynamic factor “time.” In this study’s con-
on the brand from a list of all other brands in the category). text, CRE models enable us to establish the extent
Table 1 Overview of Metrics and Models
Metrics/models Equation Model Table
Potential metric Equation (1) POTt = 6MAX − At−1 7/MAX Table 8

0
Stickiness metric Equation (3) A0t = At /4MAX − At 5 = cAt−1 X1t1 X2t2 X3t3 eut Table 8
0
Responsiveness model Equation (3) A0t = At /4MAX − At 5 = cAt−1 X1t1 X2t2 X3t3 eut Table 4
1 2 3
eut
Conversion model (mind-set route) Equation (5) St = cSt−1 At−1 Ct−1 Lt−1 Table 3

Marketing mix model (transactions route) Equation (2) applied to sales St = cSt−1 X1t1 X2t2 X3t3 eut Table 5
a 4 5 6 4 5 6 u
Transactions + Consumer attitude model Equation (6) St = cS t−1 X X X A
1t 2t 3t t−1 C L e
t−1 t−1 t Table 6
a
Equation (6) combines Equations (2) and (5) and has both marketing mix and attitudinal metrics.
to which the four criteria on translation of attitudes and varying slope model. The information criteria
to behavior vary over time and across brands. Sec- (Akaike’s information criteria and Bayesian informa-
ond, longitudinal hierarchical linear models (HLMs) tion criteria) support the latter.8 To obtain the brand-
enable us to investigate how marketing–attitude and level effects, we combine fixed and random effects at
attitude–sales relations vary by brand. We can there- the brand level. As a diagnostic check, we perform
fore assess whether higher involvement scenarios normal Q–Q plots for the standardized residuals.
imply higher responsiveness, stickiness, and sales We find no violation of the normality assumption.9
conversion of attitude metrics. Moreover, the longi-
tudinal HLM separates the variance of an outcome Mediation Test. To test our overall framework
variable into “among” and “within” variances, which (summarized in Figure 1), we conduct a formal medi-
increases the precision of estimates (Rabe-Hesketh ation analysis, using the Sobel–Goodman mediation
and Skrondal 2005). test (Sobel 1982) to determine whether a mediator
We estimate the CRE and longitudinal HLMs on (e.g., attitudinal metric) carries the influence of an
logistic-transformed (in Equation (3) in Table 1) or independent variable (e.g., marketing action) to the
log-transformed (all other equations in Table 1) data. dependent variable, sales. Full mediation would indi-
To select the final empirical model, we perform cate that the attitudinal metrics benchmark model
several tests based on our HLM and CRE model (without marketing mix) is sufficient to predict sales,
formulations (see Equations (A1) and (A4) in Web as the “transaction” route of marketing influence is
Appendix A). We discuss the results of these tests fully subsumed in the “mind-set” route. On the other
in the ensuing section in detail. Because of the large hand, no mediation would indicate that the marketing
number of equations and parameters that were esti- mix benchmark model (without attitudinal metrics) is
mated, we present only a few illustrative tables and appropriate. Finally, partial mediation would suggest
graphs. A full set of econometric results may be found that the full model with both marketing mix and atti-
in the Web appendix. tudinal metrics is superior because it acknowledges
both transactions and mind-set routes of influence.
The Sobel–Goodman tests obtained using the HLM
Estimation Results
estimation results revealed evidence of partial media-
Model Testing tion (see Web Appendix C2 for results), leading us to
The CRE and HLM estimations across 24 brands in conclude in favor of the full model with both market-
four categories allow us to make several generaliza- ing mix and attitudinal metrics, as shown in Figure 1.
tions on the four criteria that govern the attitude-to-
sales relationship. For each of the HLMs and CRE Endogeneity Test. The model we use is subject to
models, we test whether a mixed-effects specifica- a potential endogeneity issue as a result of lagged
tion with both fixed and random effects is superior regressors and the brand-level random-effects term.
to a conventional regression with fixed effects only. To deal with this, we use the dynamic panel model
Tables 2–8 report the main results. developed by Arellano and Bond (1991), using a gen-
As Tables 3–6 show, the likelihood ratio (LR) test eralized method of moments (GMM) estimator that
results are significant for all models, justifying the use relies on constructing moment conditions based on
of HLMs and CRE models.7 We also compare (i) the
varying intercept model and (ii) the varying intercept 8
The information criteria statistics are available from the authors
upon request.
7 9
The LR test results for CRE models are available from the authors The normal Q–Q plots are available from the authors upon
upon request. request.
the lagged variables.10 The results indicate that the the time variation within the brand variation in lon-
dynamic panel estimates are similar to those of longitudinal HLMs to investigate the magnitude of the
gitudinal HLMs in terms of sign and significance.11 marketing–attitude and attitude–sales relationships,
We conclude that accounting for endogeneity does with a view to understanding the nature of compet-
not impact our substantive findings, and we report itive brand advantage. The longitudinal HLM results
the HLM results. These estimates offer superior prop- for sales conversion and responsiveness are shown in
erties over dynamic panel estimation, including the Tables 3 and 4, respectively. We observe differences
ability to detect to what extent the groups (brands) across brands but also note general patterns regard-
vary in the intercept and/or slope parameters and ing attitude criteria, which can be grouped in two key
the ability to obtain estimates for the category level sets of findings.
brand level simultaneously. Furthermore, the variance
Sales Conversion vs. Stickiness for Liking vs.
of the dependent variable is split into “between” and
Awareness and Consideration. Table 3 shows the
“within” variances, which increases the precision of
liking-to-sales conversion elasticities for each cate-
estimates.
gory. The median across all brands is 0.549, imply-
Generalizations About Attitudes and ing that sales move approximately with the square root
Their Sales Conversion of liking. This affect conversion is more than three
We organize our findings around the four criteria of times the cognitive conversion of consideration for
stickiness, conversion, marketing responsiveness, and all categories. However, as shown in Table 4, con-
potential. sumer liking has two less desirable characteristics for
brands. First, it is less sticky than the cognitive atti-
Sales Conversion Is Predominantly Stable Over tude metric of awareness. Across the four categories,
Time. The CRE model results reveal that brand the median level of stickiness for awareness is 0.499,
variation is more important than time variation in whereas that for consideration is 0.238 and for lik-
attitude-to-sales conversion models. Table 2 reports ing is 0.234. Second, as shown in Table 4, brand lik-
the percentage variation due to brands and time for ing is responsive to advertising in only two of the
all these models. Except for the cereal category, we four categories, whereas consideration is responsive
observe that the variation in estimates is more brand to advertising in all but one category. Advertising
specific than time specific. This result highlights the moves the needle on all three mind-set metrics of
benefit of strong consumer attitudes favoring a brand awareness, consideration, and liking; 9 of the 12
and resulting in sales conversion.12 coefficients are statistically significant. Price moves
Brand-Specific Attitude Responsiveness to Mar- the needle on awareness, consideration, and liking,
keting Dominates. Turning to the effects of market- though only 4 of the 12 responsiveness coefficients
ing actions on attitude metrics, Table 2 shows they are statistically significant. Promotion influences con-
are also more brand specific than time specific. Atti- sideration only for the shampoo category. Liking has
tude responsiveness is typically specific to the brand high sales conversion ranging from 0.403 to 0.926,
and stable over time. This is especially pronounced whereas consideration has lower sales conversion:
in the shampoo category: for each attitude metric, the consideration conversions range from 0 to 0.193 (see
vast majority (63%–80%) of responsiveness variation Table 3). Note that the highest conversions to sales
is due to brands. from both consideration (0.193) and liking (0.926) are
Combining all results from the CRE models, we in the shampoo category, which is higher in consumer
find that attitude criteria are predominantly stable involvement than the others.
over time but vary substantially across brands within These results may be explained by consumer behav-
the same category.13 Therefore, we proceed by nesting ior theory. Feldman and Lynch’s (1988) accessibility–
diagnosticity framework predicts that a given piece
10
Their approach is as follows: first, take first differences of both of information “will be used as an input to a sub-
sides and eliminate the group effects (ui ), then look for instrumen- sequent response if the former is accessible and if
tal variables (lagged variables), and finally, use GMM to estimate
the model. We implemented these steps and compared the results
of the dynamic panel GMM estimations with the results of longi- and market share. Our data only allow us to check the correla-
tudinal HLMs. tion of the latter variable with the attitude criteria across brands.
11 The only substantial correlation is that between market share and
These are available from the authors upon request.
advertising awareness conversion (0.29). Thus, large brands tend
12
The statistical results and the time-series plots highlighting these to have a higher sales conversion of ad awareness. A plausible
findings from the CRE models are available from the authors upon rationale is that, once they notice a brand’s advertising, consumers
request. have an easier time remembering, finding, and buying larger versus
13
These variations across brands may be due to several factors such smaller brands. This is consistent with the double jeopardy effect
as the quality of the product experience, past market spending, (Ehrenberg et al. 1990).
Table 2 Variation Across Brands and Time in Attitude Responsiveness to Marketing and Sales Conversion: CRE
Category Awareness responsiveness (%) Consideration responsiveness (%) Liking responsiveness (%) Sales conversion (%)
Shampoo
Brand variation 62048 70086 80022 15014
Time variation 10052 3075 4053 6047
Bottled water
Brand variation 44007 76076 64009 71071

