Sei sulla pagina 1di 9

Journal of Business Research 69 (2016) 3168–3176

Contents lists available at ScienceDirect

Journal of Business Research

Research in reverse: Ad testing using an inductive consumer


neuroscience approach
Terry Daugherty ⁎, Ernest Hoffman, Kathleen Kennedy
Suarez Applied Marketing Research Laboratories, The University of Akron, United States

a r t i c l e i n f o a b s t r a c t

Article history: Consumer neuroscience (CNS) methodologies were adapted from scientific disciplines where research is done in
Received 1 March 2015 reverse from the scientific method that has come to inform the vast majority of marketing research. Conventional
Received in revised form 1 June 2015 applications of CNS often begin with known outcomes and implications (i.e., symptomologies), subject them to
Accepted 1 October 2015
neuro-physiological tests, and use the data from these tests to generate descriptive (e.g., diagnosis) and prescrip-
Available online 29 December 2015
tive (e.g., prognosis) hypotheses. This article offers the heretical proposition that such an approach should not be
Keywords:
limited to the “hard sciences,” but could be profoundly beneficial in fields that have come to exclusively rely upon
Consumer neuroscience the scientific method. Findings of a dEEG study conducted on the basis of known results in order to generate
Neuro-marketing novel theory and hypotheses are presented. Implications of this paradigm shift for advancing marketing research
Marketing research while helping to bridge an ever-widening scholar–practitioner divide are discussed.
Ad testing © 2015 Elsevier Inc. All rights reserved.

1. Introduction By and large the full potential of CNS to meaningfully contribute to-
wards understanding consumer behavior in particular, and marketing re-
The utilization of neurological and physiological measurement tools search in general, has yet to be unlocked. Doing so would require
to “analyze and understand human behavior in relation to markets and entertaining a heretical proposition that contradicts one of the deepest-
marketing exchanges” (Lee, Broderick, & Chamberlain, 2007, p. 200), or held assumptions of consumer behavior research within the social
“consumer neuroscience” (CNS) is a rapidly growing trend. The sciences; that the scientific method is the only suitable approach for
emergence of this once obscure area of consumer research is often at- engaging in scientific discovery. This seeming heretical contention is
tributed to a) a growing appreciation for scientific and objective mea- rooted in observing the way in which tools such as EEG and fMRI are
surement, which is preferable to more subjective research paradigms used in more naturalistic settings, such as the medical field. Rather than
(e.g., surveys, focus groups), and b) claims that such techniques consti- starting with theory-driven hypotheses that utilize physiological and
tute more accessible and cost-effective means of developing products neurological tools to yield results and implications (i.e., scientific meth-
consumers will embrace (Ariely & Berns, 2010; Pradeep, 2010). While od), the hard sciences often begin an investigation in reverse. For exam-
the potential of CNS methodologies has been met with increasing ple, a patient will present a known result (e.g., slurred speech) along with
recognition among scholars (e.g., Advertising Research Foundation implications (e.g., unable to perform one's job) and is then subjected to
NeuroStandards Project, 2013) and industry leaders (e.g., Nielsen, medical tests (i.e., EEG and fMRI) in order to locate the source or cause.
Buyology, Gallup, & Robinson, and Innerscope Research), many ques- Only then are theories and hypotheses applied to generate diagnoses,
tion the legitimacy of some claims. Consumer neuroscience has been prognoses, and prescriptive actions. Given the success of this approach
accused of failing to deliver on the promises that sparked initial popu- in the medical field, one might wonder why such a methodology has
larity, such as having the ability to identify a consumer “buy button” yet to be translated into CNS research and practice. By using known re-
in the brain (Reason, 2012) and providing a cost-effective alternative sults (e.g., advertising data) and implications (e.g., dollars invested, ads
to conventional market research techniques (Ariely & Berns, 2010). continued/discontinued, market share), consumer responses could be
Others point to a lack of transparency by companies' marketing CNS subjected to physiological and neurological tools in order to generate
(Advertising Research Foundation, 2013; Parson, 2011), as well a “pau- novel theory and hypotheses regarding the success (or failure) of mar-
city of peer-reviewed reports in the field” as indicators that “companies keting strategy. The absence of such a paradigm is likely attributable to
may be making premature claims about the power of neuroscience to consumer behavior researchers being strongly conditioned (by the scien-
predict consumer behavior” (Fisher, Chin, & Klitzman, 2010, p. 230). tific method) to start with hypotheses developed from established theo-
ry, design a study to adequately test those hypotheses, analyze the
⁎ Corresponding author at: Department of Marketing, College of Business
Administration, The University of Akron, Akron, OH 44325, United States. Tel.: +1 330
results, and discuss any pertinent implications, limitations, and future re-
972 8304. search needs (see Hudson & Ozanne, 1988, for an in-depth discussion of
E-mail address: terry.daugherty@uakron.edu (T. Daugherty). assumptions that guide consumer research). Such a deductive approach

http://dx.doi.org/10.1016/j.jbusres.2015.12.005
0148-2963/© 2015 Elsevier Inc. All rights reserved.
T. Daugherty et al. / Journal of Business Research 69 (2016) 3168–3176 3169