Time variation 12045 4029 4004 6030
Juice
Brand variation 56061 58015 54089 22093
Time variation 5076 8015 5006 4074
Cereal
Brand variation 32056 53012 30082 6035
Time variation 15025 4053 25049 17081
Note. From the cross-effects model output in the Web appendix, read as follows: “Of the total variation in the awareness responsiveness model in the shampoo
category, 62.48% is due to brands, 10.52% due to time, the remainder (27.00%) is residual variation.”
Table 3 Maximum Likelihood Fixed Effects Estimates of Sales Conversion in Longitudinal HLM
Shampoo Bottled water Juice Cereal
Coefficient SE p > z Coefficient SE p > z Coefficient SE p > z Coefficient SE p > z
Fixed effects
Constant −10722∗ 00414 00000 00273 00384 00477 −10271∗ 00344 00000 −00617∗ 00258 00017
AR(1) 00569∗ 00035 00000 00678∗ 00031 00000 00711∗ 00030 00000 00743∗ 00028 00000
Awareness (t − 1) 00011 00056 00845 00013 00029 00667 00055 00032 00080 00006 00032 00844
Consideration (t − 1) 00193∗ 00085 00023 −00017 00059 00773 00193∗ 00061 00002 00145∗ 00052 00002
Liking (t − 1) 00926∗ 00414 00000 00403∗ 00191 00035 00641∗ 00201 00001 00457∗ 00152 00003
LR test 42 = 35064, p > 2 = 0000 42 = 33040, p > 2 = 0000 42 = 36096, p > 2 = 0000 42 = 14061, p > 2 = 0001
Notes. The dependent variable is sales. t − 1 indicates one lag is taken. Because of space limitations, random effects estimates are in Table C3 in the Web
appendix.
∗
Statistically significant effects at p < 0005.
it is perceived to be more diagnostic than other acces- and 72%, respectively, whereas the potential for the
sible inputs” (p. 431). The greater the accessibility affect metric of liking averages 19% (see Table 8). For
and diagnosticity of an input for a judgment relative instance, some consumers who are not considering a
to alternate inputs, the greater the likelihood that it brand may well have tried it and not liked it, result-
will be used (Simmons et al. 1993). In high involve- ing in higher potential for consideration relative to
ment categories, such as shampoo, attitudinal changes liking. This suggests that, all else equal, brands have
in consideration make the consumer’s brand expe- higher opportunity to make progress on cognitive
rience diagnostic and accessible resulting in higher metrics. Thus consumer satisfaction (“liking”) runs
sales conversion. Purchases of low involvement prod- high across brands, indicating high product quality,
ucts, on the other hand, are not preceded by signifi- and consequently, the marketing challenges for indi-
cant attitude change, particularly as it pertains to the vidual brands have more to do with their progress in
cognitive attitudinal metrics of awareness and con- the cognitive metrics.
sideration. This shows the limitation of relying only
on attitudinal response for making marketing impact
Assessing Managerial Relevance
inferences. Even when marketing succeeds in lift-
ing an attitudinal metric, it does not imply that this Prediction Test. Given that additional costs are
specific attitude metric, in turn, converts into sales. involved in the collection of attitudinal data, man-
Accounting for the full chain reaction of events allows agers will want to ensure that these data improve the
for an actionable connection between marketing and accuracy of sales forecasts, conditional on their mar-
financial performance metrics. keting plans. We assess these improvements by com-
Attitude Potential Is Higher for Cognitive than paring conditional forecast results for the monthly
for Affect Metrics. The cognitive metrics of aware- observations of periods 85–96, where the brand’s mar-
ness and consideration have higher potential of 73% keting mix decisions for those periods are known
Table 4 Maximum Likelihood Fixed Effects Estimates of Attitude Responsiveness in Longitudinal HLM
Model 1 (DV = Awareness) Model 2 (DV = Consideration) Model 3 (DV = Liking)
Coefficient SE p > z Coefficient SE p > z Coefficient SE p > z
Shampoo
Fixed effects
Constant −00575∗ 00112 00000 −10055∗ 00132 00000 10132∗ 00179 00000
AR(1) 00431∗ 00036 00000 00228∗ 00040 00000 00194∗ 00042 00000
Price −00269∗ 00114 00018 −00086 00087 00323 −00071 00100 00481
Promotion 00028 00016 00075 00036∗ 00017 00031 00007 00021 00747
Advertising 00010∗ 00004 00009 00002∗ 00001 00009 00001 00001 00225
LR test 52 = 110082, p > 2 = 0000 52 = 174010, p > 2 = 0000 52 = 197043, p > 2 = 0000
Bottled water
Fixed effects
Constant −10050∗ 00160 00000 −00881∗ 00408 00031 00623∗ 00226 00006
AR(1) 00567∗ 00031 00000 00247∗ 00040 00000 00274∗ 00040 00000
Price −00331∗ 00137 00016 −00275 00266 00302 −00552∗ 00160 00001
Promotion 00033 00021 00111 −00002 00017 00889 00010 00026 00697
Advertising 00013∗ 00002 00000 00003∗ 00001 00003 00004∗ 00002 00029
LR test 52 = 62044, p > 2 = 0000 52 = 175082, p > 2 = 0000 52 = 149041, p > 2 = 0000
Juice
Fixed effects
Constant −00503∗ 00113 00000 −00816∗ 00288 00005 00932∗ 00228 00000
AR(1) 00667∗ 00027 00000 00382∗ 00037 00000 00559∗ 00033 00000
Price −00093 00096 00332 00053 00163 00743 −00108 00179 00546
Promotion 00014 00037 00702 00049 00048 00307 00001 00056 00996
Advertising 00009∗ 00002 00000 00002∗ 00001 00054 00004∗ 00002 00028
LR test 52 = 80027, p > 2 = 0000 52 = 133018, p > 2 = 0000 52 = 83002, p > 2 = 0000
Cereal
Fixed effects
Constant −00673∗ 00197 00001 −00621∗ 00129 00000 10388∗ 00144 00000
AR(1) 00297∗ 00038 00000 00067 00042 00110 00109∗ 00041 00008
Price 00700 00491 00154 −10094∗ 00276 00000 00185 00233 00427
Promotion 00067 00037 00072 00043 00023 00064 −00021 00030 00489
Advertising 00008∗ 00003 00004 00000 00003 00889 00001 00004 00749
LR test 52 = 140005, p > 2 = 0000 52 = 208007, p > 2 = 0000 52 = 126072, p > 2 = 0000
Notes. Because of space limitations, random effects estimates are in Table C4 in the Web appendix. DV, dependent variable.
∗
(i.e., planned) at the end of period 84. The benchmark sales predictions made by the combined “marketing
forecasts are obtained from the marketing mix mod- mix and attitudinal metrics” models outperform the
els (without attitudinal metrics) reported in Table 5, benchmark forecasts obtained using the model with
as well as from the attitudinal metrics model (with- only attitudinal metrics or the model with only mar-
out marketing mix) reported in Table 3. The com- keting mix in all cases. The combined model offers
parison forecasts are obtained from full models with sizeable improvements in prediction: across categories
both marketing mix and attitudinal metrics reported and forecast horizons, the average MAPE for the
in Table 6. These models thus allow marketing actions attitude model is 15.7%, for the marketing mix-only
to have both transaction and mind-set route effects on model is 17.7%, and for the combined model is 12.0%.
sales. As can be expected, the sales prediction improve-
We proceed with comparisons that are based on ments for one-step forecasts are lower because these
one-step ahead and multistep forecasts, i.e., projec- are more accurate in the benchmark models.
tions up to 12 periods ahead. Although the one- Marketing Mix Scenarios Test. The brand speci-
step forecasts are expected to be more accurate, the ficity of results suggests that individual brands face