is so deeply engrained that alternative perspectives are often overlooked. philosophies continue to exist on the fringe of social science generally
Nevertheless, the potential for an inductive approach in CNS to yield categorized as grounded theory (Allan, 2003). While qualitative para-
novel insights is too substantial to disregard. digms have seemingly embraced induction, this research approach is
In order to demonstrate the potential of an inductive methodology, not defined by data categorization and can be quantitative-based.
this study begins by reviewing two widely accepted beliefs and their he- Initial enthusiasm towards utilizing neurological measurement tools
retical alternatives. In particular, a) the notion that the ends fail to justify for understanding consumer behavior has given rise to considerable
the means with a CNS paradigm and b) the idea that theory and re- amounts of research in academic and practitioner realms (Advertising
search must inform practice rather than practice informing theory and Research Foundation, 2013; Ariely & Berns, 2010; Hubert, 2010;
research. Next, a reverse research paradigm approach is used originat- Plassmann, Ramsoy, & Milosavljevic, 2012). While a complete review
ing from known results leading to theoretical scrutiny and conjecture. of all measurement techniques used is beyond the scope of this article
By starting with a known result and implications, the study then (see Kenning & Linzmajer, 2011; Wang & Minor, 2008, for a systematic
works towards intriguing hypotheses and theory using EEG as a diag- review), numerous stakeholders have begun to scientifically explore the
nostic tool. Finally, the study builds from this experience to suggest the- critical question; do the ends (i.e., efficient prediction of consumer be-
oretical and practical implications of inductive research, both within havior) justify the means (i.e., more expensive and complicated meth-
and beyond a CNS paradigm. This approach contributes to the literature odologies)? Answers to this question are mixed at best. Some appear
in at least three notable ways. The review and study suggest an intrigu- to be thoroughly convinced that CNS is well worth continued invest-
ing methodological paradigm that contradicts most empirical research ment (e.g., Fugate, 2007; Lee et al., 2007; Plassmann et al., 2012),
in consumer behavior, particularly as appearing in scholarly journals. while others have seemingly reserved judgment pending further out-
Second, this approach provides a novel and practical means of bridging comes (e.g., Advertising Research Foundation, 2013; Ariely & Berns,
the scholar–practitioner divide inasmuch as practitioners have a lot of 2010; Hubert, 2010). The ethical utility of studying consumer responses
known results and implications at their disposal, while scholars have which are harder to control and manage (i.e., consent to disclosing)
the means of making sense of those results using research tools and the- than those elicited with more traditional self-report surveys and focus
oretical acumen. Third, by breathing new life into CNS as a means of groups is also a concern (Fisher et al., 2010; Murphy, Illes, & Reiner,
gaining substantial insight over conventional understanding, this re- 2008; Wilson, Gaines, & Hill, 2008).
search signifies a response to critics who question the ability of CNS to A tendency to frame CNS methodologies as a means of better
survive and thrive as a viable discipline. predicting future consumer behavior (i.e., deductive approach) is a no-
The article closes with theoretical and practical implications for table feature of the research provided in support of these convictions.
conducting inductive research both within and beyond CNS and sug- Thus, the purpose is to use neurophysiological responses to test product
gests worthwhile directions for future research. Additional work will and package design (Reimann, Zaichkowsky, Neuhaus, Bender, &
further demonstrate the value of this suggested approach, while Weber, 2010), advertising (Ambler, Ioannides, & Rose, 2000; Kato
encouraging scholar–practitioners to pursue alternate and perhaps et al., 2009; Vecchiato et al., 2011), and brand-related information
even heretical routes to scientific discovery. (Schaefer & Rotte, 2007; Stewart, Pickering, & Sturt, 2004;
Treleaven-Hassard et al., 2010) as a basis for effectively discriminating
1.1. Heresy 1: CNS methodologies were designed to understand—not between strategies. For example, a typical application of CNS might
predict—consumer behavior consist of having consumers view multiple advertisements for a product
in the ad-testing phase while brain wave activity is monitored using
Since the inception of scientific discovery, knowledge generation can EEG. The presence or absence of specific waves (e.g., P3a, Treleaven-
be characterized as developing along two divergent paths: one Hassard et al., 2010) as ad alternatives are viewed is used to make pre-
deducing empirical consequences stemming from hypothesis-driven dictions regarding whether or not the advertisement will successfully
prediction, and the second utilizing inductive observation leading to attract product purchasing. In other words, EEG technology functions
generalizable principles (Bonoma, 1985). In essence, deductive-based as a means of predicting consumer ad response.
inquiry starts at the general and moves towards formulating predictions For all of CNS's promise, prediction-oriented applications completely
to the particular, whereas induction begins with the observable and neglect two critical insights. First, anticipated consumer behavior (the
progresses to the universal (see Fig. 1). The dominance of the deductive “ends” of current CNS research) is no substitute for actual consumer be-
approach can be traced to the philosophical axiom that induction proves havior. The former is subject to educated guesswork and projection at
nothing, and as a result, generations of scientists have been taught to best, whereas the latter already exists in the databases of organizations,
think deductively (Lawrence, 1982). Nevertheless, inductive-based marketing firms, and academic institutions. Massive amounts of data
pertaining to known successes and failures are readily accessible for fur-
ther diagnosis using neurological and physiological modes of assess-
ment popular to CNS, yet no CNS investigations have attempted to
leverage these real-world results. Secondly, CNS methodologies were
Inductive Research not originally intended to facilitate prediction. Meaning, the medical
equivalence of CNS techniques as currently used might consist of a doc-
tor using magnetic resonance imagery (MRI) to predict a future bone
fracture or a neurologist using electroencephalography (EEG) to predict
future epilepsy. Neither of these methodologies becomes useful until
the precipitating phenomenon (bone fracture, epilepsy) has occurred,
and recognizing this has clear implications for the manner in which
they are used within a CNS paradigm. Acknowledging these vital
limitationsconsiders the first heretical claim; CNS researchers and prac-
titioners have long been attempting to fit a square peg into a round hole.
In an attempt to fashion CNS methodologies to fit into the rubric of the
Deductive Research deductive scientific method, the original function and purpose of these
methodologies have been overlooked. Techniques such as fMRI and
EEG are used in other contexts (e.g., medicine, neuroscience) to illumi-
Fig. 1. A “research in reverse” approach. nate existing phenomena and not to predict or anticipate as-of-yet
3170 T. Daugherty et al. / Journal of Business Research 69 (2016) 3168–3176