multistep predictions are more realistic and strategi- unique circumstances that should govern their mar-
cally valuable in a 12-month marketing planning sce- keting moves. Therefore, in theory, we could perform
nario. Table 7 shows the comparative results, with a formal optimization of marketing mix spending by
focus on prediction accuracy, as measured by mean brand and by period, as done, for example, by Fischer
absolute percentage error (MAPE). Importantly, the et al. (2011a) in the global express delivery sector.
Table 5 Maximum Likelihood Fixed Effects Estimates of Marketing Mix Models (Transaction Route) in Longitudinal HLM
Fixed effects
Constant 00310∗ 00121 00011 00611∗ 00158 00000 00276∗ 00093 00003 00901∗ 00244 00000
AR(1) 00530∗ 00036 00000 00717∗ 00030 00000 00765∗ 00026 00000 00478∗ 00034 00000
Price −00253∗ 00106 00017 −00446∗ 00085 00000 −00183∗ 00071 00010 −00018 00374 00961
Promotion 00113∗ 00018 00000 00034∗ 00016 00032 00094∗ 00022 00000 00094∗ 00015 00000
Advertising 00005∗ 00001 00001 00003 00002 00139 00004∗ 00001 00000 00009∗ 00004 00020
LR test 42 = 38075, p > 2 = 0000 42 = 61024, p > 2 = 0000 42 = 21079, p > 2 = 0000 42 = 89043, p > 2 = 0000
Notes. The dependent variable is sales. Because of space limitations, random effects estimates are in Table C5 in the Web appendix.
∗
Table 6 Maximum Likelihood Fixed Effects Estimates of Transactions + Consumer Attitude Models in Longitudinal HLM
Fixed effects
Constant −10443∗ 00402 00000 −00026 00343 00939 −10460∗ 00280 00000 −00610∗ 00309 00048
AR(1) 00420∗ 00036 00000 00658∗ 00031 00000 00658∗ 00029 00000 00503∗ 00032 00000
Price −00366∗ 00096 00000 −00443∗ 00111 00000 −00202∗ 00050 00000 −00209 00266 00433
Promotion 00127∗ 00016 00000 00038∗ 00019 00048 00098∗ 00020 00000 00088∗ 00015 00000
Advertising 00005∗ 00001 00000 00004∗ 00002 00040 00004∗ 00001 00000 00009∗ 00003 00003
Awareness (t − 1) 00035 00054 00510 00080 00050 00108 00086∗ 00024 00000 00033 00041 00417
Consideration (t − 1) 00166∗ 00079 00050 00009 00058 00876 00156∗ 00056 00006 00168∗ 00069 00014
Liking (t − 1) 00897∗ 00261 00001 00317 00189 00092 00674∗ 00176 00000 00426∗ 00138 00002
LR test 72 = 36013, p > 2 = 0000 72 = 47089, p > 2 = 0000 72 = 22092, p > 2 = 0001 72 = 82010, p > 2 = 0000
Notes. The dependent variable is sales. t − 1 indicates one lag is taken. Because of space limitations, random effects estimates are in Table C6 in the Web
appendix.
∗
Such an exercise requires the use of brand-specific spending scenarios. Using our framework, we diag-
cost and profit margins as well as a clear under- nose the brands at the beginning of the holdout
standing of each brand’s business objective (for exam- period and offer recommendations for changes in the
ple, share gains versus profit maximization). Absent marketing mix, i.e., should the brand’s preexisting
such financial and strategic information in the current marketing support be increased, maintained, or cut,
application, we will provide diagnostic information with the goal of increasing sales. Then we compare
for several brands, based on a simulation of different their business outcomes in function of their actual
marketing spending decisions. For each category, we
choose the two top selling brands: SA and SB in
Table 7 Predictive Performance (MAPE) for the Combined Model vs.
shampoo, WA and WB in bottled water, JA and JB
the Consumer Attitude and Marketing Mix Models (Holdout
Sample: Periods 85–96) in juice, and CA and CB in cereal. Leaving periods
85–96 of our data as a holdout sample, we summa-
Forecast Consumer Marketing Combined rize the brands’ market positions in time period 84.
solution Category attitude model (%) mix model (%) model (%)
As shown in Table 8, we estimate individual brand-
One-step Shampoo 27063 29028 26037 level response models for these focal brands and
ahead Bottled water 6062 5005 3031 examine the shifts in marketing spending that these
Juice 9033 9092 9002 brands experienced in periods 85–96 to draw conclu-
Cereal 5086 7054 3081
sions regarding marketing mix decisions.
Multistep Shampoo 33013 37077 26050
As an example of brand diagnostics, shampoo
ahead Bottled water 16067 15069 10079
Juice 16098 23032 9005 brand SA has ample room for mind-set expansion
Cereal 9074 12084 7024 across the board: awareness is 27%, consideration
17%, and liking 71% (5 of 7); potential for aware-
Notes. One-step ahead forecasts update each consecutive period, whereas
multistep forecasts predict 1–12 periods ahead without updating. MAPE ness is therefore 73%, consideration 83%, and liking
denotes the mean absolute percentage error over the 12-month forecast 29%. As a result, the brand’s prospects in attitudi-
period. nal space are high, especially when compared with its
546
Table 8 Diagnostics for the Two Top Brands in Each Category, T = 84
Awareness
SA SB WA WB JA JB CA CB
Potential (%) = 73 75 79 64 84 62 64 86
Stickiness = 0.35 0.33 0.48 0.48 0.56 0.57 0.30 0.30
Responsiveness Coeff. SE p > z Coeff. SE p > z Coeff. SE p > z Coeff. SE p > z Coeff. SE p > z Coeff. SE p > z Coeff. SE p > z Coeff. SE p > z
Advertising 00141 00056 00012 00095 00074 00202 00213 00088 00015 00194 00088 00027 00007 00044 00879 00021 00044 00624 00513 00147 00001 20231 00222 00000
Promotion 00027 00020 00165 00035 00022 00117 00034 00008 00000 00034 00022 00125 00007 00039 00867 00130 00040 00001 00104 00038 00006 00148 00054 00006
Price −00259 00114 00023 −00273 00114 00017 −00332 00019 00000 −00328 00137 00016 −00146 00105 00181 −00137 00105 00195 10191 00241 00000 00700 00491 00154
Sales conversion 00030 00059 00611 00091 00010 00000 00052 00048 00278 00072 00039 00062 00048 00036 00179 00031 00036 00390 00030 00008 00000 −00001 00034 00978
Consideration
Potential (%) = 83 71 50 70 69 81 73 80
Stickiness = 0.16 0.16 0.24 0.24 0.27 0.27 0.03 0.03
Advertising 00002 00015 00915 00004 00015 00806 10079 00219 00000 00788 00227 00001 00146 00200 00465 10237 00298 00000 00434 00121 00000 00721 00176 00000
Promotion 00073 00020 00000 00044 00027 00112 00001 00025 00980 00016 00026 00541 00056 00077 00469 00145 00053 00006 00067 00027 00013 00092 00036 00011
Price −00086 00087 00323 −00086 00087 00323 −00275 00266 00302 −00275 00266 00302 00193 00287 00503 00198 00287 00490 10370 00076 00000 10094 00276 00000
Sales conversion 00198 00089 00027 00288 00010 00000 00001 00068 00988 00065 00013 00000 00214 00063 00001 00192 00063 00002 00187 00006 00000 00136 00051 00008
Liking
Potential (%) = 29 11 2 22 9 27 26 24
Stickiness = 0.14 0.14 0.25 0.25 0.36 0.36 0.09 0.09
Advertising 00001 00001 00035 00001 00021 00953 00009 00123 00943 00022 00097 00819 00019 00002 00000 00008 00002 00000 00620 00147 00000 00218 00191 00256
Promotion −00014 00064 00827 −00014 00037 00702 −00018 00027 00495 −00017 00027 00524 00026 00059 00655 −00037 00059 00530 00065 00042 00122 −00010 00045 00819
Price −00078 00100 00434 −00078 00100 00435 −00582 00160 00000 −00582 00160 00000 −00218 00209 00298 −00218 00209 00298 00185 00233 00427 00183 00233 00431
Sales conversion 00756 00286 00001 00758 00082 00000 00254 00215 00236 00251 00047 00000 00458 00213 00032 00457 00213 00032 00467 00028 00000 00456 00162 00005
Note. Bold indicates statistically significant effects at p < 0005.