nonexistent phenomena. By the time a patient presents themselves for important consideration to note with this particular case is that the re-
one of these tests, known results have already occurred. Understanding search team was intentionally kept unaware of each ad's classification
those results for the purpose of refining theories and offering prescrip- until completion of the data collection. While the firm sought to deter-
tive advice is the primary purpose of subjecting patients to these tests. mine whether neurological differences exist across the designated ad
Adapting this insight to the sub-discipline of CNS would require an in- categories, the research team recognized this as an opportunity to
ductive approach, essentially conducting “research in reverse” from explore the specific nature and meaning of any differences identified
the conventional method that all mainstream academic journals adhere. in view of neurological theory, resulting in the following research
Rather than starting with established theory and hypotheses, consumer question:
research would begin with known results and implications. From this
point, neurophysiological methodologies could be used as a means of RQ: Will measurable neural activation differences exist among con-
better understanding these results on the way to developing more re- sumers viewing successful versus unsuccessful direct response
fined and nuanced theories and hypotheses regarding various elements product advertising?
of consumer behavior. The benefits of doing “research in reverse” how-
ever are poised to extend far beyond consumer neuroscience. Starting 2. Method
with real-world data, and applying methodologies in order to generate
novel theory and hypotheses, has broad potential that extends into A within-subjects design was used to measure consumer neural
marketing and business research as a whole. The following section activity using dense-array electroencephalography (dEEG) with direct
briefly explores this second heretical claim in depth. response advertisements serving as the independent variables.
Within-subjects research is common in CNS because activation patterns
1.2. Heresy 2: reversing the scientific method could lead to valuable insights are highly reliable within participants but not necessarily between
participants (see Corsi-Cabrera, Galindo-Vilchis, Del-Rio Portilla, Arce,
An overwhelming majority of marketing research relies upon the & Ramos-Loyo, 2007), suggesting that individuals provide their own
scientific method, which remains surprisingly uncontested by scholars optimal baseline for measuring activation differences.
(see Anderson, 1983 for one exception). Some have questioned whether
scholars adopting this approach achieve their ultimate aim of engaging 2.1. Participants
in “scientific” research (e.g., Hodson, 1996). For instance, pre-existing
theory can act as a biasing factor that hinders scientific inquiry, which Twenty-three subjects (n = 23) participated in the study. A purpo-
ought to be objective and observation-based. Meanwhile, inductive sive sample was used with participants screened to match the general
approaches (Haig, 1995) begin with participant observations and use target audience for direct response advertising, as defined by the direct
them as a basis for establishing novel theory and propositions. Similar marketing firm. Specifically, participants were required to; (1) demon-
insights could be gained when business researchers begin with quanti- strate a propensity to respond to direct response television advertising,
tative data (e.g., sales figures) and use existing methodologies as a as measured by a recent direct response purchase by phone or online
means of translating observed phenomena into coherent and distinct after viewing an ad, (2) be 30 years in age or older, and (3) report a
theoretical frameworks. Although the research example presented ap- household income of at least $30,000 a year. Subjects were recruited
plies this proposition from a CNS perspective, this application extends from the local area and monetarily compensated ($100) for their
beyond the realm of business where known results (e.g., employee participation.
turnover, financial performance, work-related injuries) are readily
available. 2.2. Stimuli
By proposing an inductive “research in reverse” approach, the inten-
tion is not to discredit or discourage research that proceeds according to A total of six direct response television ads were selected by the di-
the deductive based scientific method. To the contrary, a vast amount of rect marketing firm to serve as the stimuli for the study. The ads varied
insight has been gained via deductive research with progress made as a across a number of product categories, including housewares (Ad1: Liq-
result of this approach. Nevertheless, the view taken here is that addi- uid Dispenser & Ad3: Food Storage), home repair (Ad2: Spray Sealant),
tional progress is possible, particularly within the domain of CNS, by health and beauty (Ad4: Skin Care Cream & Ad6: Self-Tanning Lotion)
attempting to execute this prevailing research paradigm in reverse. and collectables (Ad5: Commemorative Coin), yet maintained a similar
The following case study is presented to illustrate the potential benefits price point ($19.99). In addition, each ad was 60 s in length and follow-
of inductive research using CNS measurement methods as a way to bet- ed an analogous informational design by presenting a consumer prob-
ter understand known responses to a series of product advertisements. lem (e.g. dry skin), introducing the product, communicating a solution
via product benefits and ultimately requesting consumers to order the
1.3. Case study: using CNS to distinguish between “successful” and product immediately by phone or online. An important point of empha-
“unsuccessful” product advertising sis here is to reinforce that the purpose of this case study is not to decon-
struct the executional elements of the ads as the units of analysis.
The following study emerged as part of an ongoing research rela- Meaning, while many similarities occur among the ads, each is obvious-
tionship with a large Midwestern direct marketing firm. The firm ly different and the goal is not to conduct A/B tests or delineate structur-
approached the research team questioning the value of CNS and pro- al differences between the ads. Rather, the purpose is to determine
posed a challenge to determine whether or not CNS techniques would whether or not neural activity differs among individuals when exposed
be able to differentiate between successful and unsuccessful direct re- to all of the ads, and if so, is this activity meaningfully different accord-
sponse product advertisements. Six direct response style ads (three suc- ing neurological theory.
cessful, three unsuccessful) were selected by the firm with successful
ads operationalized as those generating N300 units sold during the 2.3. Dependent measure
first week of the media schedule, and unsuccessful ads resulting in
b300 units sold. This is an internal benchmark used by the firm for allo- The dependent variable in this study is brain wave activity measured
cating media spending on a specific campaign. A defining feature of di- by means of dEEG using a 256-channel Hydrocel Geodesic Sensor Net
rect response advertising is the goal of eliciting a direct behavioral (Electrical Geodesics Inc., Eugene, Oregon), which is soaked in a potas-
response translating into immediate sales. If unit sales do not occur at sium chloride saline solution and then applied to each participant's
a profitable level within the first week the ad is discontinued. An scalp. Brain wave activity is operationalized as visually-evoked event-
T. Daugherty et al. / Journal of Business Research 69 (2016) 3168–3176 3171

related potentials (ERPs), or fluctuations in electrical neural activity ERPs over the course of multiple exposures has been shown to provide
(i.e., voltage) that occur as subjects are exposed to visual stimuli. an effective means of distinguishing genuine activation differences from
Event-related potentials are measured in milliseconds, given the rapid noise artifacts and ongoing brain activity (e.g., Arieli, Sterkin, Grinvald, &
response of the brain to external stimuli, and then combined across Aertson, 1996). The image presentations were separated by a) a fixation
channels in theoretically-relevant ways (e.g., to isolate specific geo- screen (i.e., black screen with a white cross in the center) shown for a
graphical areas of the brain). Data can thus be analyzed in terms of sig- random interval between 800–1400 ms and b) a blank black screen
nificant time epochs (e.g., 200–350 ms) as well as segmented wave that was displayed for 200 ms. Random fixation intervals are frequently
functions (e.g., maximum amplitude). Differences in brain wave activity used in this type of research to mitigate practice effects. All stimuli were
are assessed on the basis of ERPs collected as participants view and re- of similar production value, screen resolution, and counterbalanced in
spond to multiple stimuli exposures. order of presentation in both the learning phase and ERP phase using
E-Prime (Psychology Software Tools, Inc.). Overall, the active data col-
2.4. Procedure lection took approximately 55 min per subject with Fig. 2 illustrating
this procedure.
Upon entering the lab participants were informed of the procedures
with consent obtained for the study. Preparation was then taken to fit
participants with a sensor net followed by single-trial dEEG node 3. Results
response tests to ensure that each of the channel sensors was properly
recording data. Subjects were then exposed to the stimuli via an exper- 3.1. Data analysis
imental station using a two-phase process common to ad testing
(Young & Robinson, 1987): a learning phase and ERP phase. First, partic- The sample consisted of 9 males (36%) and 14 females (64%) with an
ipants viewed the ads in their entirety separated by the momentary pre- average age of 55. This sample size is comparable to other studies
sentation of a black empty screen. Following the learning phase a brief (e.g., Fallani et al., 2008; Rothschild & Hyun, 1990; Vecchiato et al.,
filler task (demographic survey) was used to separate and transition 2010) utilizing EEG to explore within-subject activation differences as
to the primary ERP portion of the study. subjects respond to television commercials. NetStation 4.2 software
Second, during the ERP phase subjects were exposed to a series of (Electrical Geodesics Inc., Eugene, Oregon) was used to process the
distinct still images from each of the ads (4 per ad, 24 total) and an ad- raw ERP data. EEG recordings were filtered using 0.3–45 Hz band-pass
ditional set of images selected from the International Affective Pictures filters. Event-related potential trials were base-lined at 100 ms and off-
System database (IAPS: Lang, Bradley, & Cuthbert, 2008). The IAPS im- set to 15 ms based on a visual timing test of the data collection system
ages are extensively validated in terms of their perceived valence used. Bad channels were marked, and artifacts were detected using
(pleasantness) and intensity. By selecting and presenting subjects a se- the NetStation auto detection tool and subsequent inspection. ERP seg-
ries of ten pleasant, unpleasant, and neutral images (30 total) where in- ments were rejected if they contained more than 10 bad channels or any
tensity is held constant (i.e., moderate), subjects are provided with a artifacts. Data were epoched at 200–350 ms, 350–500 ms, and 500–
standard of comparison as they make judgments regarding the focal 800 ms. Scalp channel data were converted into source space by
stimuli. Subjects were asked to respond to each still image they viewed means of EGI GeoSource Source-Imaging Specification (Electrical
by using a keypad to rate the image as a) pleasant, b) unpleasant, or Geodesics Inc., Eugene, Oregon) using the Finite Difference Model
c) neutral while ERP data was recorded. Images of the stimuli were pre- constrained by standardized Low-Resolution Brain Electromagnetic To-
sented 10 times for 3 s each and in 5 blocks, separated by user- mography (sLORETA) with a Tikhonov regularization of 10-3. The re-
controlled breaks in order to reduce participant fatigue. Averaging sults of this analysis are illustrated below with data derived using

Preparation Learning Phase Transition ERP Phase

Attach Net and Subject Views Brief Filler Brain Activity


Test Sensors All Ads Task Recorded

Demogr Image
Survey Block
Ad1 Ad2

User-
Controlled
Ad3 Ad4
Break

Ad5 Ad6

15 min 15 min 5 min 20 min

Fig. 2. Experimental procedure and timeline.


3172 T. Daugherty et al. / Journal of Business Research 69 (2016) 3168–3176

source montages for the voxels corresponding with known regions of observed neurological difference is also of primary interest. Upon com-
the brain (i.e., Brodmann areas). pletion of the data collection, the direct marketing firm identified Ad2
(Spray Sealant), Ad3 (Food Storage) and Ad4 (Skin Care Cream) as suc-
3.2. Research question cessful in market direct response ads with Ad1 (Liquid Dispenser), Ad5
(Commemorative Coin) and Ad6 (Self-Tanning Lotion) as unsuccessful.
The research question reflects a central goal of determining whether Differences in brain wave activity were first assessed for the data as a
differences in neural activation will be found among the direct response whole (i.e., across all nodes of the EEG net) with the three time epochs
ads that were known to be successful in the market, as opposed to those (200–350 ms, 350–500 ms, and 500–800 ms) tested because of their
that were unsuccessful. As a means of moving from known results common use in research pertaining to episodic memory (e.g., Bland &
to novel theory generation, the specific nature and meaning of any Schaefer, 2011; Hsu, Garside, Massey, & McAllister-Williams, 2003).

Fig. 3. Direct response television advertising EEG GeoSource estimates.