Marketing Science 33(4), pp. 534–550, © 2014 INFORMS
competitor, shampoo brand SB. By contrast, the areas Resource Allocation Implications
where bottled water brand WA has more potential We explore two additional ways in which our results
than its competitor WB are less marketing actionable. are helpful for guiding marketing decision making.
For example, WA has more advertising awareness First, at any point in time, a manager can gauge
potential and less liking potential than WB. However, the overall attractiveness of investing in marketing
the latter matters much more in this low involvement actions that move the needle on different attitudinal
category. Thus, any marketing effort that stimulates metrics. As an illustration, consider the diagnostics
attitude metrics other than liking is likely to have only (at time = 84) for shampoo brand SA versus juice
negligible demand effects. brand JB (in Table 8). Brand SA can make progress
Two brands implemented major shifts in their mar- in its liking through advertising, which affects its
keting allocations after T = 84. Shampoo brand SA preference score; this, in turn, converts into higher
increased its advertising spending by 50% and kept sales. By contrast, brand JB needs to use promo-
its prices the same. In contrast, water brand WA tions to increase consideration that converts into sales.
cut its advertising spend by 42% while also keep- Numerous comparative scenarios can be derived from
ing prices the same. What are the consequences of Table 9, but these are limited to top-line inferences;
these brands’ strategic actions? We make directional i.e., there are no cost and profit implications of these
sales forecasts up to 12 months later, based on their scenarios.
attitude criteria shown in Table 8. As an illustra- Second, we conduct a more formal analysis of opti-
tive example, for shampoo brand SA, there is a high
mal marketing mix spending using dynamic pro-
responsiveness of attitudinal metrics to advertising.
gramming. To illustrate how to make marketing mix
Specifically, an increase in advertising moves the nee-
decisions by taking into account a mind-set metric,
dle on both the attitudinal metrics of awareness and
we pick two different shampoo brands, SC and SD,
liking. The awareness metric has an ample potential
as they have similar sales levels but varying levels of
of 73%, whereas the liking metric has a potential of
awareness; shampoo brand SC has higher awareness
29%. Finally, liking converts to sales resulting in a
than brand SD. We assume that both brands have the
forecasted increase in sales, which we denote with a
“↑” in Table 9. same 10% growth targets in terms of sales and aware-
Similar calculations through the chain of events ness over the last 12 periods. Our goal is to obtain the
from marketing actions → attitudinal metrics → optimal marketing mix path for both brands and val-
sales conversion are performed for each of the four idate our results using out-of-sample data from t = 85
brands in the analyses. In Table 9, we offer model- to t = 96.
based recommendations on changes in marketing We address the optimal marketing path by consid-
mix decisions for advertising, i.e., increasing (“↑”), ering two main steps (Gupta and Steenburgh 2008).
decreasing (“↓”), or maintaining (“−−”) with a view In step 1, we estimate sales response and attitude
to increasing brand sales. As Table 9 shows, brands response (awareness) models using time-series econo-
that followed a different course from the model- metrics. In step 2, we find the optimal marketing
based recommendations on marketing mix decisions path by using the estimated parameters in the opti-
(as depicted in the column “Agreement with model- mization part. Specifically, we follow a DP approach
based recommendation on spend”) performed worse used for solving multistage optimization problems
in terms of actual sales outcomes compared with (Rust 2006). The central idea of the DP is to maxi-
brands that followed a course consistent with model- mize (minimize) the sum of today’s reward (loss) and
based recommendations. Thus, diagnosing attitudi- the discounted expected reward (loss) from the future
nal brand metrics can help directionally predict the periods (Bellman 1957). The details of these two steps
impact of different marketing mix decisions on sales. are described in Web Appendix B.
Table 9 Illustrations of Model-Based Marketing Recommendations and Sales Outcomes
Sales outcome
Advertising spend Agreement with model-based Forecast conditional on
Brand Recommend Actual recommendations on spend agreement with recommendation Actual
SA ↑ ↑ Yes ↑ ↑
SB ↓ ↓ Yes ↑ −−
WA ↑ −− No ↑ ↓
WB ↑ −− No ↑ ↓
Note. ↑ denotes an increase, ↓ denotes a decrease, and −− denotes no change.