T. Daugherty et al. / Journal of Business Research 69 (2016) 3168–3176 3173

Within each time epoch, six metrics of the sinusoidal ERP wave were Collectively, these five areas are associated with episodic memory
analyzed. Maximum amplitude, minimum amplitude, mean amplitude, retrieval (Lepage, Ghaffar, Nyberg, & Tulving, 2000), a process that
and adaptive mean were measured in Megavolts, which is a unit of elec- subjects in this study engage in as they link still images of stimuli to
trical brain activity. Two additional variables, latency to positive peak the advertisements they came from. Each of these areas has bilateral
and latency to negative peak, were measured in milliseconds. symmetry, but left versus right hemispheric activation differences are
GeoSource localization of neural activity for each of the ads across fairly common, so the left and right hemispheres were analyzed sepa-
epochs is constructed to visually illustrate how activation changes rately for each BA, resulting in 10 analysis areas.
over time (see Fig. 3). Subsequently, significant neural activation differ- As Table 1 illustrates, significant group differences exist in five dis-
ences were detected across the six direct response ads at the 200– tinct Brodmann areas. Note that six significant differences are reported,
350 ms epoch (maximum amplitude; F1, 23 = 2.89, p b .05), but not but two of these differences are localized to the left and right hemi-
for 350–500 ms epoch (maximum amplitude; F1, 23 = 2.14, p N .05), spheres for BA 8. Also, each of these BAs is geospatially located in the
or 500–800 ms epoch (maximum amplitude; F1, 23 = 2.64, p N .05). prefrontal region of the brain (see Fig. 4). Brodmann area 8 has been as-
The lack of significant differences across epochs is not unexpected sociated with the processing of uncertainty (e.g., Volz, Schubotz, & von
when examining overall neural activity as reactions to stimuli begins Cramon, 2005). Analysis results suggest statistically significant ERP
to normalize as the brain adjusts (Hsu et. al., 2003). While both observ- differences between successful and unsuccessful ads in BA8 in the
able and statistically significant neural activity differences have been right hemisphere in the 200–350 ms epoch (for maximum amplitude;
detected, these differences are not necessarily meaningful in the sense F1, 23 = 7.41, p b .01) as well as in the left hemisphere in the 350–
that they do not allow us to distinguish between successful versus 500 ms epoch (for minimum amplitude; F1, 23 = 4.66, p b .05). A
unsuccessful ads. The reason is because overall neural activity does comparison of means in both cases suggests that successful ads (m8R
not provide differentiation along key neural correlates associated with success = 2.05, m8L success = 8.44) prompt less uncertainty than
specialized cognitive functions. unsuccessful ads (m8R unsuccessful = 2.29, m8L unsuccessful =
In order to meaningfully interpret activation differences, localization 9.69). Activation differences in the right hemisphere for BA9 in the
of the data to distinct areas of the brain that are known to engage in spe- 200–350 ms epoch for maximum amplitude (F1, 23 = 8.48, p b .01)
cific types of neurological processes is necessary. During this analysis, might suggest differences in working memory activation (see Zhang,
Ad2, Ad3, and Ad4 are designated as “successful” ads while Ad1, Ad5, Leung, & Johnson, 2003), such that successful direct response advertis-
and Ad6 as “unsuccessful”. The Brodmann (1909) classification is wide- ing (m9R success = 1.86) require less effort to retrieve than their unsuc-
ly used in neuroscience to distinguish specific regions of the brain in cessful counterparts (m9R unsuccessful = 2.02). The same could be said
terms of structure and functional purpose. Scholars generally agree for observed differences in maximum amplitude for the left hemisphere
that there are 52 Brodmann areas (BAs), which are distinct from each of BA10 (F1, 23 = 4.68, p b .05) during the same time interval. Like BA9,
other in terms of neuronal configuration (see Fig. 4). For the present scholars believe that BA10 contributes to working memory by assisting
study, EEG data was localized to five of these areas and tested for differ- in the process of cognitive branching, whereby one working task is put
ences on the basis of ad success; in particular, BAs 8, 9, 10, 45, and 47. on hold until another is complete (Koechlin & Hyafil, 2007). Observed

Left Hemisphere

Right Hemisphere

Fig. 4. Brodmann areas.


3174 T. Daugherty et al. / Journal of Business Research 69 (2016) 3168–3176

Table 1
ERP wave differences by Brodmann's area for direct response television advertising.

Brodmann area Successful adsc Unsuccessful adsc Type III sum of squares df Mean square F Sig.

8La 8.44 9.69 0.06 1 0.06 5.92 p b .05


8Rb 2.05 2.29 0.36 1 0.36 7.41 p b .01
9Rb 1.86 2.02 0.17 1 0.17 8.48 p b .01
10Lb 2.49 2.63 0.12 1 0.12 4.68 p b .05
45Lb 5.72 6.14 1.16 1 1.16 6.60 p b .05
47La 1.02 1.15 0.12 1 0.12 4.66 p b .05

Note. L = left brain hemisphere; R = right brain hemisphere.


a
Brodmann area with significant differences in minimum amplitude between 350 and 500 ms.
b
Brodmann area with significant differences in maximum amplitude between 200 and 350 ms.
c
Recorded ERP mean scores by ad category per Brodmann area.

mean differences between successful (m10L success = 2.49) and unsuc- and empirical example has been provided that appears to corroborate
cessful (m10L unsuccessful = 2.63) ads suggest that ineffective ads the potential benefits of taking an inductive approach. Using market-
might elicit a greater degree of cognitive branching. Activation differ- driven data to designate “successful” and “unsuccessful” product adver-
ences are also significant in the left hemisphere of BA45 (F1, 23 = tisements offers new insights and directions for theoretical exploration
6.60, p b .05), a region of the brain that is believed to retrieve and eval- by using dEEG as a diagnostic tool. In particular, the observation of sig-
uate semantic information, as well as in the left hemisphere of BA47 (F1, nificant differences in brain wave activity as participants engage with
23 = 4.66, p b .05), which is also associated with semantic processing successful, as opposed to unsuccessful, ads is localized to specific re-
(e.g., Chou, Booth, Bitan, Burman, Bigio, Cone, Lu, & Cao, 2006). Consid- gions of the brain whose function and purpose are already supported
ered together, these mean differences suggest that components of suc- by existing neurological research. Utilizing dEEG measurement accord-
cessful ads (m45L succeed = 5.72, m47L succeed = 1.02) require less ing to originally intended purposes (diagnostic) results in a shift in focus
effort to process and evaluate than components of unsuccessful ads from attempting to predict ad success on the basis of neurological differ-
(m45L unsuccessful = 6.14, m47L unsuccessful = 1.15). ences to describing the characteristics as consumers actively process
In summary, significant activation differences were ascertained them. Accordingly, this research is the first to link the success or failure
between successful and unsuccessful direct response ads with these of direct response advertising to areas of the brain associated with un-
differences localized to specialized areas of the brain. Tracing these dif- certainty reduction, working memory, and semantic processing.
ferences in neural activation to specific sources allows for novel theories
and hypotheses regarding possible reasons for advertisement effective-
ness, or ineffectiveness, to be interpreted. In this particular case, episod- 4.1. Limitations
ic memory retrieval appears to be easier for images associated with
successful as opposed to unsuccessful ads. Despite the intriguing nature of an inductive paradigm, a number of
limitations ought to be noted as they could facilitate further exploration.
4. Discussion First, this article reports on a single applied example, which relies on a
dataset produced from a single set of direct response product advertise-
Taking a “research in reverse” approach, whereby known results and ments. What remains to be seen is whether a larger design consisting of
their implications are subjected to research methodologies for the pur- a broader pool of “successful” and “unsuccessful” ads selected on the
pose of hypothesis generation and theory building, is a worthwhile av- basis of varying levels of usage/familiarity, or from additional product
enue for consumer research in particular and marketing research as a categories (or brand focused), would yield a richer and/or more nu-
whole. Such a paradigm defies conventional deductive research logic, anced set of observed differences in brain wave activity. Surprisingly,
which recommends proceeding from hypotheses and theory to generat- when diagnostic tools such as dEEG are applied in more traditional
ing a sufficient test of that theory to exploring implications and bound- medical contexts, measurable phenomena (e.g., epilepsy) manifest
ary conditions of obtained results, consistent with the scientific method. themselves quite consistently across patients. This would seem to sug-
Historically speaking, marketing scholars have long been motivated gest that observed differences in brain wave activity would be similar
to substantiate the field of marketing as a scientific discipline to what is reported in this study, even in the presence of a broader
(Anderson, 1983). As Anderson (1983) notes, this has led to the sampling of products and ad types. Nevertheless, future studies ought
adoption of two primary forms of inquiry; logical empiricism and to be undertaken in an effort to scrutinize these findings and whether
falsification. A shared feature of these two paradigms is their basis in a or not they are replicable.
method of scientific inquiry whereby theory generation precedes Another suggestion made in this article is that an inductive ap-
hypothesis development, which precedes empirical verification. proach, although heretical, could be applied with similar benefits to
Deshpande (1983) asserts that this over-reliance on the scientific meth- other research methodologies within and beyond the purview of CNS.
od has obscured attention towards alternative paradigms that could However, no direct empirical evidence is provided to support this
give rise to the development of novel theory. He goes on to suggest claim, which suggests two directions for future research. First, addition-
that primarily inductive approaches could contribute to the literature al studies are needed to determine whether a “research in reverse” ap-
by generating novel insights on the basis of what Glaser and Strauss proach is beneficial when applied to other types of CNS methodologies,
(1967) refer to as “grounded events.” Few have since chosen to heed such as functional magnetic resonance imagery (fMRI), positron emis-
this promising suggestion. Indeed, a recent content analysis of research sion tomography (PET), and or other physiological measurement
published in prominent marketing journals between 1990-2009 (Davis, (e.g., eye-tracking). Second, the utility of a “research in reverse” ap-
Golicic, Boerstler, Choi, & Oh, 2013) illustrates a trend of decreasing proach should be tested with a broader array of research methodolo-
methodological diversity despite the growing complexity of marketing gies. Although data collection methods are commonly used in the
phenomena. The authors attribute this to an overreliance on the service of testing hypotheses and/or products, they can also be used in
scientific method and suggest that scholars consider alternative para- an attempt to diagnose and understand the known success or failure
digms that account for aspects, such as contextual realism, which is ac- of business-related interventions beyond common qualitative methods,
knowledged here within a research-in-reverse paradigm. An applied as this study demonstrates.
T. Daugherty et al. / Journal of Business Research 69 (2016) 3168–3176 3175