Figure 2 Optimal Price and Advertising Policies for Brands SC substantially increase projected sales revenues: 40%
and SD for brand SC and 34% for brand SD.
Optimal price policy
3
Brand SC
Brand SD
Managerial Implications and
2 Conclusions
Price
We argued in our introduction that the CFO’s needs

1
for financial accountability of marketing may well be
met by traditional marketing mix models on transac-
0
84 86 88 90 92 94 96 tions data. However, the chief marketing officer also
Time period needs to understand the consumer behavior reasons why
marketing does or does not affect business perfor-
Optimal advertising policy
mance. Our paper has demonstrated that the objec-
10
tives of both stakeholders can be met by recognizing
8
the unique properties of attitudinal metrics and their
Advertising
6
relationship to sales performance. In particular, these
4 measures have potential, stickiness, and responsive-
2 ness to marketing that can be assessed from the data.
0 Furthermore, the relevance of these metrics may be
84 86 88 90 92 94 96
assessed by their conversion into sales performance,
Time period
which provides the critical accountability link with
the CFO’s needs. By applying our approach, man-
Figure 2 displays the optimal marketing mix path agers can develop actionable guidelines on how to
over 12 periods to achieve the targeted sales (+10%) apply closed-loop learning on the attitude metrics
and targeted awareness (+10%). In addition, we com- (e.g., “if one observes metrics with the following val-
∗ ∗
pute (Xp1 t × St − Xa1 t ), where t is the time index; ues/characteristics, then this marketing action will be
∗ ∗
Xp and Xa are the optimal price and advertising most effective”).
policies, respectively; and S is the observed sales Different product categories and brands within
level. The cost of increasing revenue performance is them vary significantly in the magnitude of the four
through increased advertising or lowering price, or proposed criteria, and these differences form the basis
both. Note that despite the similar sales starting posi- for formulating marketing mix strategies that are
tion and target, our model recommends different mar- more likely to succeed. Table 10 provides an overview
keting policies for each brand. of four corner cases. The estimates reported in Table 8
Figure 3 shows the optimum sales revenues allow a classification of the brands into the four cells.
achieved upon implementing the optimal price and First, if a brand has low sales conversion from con-
advertising actions from t = 85 to t = 96. Both brands sumer attitudinal metrics, and low responsiveness to
marketing, we label that scenario a transactions effect at
Figure 3 Optimal Revenues for Brands SC and SD best. For the eight brands with three attitudinal met-
/PTIMALREVENUES rics in Table 8, only one can be classified into this cell.

"RAND3#²OPTIMAL
Thus, for most brands, marketing mix strategies result
"RAND3#²ACTUAL in sales conversion through the “mind-set effect”; i.e.,
2EVENUES
at least one attitudinal metric/marketing mix combina-

tion is sales relevant for all of these brand scenarios,
lending strong support to our current approach.

Turning to the second case, if a brand has low
conversion to sales from consumer attitudinal met-
4IMEPERIOD rics but high responsiveness to marketing, we label

"RAND3$²OPTIMAL
"RAND3$ACTUAL Table 10 Strategic Importance of Attitudinal Metrics
2EVENUES
Responsiveness Sales conversion

of attitude

to marketing Low High

Low Transactions effect at best Ineffective marketing lever
4IMEPERIOD High Ineffective marketing focus Long-term effect potential
that scenario an ineffective marketing focus. For exam- (e.g., weekly, monthly, quarterly) and including data
ple, brands that invest substantially in consideration on competitors. Comparisons between brands could
set-enhancing advertising may fail to see a substan- also be made in a more systematic way on the basis
tial sales lift. A case in point is water brand WA with of their strategic orientation—for instance, in terms
respect to advertising. Increases in advertising gen- of their degree of differentiation. Degree of differen-
erate awareness and consideration lifts that do not tiation and other factors such as market share may
convert to sales. Hence, for water brand WA, advertis- be important drivers of the cross-brand differences
ing represents an ineffective marketing focus that may in mind-set metric criteria. A company interested in
please managers focused on awareness and consider- brand equity may, in addition, want to examine to
ation metrics but not managers focused on increasing which extent its brand can command a revenue pre-
the top line. mium compared with other brands and private labels
Third, if the attitudinal metric has high sales con- (Ailawadi et al. 2003). The revenue premium can itself
version but does not respond well to increased mar- be modeled as a function of mind-set metrics. More-
keting spending, that would result in an ineffective over, data on the profits gained from better decisions
marketing lever scenario. This is the situation that would enable managers to weigh them against the
shampoo brand SB finds itself in with regard to cost of collecting attitudinal metrics, thus providing a
advertising. Consideration and liking have high sales return on investment measure for such data.
conversion for SB, but they do not respond well to If individual-level attitude metrics are available,
marketing spending. Managers can use such insights these could be used in more granular response model
to motivate a detailed analysis of the reasons, which specifications. Indeed, the lack of attitudinal metrics
may include the wrong message, the wrong execu- that match the transactional records is a limitation
tion, the wrong communication channel, the wrong of our approach. Attitudinal tracking data are typ-
timing, etc. In contrast, increases in shampoo brand ically survey based, which is costly and subject to
SA’s advertising generate liking that converts into sampling error. The digital age offers new opportu-
sales. Hence, advertising is an “effective marketing nities in this regard. Instead of surveying consumers,
lever” for shampoo brand SA to generate sales con- one can observe how they express themselves on the
version from a lift to liking. Internet via searches, chat rooms, social media, social
Finally, if the attitude metric has high sales conver- network sites, blogs, product reviews, and similar
sion, and there is high responsiveness to marketing, online word-of-mouth forums. Some preliminary evi-
we label that as a situation with long-term poten- dence suggests that “Internet-derived consumer opin-
tial. For example, cereal brand CA has sales conver- ions” are predictive of subsequent behavior (e.g., Shin
sion from awareness and consideration, which have et al. 2013). Future research should examine which
a high responsiveness to all marketing actions. This Internet-derived attitudinal metrics are the most rel-
offers an opportunity to allocate marketing resources evant and investigate the extent to which measure-
to move the needle on the consumer attitudinal met- ment error in online versus off-line metrics may mat-
ric of awareness and consideration and eventually to ter. These metrics could then be substituted for the
a long-term sales lift. survey-based measures that were used in this paper.
Our research opens up several avenues for future
work. One area is to examine alternative functional Supplemental Material
forms on the relationship between attitudes and sales, Supplemental material to this paper is available at http://dx
with an assessment of the relative performance of log- .doi.org/10.1287/mksc.2013.0841.
log models, log-linear models, as well as other forms
of nonlinearity. Moreover, the effectiveness of market- Acknowledgments
ing actions (e.g., promotions and advertising) could This paper was originally intended to appear in the special
be modeled as functions of awareness, consideration, issue on Theory and Practice in Marketing Conference Spe-
and liking. In addition, we model potential as the cial Section, but the review process was not concluded in
remaining distance to the maximum for its ease of time. The authors are listed in alphabetical order. The authors
elasticity definition, but future research should com- thank Kantar Worldpanel France for providing the data used
in this paper. They also thank participants at the Market-
pare alternative operationalizations (e.g., square root
ing Science Conference, the Marketing Dynamics Confer-
of the remaining distance to the maximum) as well ence, and the Theory+Practice Conference for their valuable
as alternative models and estimation techniques (e.g., suggestions.
Bayesian vector autoregressive models).
Future research should also explore category com-
parisons with even higher levels of consumer involve- References
ment, such as durables and high-value services, Ailawadi KL, Lehmann DR, Neslin SA (2003) Revenue premium as
possibly using observations at different time intervals an outcome measure of brand equity. J. Marketing 67(4):1–17.
Arellano M, Bond S (1991) Some tests of specification for panel Hardie BGS, Fader PS, Wisniewski M (1998) An empirical com-