Finally, small sample sizes, such as used in this study, are typical of success in this realm, attempts to do similar work that proceeds from
CNS research due to the remarkable consistency of brain wave activity known and quantifiable results are notably lacking. Given the vast
differences observed across participants. Additionally, the average age amount of archival data that exists (i.e., Big Data) both for public con-
of the sample was relatively high, and although this is consistent with sumption and within organizations, a reverse research methodology
the targeted age demographic for direct marketing advertisements, a could serve as a catalyst for significant theoretical advances through
more representative age range could further enrich the generalizability open research collaboration. Existing theoretical frameworks could
of the results. Medical applications of dEEG and related technologies also be scrutinized in terms of proposed inductive methodology. Speci-
have benefitted from sample sizes in the millions through the culmina- fying a theory and developing a specific set of measures and analyses
tion of numerous studies as a means of increasing the diagnostic preci- around that theory is one step, however a far more challenging under-
sion of these tools and the people using them. Additional research is taking to discern whether the elements of an established theory emerge
recommended that explores the effects observed in this study among from real-world outcome data and business decisions. An argument
consumers of varying age groups, geographical locations, and other de- commonly made in opposition for starting with real-world data is that
mographic features in an effort to confirm that these findings generalize such data lack the necessary metrics (i.e., valid and reliable measures
to a broader consumer population. of relevant constructs) to enable a true test of the theory in question.
However, the paradigm presents a mediating phase, where known re-
5. Managerial implications sults can be subjected to additional testing, such as dense-array EEG,
which allows for the usage of scientifically-acceptable measurements
Taking an inductive approach towards research implies that there is of pertinent constructs.
more to be discovered by managers after the results and implications of An optimal paradigm might be one where inductive based research
a marketing strategy (e.g., a product advertisement) are known. Beyond is used as a form of reciprocal dialog. That is, findings gleaned from a
deciding whether or not to continue with a chosen strategy, practi- study such as this could then be subjected to a more traditional
tioners can use the approach outlined here to determine the basis for hypothesis-based investigation. Observations from this research could
success or non-success with regard to specific tactics. Gaining insight then be used as feedback to inform future “research in reverse” applica-
into why something worked (or failed to work) can serve as an invalu- tions, which could then provide an impetus for imminent studies
able source of information as efforts are undertaken to develop subse- employing the scientific method and so forth. In this way, the undeni-
quent strategies. Consumer advertising is frequently tested prior to able value of the scientific method continues to be acknowledged, as
making a substantial investment in media spending to reach a broader is the fine line that exists between heresy and apostasy.
market audience. The approach advanced in this article implies that
also conducting follow-up testing using research methodologies as
diagnostic tools may enable organizations to maximize their return- 7. Conclusion
on-investment.
Research in marketing, as in many fields, has long been plagued with Although the research reviewed here is limited to a single and fairly
a scholar–practitioner divide (see Rossiter & Percy, 2013, for a recent specialized (i.e., dense array EEG) study, an inductive “research in re-
discussion of this issue). As such, scholars frequently lament a lack of verse” approach warrants additional focus and consideration. The argu-
theory-driven practice, while practitioners perceive a growing discon- ment for starting with known results and working towards greater
nect between the research conducted in academia and the real world. theoretical understanding has been based on the traditional usage of
The approach taken here suggests a novel and perhaps heretical physiological and neurological measurement tools, but such an ap-
means of bridging this gap. Existing real-world data, rather than proach could be generalized to more traditional marketing research
academically-derived hypotheses, could serve as the knowledge-base methodologies with equal benefit. At a minimum, conducting “research
from which research could assist in the generation of useful and practi- in reverse” assists in bridging the persistent divide between theory and
cal theories or models. In keeping with the notion of an inductive ap- practice. At best, an inductive approach is poised to add a substantial
proach, practice would give rise to theory rather than theory giving amount of further insight and depth to CNS and throughout the various
rise to practice (heresy!). fields and subfields of marketing research. Similar things were said of
In the grand scheme of social science, the study of CNS is ultimately the scientific method, which has come to inform the vast majority of
in its infancy. While neuroscience researchers continue to increase the marketing research, in a time before the paradigm rose to prominence
understanding of the brain, a significant amount of hard work lies and was viewed as “heresy” within the scientific community.
ahead for those paving the way in CSN. Nevertheless, the potential to
measure conscious and unconscious brain activity in ad testing cannot
References
be ignored as the opportunity to better understand message processing
(attention, capacity, encoding, and retrieval), cognitive workload, and Advertising Research Foundation (2013). Neurostandards collaboration project [white
emotional engagement (affinity, loyalty, advocacy) is of great value to paper]. Retrived from www.thearf.org
Allan, G. (2003). A critique of using grounded theory as a research method. Electronic
marketing researchers. The notion of the ever elusive “buy button” in Journal of Business Research Methods, 2(1), 1–10.
the brain may be hyperbole, but the scientific pursuit to utilize CNS to Ambler, T., Ioannides, A., & Rose, S. (2000). Brands on the brain: Neuro-images of adver-
observe human thought processes is not. tising. Business Strategy Review, 11(3), 17–30.
Anderson, P. F. (1983). Marketing, scientific progress, and scientific method. Journal of
Marketing, 47(4), 18–31.
6. Theoretical implications Arieli, A., Sterkin, A., Grinvald, A., & Aertson, A. (1996). Dynamics of ongoing activity: Ex-
planation of the large variability in evoked cortical responses. Science, 27, 1868–1871.
Ariely, D., & Berns, G. S. (2010). Neuromarketing: The hope and hype of Neuromarketing
The ability of inductive research to promote the development of in business. Nature Reviews Neuroscience, 11(4), 284–292.
novel and interesting theories on the basis of observed phenomena is Bland, A. R., & Schaefer, A. (2011). Electrophysiological correlates of decision making
a central theme here. The theory and findings advanced in this article il- under varying levels of uncertainty. Brain research, 1417, 55–66.
Bonoma, T. (1985). Case research in marketing: Opportunities, problems, and a process.
lustrate the potential for “research in reverse” to generate results that
Journal of Marketing Research, 22(2), 199–208.
may not be intuitive, enriching the existing theory base and perhaps Brodmann, K. (1909). Vergleichende Lokalisationslehre der Groshirnrinde. Leipzig: Barth.
even challenging common practices. As mentioned previously, this is Chou, T. L., Booth, J. R., Bitan, T., Burman, D. D., Bigio, J. D., Cone, N. E., ... Cao, F. (2006). De-
frequently done with qualitative data using methodological techniques velopmental and skill effects on the neural correlates of semantic processing to visu-
ally presented words. Human brain mapping, 27(11), 915–924.
such as grounded theory (Strauss & Corbin, 1994) to proceed from Corsi-Cabrera, M., Galindo-Vilchis, L., Del-Rio Portilla, Y., Arce, C., & Ramos-Loyo, J. (2007).
participant- generated observations to systematic frameworks. Despite Within-subject reliability and inter-session stability of EEG power and coherent
3176 T. Daugherty et al. / Journal of Business Research 69 (2016) 3168–3176