data: Monte Carlo evidence and an application to employment parison of new product trial forecasting models. J. Forecasting
equations. Rev. Econom. Stud. 58(2):277–297. 17:209–229.
Baddeley A, Eysenck MW, Anderson MC (2009) Memory (Psychol- Jamieson LF, Bass FM (1989) Adjusting stated intention measures
ogy Press, London). to predict trial purchase of new products: A comparison of
Bagozzi RP, Silk AJ (1983) Recall, recognition, and the measurement models and methods. J. Marketing Res. 26(3):336–345.
of memory for print advertisements. Marketing Sci. 2(2):95–134. Johansson JK (1979) Advertising and the S-curve: A new approach.
Baltagi BH (2005) Econometric Analysis of Panel Data, 3rd ed. (John J. Marketing Res. 16(3):346–354.
Wiley & Sons, London). Kardes FR (1986) Effects of initial product judgments on subsequent
Batra R, Vanhonacker WR (1988) Falsifying laboratory results memory-based judgments. J. Consumer Res. 13(1):1–11.
through fields tests: A time-series methodology and some Keller KL, Lehmann DR (2006) Brands and branding: Research find-
results. J. Bus. Res. 16(4):281–300. ings and future priorities. Marketing Sci. 25(6):740–759.
Bauer RA (1967) Consumer behavior as risk taking. Cox DF, ed. Risk Kotler P, Keller KL (2012) Marketing Management, 14th ed. (Pearson
Taking and Information Handling in Consumer Behavior (Harvard Prentice-Hall, Upper Saddle River, NJ).
University Press, Boston), 23–33. Lautman M, Pauwels K (2009) What is important? Identifying met-
Baumgartner MH, Leippe MR, Ronis DL, Greenwald AG (1983) rics that matter. J. Advertising Res. 49(3):339–359.
In search of reliable persuasion effects: II. Associative interfer- Lingle JH, Ostrom TM (1979) Retrieval selectivity in memory-based
ence and persistence of persuasion in a message-dense envi- impression judgments. J. Personality Soc. Psych. 37(2):180–194.
ronment. J. Personality Soc. Psych. 45(3):524–537. Nelson P (1970) Information and consumer behavior. J. Political
Belch GE (1982) The effects of television commercial repetition on Econom. 78(2):311–329.
cognitive response and message acceptance. J. Consumer Res. Norris IB, Peters K, Naik PA (2012) Discovering how advertising
9(1):56–65. grows sales and builds brands. J. Marketing Res. 49(6):793–806.
Bellman R (1957) Dynamic Programming (Princeton University Press, Pauwels K, Erguncu S, Yildirim G (2013) Winning hearts, minds
Princeton, NJ). and sales: How marketing communication enters the purchase
Bettman JR (1979) Memory factors in consumer choice: A review. process in emerging and mature markets. Internat. J. Res. Mar-
J. Marketing 43(2):37–53. keting 30(1):57–68.
Burke RR, Srull TK (1988) Competitive interference and consumer Peter JP, Tarpey LX Sr (1975) A comparative analysis of three con-
memory for advertising. J. Consumer Res. 15(1):55–68. sumer decision strategies. J. Consumer Res. 2(1):29–37.
Deighton J, Henderson CM, Neslin SA (1994) The effects of adver- Rabe-Hesketh S, Skrondal A (2005) Multilevel and Longitudinal Mod-
tising on brand switching and repeat purchasing. J. Marketing eling Using Stata (Stata Press, College Station, TX).
Res. 31(1):28–43. Rust J (2006) Dynamic programming. Entry for consideration by the
New Palgrave Dictionary of Economics. Accessed September 1,
Dekimpe M, Hanssens DM (1999) Sustained spending and persis-
2013, http://gemini.econ.umd.edu/jrust/research/papers/dp
tent response: A new look at long-term marketing profitability.
.pdf.
J. Marketing Res. 36(4):397–412.
Schank RC (1982) Dynamic Memory: A Theory of Reminding in Com-
Ehrenberg ASC (1974) Repetitive advertising and the consumer.
puters and People (Cambridge University Press, New York).
J. Advertising Res. 14(2):25–34.
Shin HS, Hanssens DM, Kim KI, Choe JA (2013) Positive vs. neg-
Ehrenberg ASC, Goodhardt GJ, Barwise TP (1990) Double jeopardy
ative e-sentiment and the market performance of high-tech
revisited. J. Marketing 54(3):82–91.
products. MSI Working Paper 13-123, University of California,
Feldman JM, Lynch JG Jr (1988) Self-generated validity and other Los Angeles, Los Angeles.
effects of measurement on belief, attitude, intention and behav- Simmons CJ, Bickart BA, Lynch JG Jr (1993) Capturing and creat-
ior. J. Appl. Psych. 73(3):421–435. ing public opinion in survey research. J. Consumer Res. 20(2):
Fischer M, Giehl W, Freundt T (2011a) Managing global brand 316–329.
investments at DHL. Interfaces 41(1):35–50. Sims CA (1986) Are forecasting models usable for policy analysis?
Fischer M, Albers S, Wagner N, Frie M (2011b) Dynamic marketing Federal Reserve Bank of Minneapolis Quart. Rev. 10(1):2–6.
budget allocation across countries, products, and marketing Sobel ME (1982) Asymptotic confidence intervals for indirect effects
activities. Marketing Sci. 30(4):568–585. in structural equation models. Sociol. Methodol. 13:290–312.
Fourt LA, Woodlock JW (1960) Early prediction of market success Srinivasan S, Vanhuele M, Pauwels K (2010) Mindset metrics in
for new grocery products. J. Marketing 25(2):31–38. market response models: An integrative approach. J. Marketing
Franses PH, Vriens M (2004) Advertising effects on awareness, con- Res. 67(4):672–684.
sideration and brand choice using tracking data. ERIM Report Stahl F, Heitmann M, Lehmann DR, Neslin SA (2012) The impact
Series ERS-2004-028-MKT, Erasmus University Rotterdam, of brand equity on customer acquisition, retention, and profit
Rotterdam, The Netherlands. margin. J. Marketing 76(4):44–63.
Gupta S, Steenburgh TJ (2008) Allocating marketing resources. Steenkamp J-B, Nijs VR, Hanssens DM, Dekimpe MG (2005) Com-
HBS Working Paper 08-069, Harvard Business School, Boston. petitive reactions and the cross-sales effects of advertising and
http://www.hbs.edu/faculty/Pages/item.aspx?num=32002. promotion. Marketing Sci. 24(1):35–54.
Haley R, Case PB (1979) Testing thirteen attitude scales for agree- Tulving E, Pearlstone Z (1966) Availability versus accessibility of
ment and brand discrimination. J. Marketing 43(4):20–32. information in memory for words. J. Verbal Learn. Verbal Behav.
Hanssens DM, Parsons LJ, Schultz RL (2001) Market Response Mod- 5(4):381–391.
els: Econometric and Time Series Analysis, 2nd ed. (Kluwer Aca- Vakratsas D, Ambler T (1999) How advertising works: What do we
demic Publishers, Norwell, MA). really know? J. Marketing 63(1):26–43.

Consumer Attitude Metrics For Guiding Marketing Mix Decisions

Uploaded by

Copyright:

Available Formats

Consumer Attitude Metrics For Guiding Marketing Mix Decisions

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Consumer Attitude Metrics For Guiding Marketing Mix Decisions

Uploaded by

Copyright:

Available Formats

This article was downloaded by: [128.187.103.

98] On: 10 January 2015, At: 01:53

Consumer Attitude Metrics for Guiding Marketing Mix

To cite this article:

Full terms and conditions of use: http://pubsonline.informs.org/page/terms-and-conditions

Copyright © 2014, INFORMS

Please scroll down for article—it is on subsequent pages

Consumer Attitude Metrics for Guiding

Introduction on financial returns (including the chief financial offi-

brand consideration stands at 70% while brand lik- http://dx.doi.org/10.1287/mksc.2013.0841). First, it

mind-set route and the latter as the transaction route

Figure 1 Conceptual Framework

Responsiveness refers to marketing’s ability to “move For example, in an awareness-to-advertising relation-

After a discussion with the data provider, we Econometric Modeling

Table 1 Overview of Metrics and Models

Metrics/models Equation Model Table

Potential metric Equation (1) POTt = 6MAX − At−1 7/MAX Table 8

Brand variation 44007 76076 64009 71071

Shampoo Bottled water Juice Cereal

Model 1 (DV = Awareness) Model 2 (DV = Consideration) Model 3 (DV = Liking)

Coefficient SE p > z Coefficient SE p > z Coefficient SE p > z

Shampoo Bottled water Juice Cereal

Shampoo Bottled water Juice Cereal

Table 8 Diagnostics for the Two Top Brands in Each Category, T = 84

Stickiness = 0.35 0.33 0.48 0.48 0.56 0.57 0.30 0.30

Stickiness = 0.16 0.16 0.24 0.24 0.27 0.27 0.03 0.03

Stickiness = 0.14 0.14 0.25 0.25 0.36 0.36 0.09 0.09

Note. Bold indicates statistically significant effects at p < 0005.

Table 9 Illustrations of Model-Based Marketing Recommendations and Sales Outcomes

Note. ↑ denotes an increase, ↓ denotes a decrease, and −− denotes no change.

We argued in our introduction that the CFO’s needs

at least one attitudinal metric/marketing mix combina-

 Responsiveness Sales conversion

You might also like

Coefficient SE p > z Coefficient SE p > z Coefficient SE p > z

Responsiveness Sales conversion