activity in women evaluated monthly over nine months. Clinical Neurophysiology, Parson, A. (2011). Neuromarketing: Prove thyself & protect consumers [briefing paper].
118(1), 9–21. Retrieved from www.dana.org/News/Details.aspx?id=43519
Davis, D. F., Golicic, S. L., Boerstler, C. N., Choi, S., & Oh, H. (2013). Does marketing research Plassmann, H., Ramsoy, T. C., & Milosavljevic, M. (2012). Branding the brain: A critical re-
suffer from methods myopia? Journal of Business Research, 66(9), 1245–1250. view and outlook. Journal of Consumer Psychology, 22, 18–36.
Deshpande, R. (1983). “Paradigms lost”: On theory and method in research in marketing. Pradeep, A. K. (2010). The buying brain: Secrets for selling to the subconscious mind. New
Journal of Marketing, 47(4), 101–110. York, NY: Wiley Publishers.
Fallani, F. D. V., Astolfi, L., Cincotti, F., Mattia, D., Marciani, M. G., Gao, S., ... Babiloni, F. Reason, G. (2012). Neuromarketing is dead: Long live neuromarketing [blog post]. Re-
(2008). Structure of the cortical networks during successful memory encoding in trieved from www.essencedigital.com/blog/2012/11/neuromarketing-is-dead-long-
TV commercials. Clinical Neurophysiology, 119, 2231–2237. live-neuromarketing
Fisher, C. E., Chin, L., & Klitzman, R. (2010). Defining neuromarketing: Practices and pro- Reimann, M., Zaichkowsky, J., Neuhaus, C., Bender, T., & Weber, B. (2010). Aesthetic pack-
fessional challenges. Harvard Review of Psychiatry, 18(4), 230–237. age design: A behavioral, neural, and psychological investigation. Journal of Consumer
Fugate, D. L. (2007). Neuromarketing: A layman's look at neuroscience and its potential Psychology, 20, 431–441.
application to marketing practice. Journal of Consumer Marketing, 24(7), 385–394. Rossiter, J. R., & Percy, L. (2013). How the roles of advertising merely appear to have
Glaser, B., & Strauss, A. (1967). The discovery of grounded theory. Chicago, IL: Aldine. changed. International Journal of Advertising, 32(3), 391–398.
Haig, B. D. (1995). Grounded theory as scientific method. Philosophy of Education, 28(1), Rothschild, M. L., & Hyun, Y. J. (1990). Predicting memory for components of TV commer-
1–11. cials from EEG. Journal of Consumer Research, 16, 472–478.
Hodson, D. (1996). Laboratory work as scientific method: Three decades of confusion and Schaefer, M., & Rotte, M. (2007). Thinking on luxury or pragmatic brand products: Brain
distortion. Journal of Curriculum Studies, 28(2), 115–135. responses to different categories of culturally based brands. Brain Research, 1165,
Hsu, F. C., Garside, M. J., Massey, A. E., & McAllister-Williams, R. H. (2003). Effects of a sin- 98–104.
gle dose of cortisol on the neural correlates of episodic memory and error processing Stewart, A. J., Pickering, M. J., & Sturt, P. (2004). Using eye movements during reading as
in healthy volunteers. Psychopharmacology, 167(4), 431–442. an implicit measure of the acceptability of brand extensions. Applied Cognitive
Hubert, M. (2010). Does neuroeconomics give new impetus to economic and consumer Psychology, 18, 696–709.
research? Journal of Economic Psychology, 31, 812–817. Strauss, A., & Corbin, J. (1994). Grounded theory methodology. Handbook of qualitative re-
Hudson, L. A., & Ozanne, J. L. (1988). Alternative ways of seeking knowledge in consumer search (pp. 273–285).
research. Journal of Consumer Research, 14(4), 508–521. Treleaven-Hassard, S., Gold, J., Bellman, S., Schweda, A., Ciorciari, J., Critchley, C., & Varan,
Kato, J., Ide, H., Kabashima, I., Kadota, H., Takano, K., & Kansaku, K. (2009). Neural corre- D. (2010). Using the P3a to gauge automatic attention to interactive television adver-
lates of attitude change following positive and negative advertisements. Frontiers in tising. Journal of Economic Psychology, 31, 777–784.
Behavioral Neuroscience, 3, 1–13. Vecchiato, G., Astolfi, L., Fallani, F. D. V., Cincotti, F., Mattia, D., Salinari, S., ... Babiloni, F.
Kenning, P., & Linzmajer, M. (2011). Consumer neuroscience: An overview of an emerg- (2010a). Changes in brain activity during the observation of TV commercials by
ing discipline with implications for consumer policy. Journal of Consumer Protection using EEG, GSR, and HR measurements. Brain Topography, 23(2), 165–179.
and Food Safety, 6, 111–125. Vecchiato, G., Toppi, J., Astolfi, L., Fallani, F. D. V., Cincotti, F., Mattia, D., ... Babiloni, F.
Koechlin, E., & Hyafil, A. (2007). Anterior prefrontal function and the limits of human de- (2011). Spectral EEG frontal asymmetries correlate with the experienced pleasant-
cision-making. Science, 318(5850), 594–598. ness of TV commercial advertisements. Medical & Biological Engineering &
Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (2008). International affective picture system Computing, 49, 579–583.
(IAPS): Affective ratings of pictures and instruction manual. Technical report A-8. Volz, K. G., Schubotz, R. I., & von Cramon, D. Y. (2005). Variants of uncertainty in decision-
Lawrence, J. R. (1982). To hypothesize or not to hypothesize? The ‘correct’ approach to making and their neural correlates. Brain research bulletin, 67(5), 403–412.
survey research. Journal of the Market Research Society, 24(4), 335–343. Wang, Y.J., & Minor, M.S. (2008). Validity, reliability, and applicability of psychophysiolog-
Lee, N., Broderick, A. J., & Chamberlain, L. (2007). What is ‘neuromarketing’? A discussion ical techniques in marketing research. Psychology and Marketing, 25(2), 197–232.
and agenda for future research. International Journal of Psychophysiology, 63, Wilson, R.M., Gaines, J., & Hill, R.P. (2008). Neuromarketing and consumer free will.
199–204. Journal of Consumer Affairs, 42(3), 389–410.
Lepage, M., Ghaffar, O., Nyberg, L., & Tulving, E. (2000). Prefrontal cortex and episodic Young, C.E., & Robinson, M. (1987). Guidelines: Tracking the commercial viewers wan-
memory retrieval mode. Proceedings of the National Academy of Sciences, 97(1), dering attention. Journal of Advertising Research, 27(3), 15–22.
506–511. Zhang, J. X., Leung, H. C., & Johnson, M. K. (2003). Frontal activations associated with
Murphy, E. R., Illes, J., & Reiner, P. B. (2008). Neuroethics of neuromarketing. Journal of accessing and evaluating information in working memory: an fMRI study.
Consumer Behaviour, 7, 293–302. Neuroimage, 20(3), 1531–1539.

Potrebbero piacerti anche